nuffield bursary 19
Post on 07-Apr-2018
228 Views
Preview:
TRANSCRIPT
-
8/4/2019 Nuffield Bursary 19
1/131
Nuffield Bursary
Step one in the Nuffield bursary project
What is my report on....?
A project should allow you scope to use your initiative. The criteria for selectingprojects are that they should:
have clear scientific or technological content; contribute to the work of the host organisation; be well defined, having a clear outcome in mind from the beginning; give students a chance to work alongside practising scientists; allow scope for initiative on the part of the student.
Initial Ideas-Comparing different medical imagings weakness and flaws-Looking at the contrast mediums, (what new mediums are being processed)
Planning Process
Initial Targets for Tuesday 26th July-decide on a project idea-understand more about SPECT process
Statistics Side
Spotfire- company which help you analyse and explore data, using differentgraphs, simple, predictive applications.
Username: GEHEAL-SILVER-80870Password: 4S3N6USR5!
Brunettebimbo details
Username: GEHEVL -79445Password: TYCO6QW3Q!
1- Gene expression clusteringhttp://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Expression%20Clustering%20in%20Heat
%20Map&waid=ffa011023552264eab9cf-ed69Hierarchical Clustering = clustering like bits of data
2- Pharmaceutical companies projectshttp://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R%26D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69Bar charts to compare amount of each company (discrete) sub sectionedinto phasesColour for phases accompanied by a key
3- Pharmaceutical companies area of researchhttp://ondemand.spotfire.com/Public/ViewAnalysis.aspx?
file=Public/Pharma%20R%26D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69Using area = %, colour and word labelling to distinguish between
http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Expression%20Clustering%20in%20Heat%20Map&waid=ffa011023552264eab9cf-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Expression%20Clustering%20in%20Heat%20Map&waid=ffa011023552264eab9cf-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Expression%20Clustering%20in%20Heat%20Map&waid=ffa011023552264eab9cf-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Expression%20Clustering%20in%20Heat%20Map&waid=ffa011023552264eab9cf-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Expression%20Clustering%20in%20Heat%20Map&waid=ffa011023552264eab9cf-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Expression%20Clustering%20in%20Heat%20Map&waid=ffa011023552264eab9cf-ed69 -
8/4/2019 Nuffield Bursary 19
2/131
4- Area of medicine, how much there was in each area and what phasehttp://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R%26D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69Check boxes so you could compare only certain phases for example phase
1 of all areas5- Area of research and diseasehttp://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R%26D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69Used a tree map: displays data hieratically by using nesting triangles, sosplitting a box into rectangles proportional to size of dataEach branch (area of research is given a triangle)Advantage: good use of space, colouring helps identify patternsIn this case larger squares (therapeutic areas TA) = number of drugs inthat TAShading = more drugs in phase |||
See which drugs are particularly progressing (metabolic type 2 diabetes)6- How many drugs each company has in each area of research
http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R%26D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69Basic table of company and number of drugs per areas of researchLots of informationSee which area has the most research (Oncology) and which companydoes the most (GSK)
7- Comparing 1 company to all the othershttp://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R%26D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69See which company is most successful, who the main competitors are
8- Cell death http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/High%20Content%20Screening%20For%20Toxicity&waid=fda1267de7c209ea15b19-ed69Uses colour shape and size to show cell death, clicking on icons to workout % of cell death
9- Mechanism of cellhttp://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/High%20Content%20Screening%20For%20Toxicity&waid=fda1267de7c209ea15b19-ed69
Displays controls as well side my side, so all products can be comparedPie charts with colouring is used to compare things, in lines so differentfactors can be compared
PET expansion: new radiotracers, improvement in instruments, improvement inimage analysis
In Vivo: experiment using a living organism
Factors affect reliability of medical imaging
1) A lot trial and error as risk can be predicted but not success. Even if
radiotracer meets all requirements, chemical and pharmaceuticalproperties dont mean the radiotracer will be successful.
http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/High%20Content%20Screening%20For%20Toxicity&waid=fda1267de7c209ea15b19-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/High%20Content%20Screening%20For%20Toxicity&waid=fda1267de7c209ea15b19-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/High%20Content%20Screening%20For%20Toxicity&waid=fda1267de7c209ea15b19-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/High%20Content%20Screening%20For%20Toxicity&waid=fda1267de7c209ea15b19-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/High%20Content%20Screening%20For%20Toxicity&waid=fda1267de7c209ea15b19-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/High%20Content%20Screening%20For%20Toxicity&waid=fda1267de7c209ea15b19-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/Pharma%20R&D%20Pipelines&waid=cb7db8464bf4fb8a911fa-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/High%20Content%20Screening%20For%20Toxicity&waid=fda1267de7c209ea15b19-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/High%20Content%20Screening%20For%20Toxicity&waid=fda1267de7c209ea15b19-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/High%20Content%20Screening%20For%20Toxicity&waid=fda1267de7c209ea15b19-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/High%20Content%20Screening%20For%20Toxicity&waid=fda1267de7c209ea15b19-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/High%20Content%20Screening%20For%20Toxicity&waid=fda1267de7c209ea15b19-ed69http://ondemand.spotfire.com/Public/ViewAnalysis.aspx?file=Public/High%20Content%20Screening%20For%20Toxicity&waid=fda1267de7c209ea15b19-ed69 -
8/4/2019 Nuffield Bursary 19
3/131
Successful Radiopharmaceutical qualities:1) Biological Targets2) Radio synthetic considerations3) In vitro/in vivo characteristics of ligand to be radiolabeled (including
affinity, protein binding, lipophilicity, metabolism, etc.)4) Toxicology and radiation dosimetry5) Species differences in (3) and (4) above
Protein -Protein interaction: 2 or more proteins bind together to carry out afunction, mapping follows this
Structure of protein determined by-nuclear magnetic resonance (NMR): when magnetic nuclei in a magnetic fieldabsorb and re emit electromagnetic radiation-spectroscopy
ADME-pharmacokinetics: drug-drug interactions, absorption, distribution,metabolism, excretion
Blood Brain Barrier (BBB): separation of circulating blood and brain extracellularfluid in the central nervous system
PET benefits-detect Alzheimers earlier allowing safer and more effective therapies to beused-early detection of cancers-non invasive
-safe-cost effective
-
8/4/2019 Nuffield Bursary 19
4/131
-can assess dose related ADME properties-used in oncology-used in cardiology-can administer radiotracer several times in the space of a short amount of time
Michaelis Menten Kinetics-made by 2 biochemist called Michaelis + Menten-its the most simple + best known model for enzymes-involves equation to explain the rate of enzyme reactionsV = V max [s]/ Km + [s]V max = maximum rate achieved by the systemKm = a constant, is the substrate constant when the reaction rate is V max[s] = concentration of the substrateV= reaction rate
-Limitations: law of mass action, to do with free diffusion (true in liquids)however cytoplasm acts more like a gel then a liquid.
K- means clustering-puts values into clusters with the nearest mean-attempts to find centres of natural clusters-several random points found-clusters are made by associating points to the nearest mean-centre of each cluster becomes the new mean-step 2 + 3 repeated until clusters converge-needs line graph to do K-means clustering
Parallel Coordinates-way of visualizing data with more than one statisticalvariable-points represented by lines joined together (poly line) with vertical lines onparallel axes-each row in a table is represented by a line-e.g. a line represents an apple and different points show how muchfructose/glucose/sucrose are in the apple, the points are joined up to show thedistribution. Straight line = equal distribution.
-values always normalised, so fruit with highest fructose = 100%, fruit withlowest sucrose = 0%
-scale of glucose column is a different scale than the fructose column, so notcomparable as columns could have different units (grams, concentration)
-
8/4/2019 Nuffield Bursary 19
5/131
-you can see which TV is the best, maybe orange as it is medium size, best type,medium warranty, second highest cost compared to pink which has the same
screen size, but a worst type + warranty however its the cheapest!-could use this on different radiopharmaceuticals and have different variables(half life, blood brain barrier, affinity)
Heat Maps-like a gradient, bright red = max-can be combined with hierarchy clustering to form a dendrogram, so if test ADEare similar cluster them together-can cluster both rows and columns not just by putting them next to each otherbut putting a grouping hug line
-Dendrograms, dendro=tree as lines look like a tree map
3/08/11 aims
-to finish amide mouse study-to conquer K-means clustering-to love another statistic software
4/08/11 aims-use amide in a fully body scan-use amide tools smoothing and filtering-write diary 4 yesterday
5/08/11 aims-finish hierchal clustering-find out what the two variables mean
-parallel co-ordinate plot
-
8/4/2019 Nuffield Bursary 19
6/131
Project 1: Statistic analysis of biological data
Objectives-use as many statistic tools as possible
-compare statics programs-tackle an ethical issue, considering both sides, and come to your ownconclusion.
K-means clusteringMy project supervisor Sven told me about K-means clustering, which was astatistical tool used to cluster data. There was an automatic tool called K-meansclustering in Spotfire, but before I used it I wanted to understand the process andbenefits of the tool. Initially I found the process for K- means clusteringis asfollows; choose several random points, form clusters by associating points to thenearest mean, find the centre of each cluster which becomes the new mean and
the previous two steps are repeated until clusters converge. [1]. Now Iunderstood the process I wanted to figure out why its useful. (INSERT why K-means clustering is useful).
I then went on to use the K-meansclustering tool in Spotfire, and decidedto set up a dummy set of data with twodistinguished clusters.
When I opened my data into Spotfire it automatically plotted it as a scattergraph.
-
8/4/2019 Nuffield Bursary 19
7/131
So when I tried to use the K-means clustering tool it said There is no valid linechart available, more information needed for K-means clustering. I decided Ineeded to know why a line chart was needed for K-means clustering. Idiscovered that the K-means clustering algorithm used in Spotfire involves linechart visualation.[2] I then used to new line chart tool to construct a line graph ofmy data;
and used to K-means clustering tool from that. Nothing happened to the line
chart or the scatter graph so I tried changing the X axis of the line chart to K-means clustering which gave me this;
-
8/4/2019 Nuffield Bursary 19
8/131
Which showed that two values had a hair length of 12cm and that the other hairlengths only had one value. In my mind this was not the purpose of the K-meansclustering tool. I then produced a heat map of my data;
As you can see the K-means clustering was all average, and all gave a value of 1.On research one reason of poor K-means clustering results was an incorrectnumber of clusters. The rule of thumb number of clusters = (number of piecesof data/2)[3]. So I my case I had 9 pieces of data, so 9/2 = 4.5 and the squareroot of 4.5 is 2.12 (3sf). Coincidently I has already been using 2 clusters whentrying the K-means clustering tool. So this was not the problem of my poor K-means clustering results. I clicked on the help box and then on to K-means
clustering and read through the instructions. One step I had missed out was tosplit the lines into at least one column so multiple lines are created. I readthrough the instructions and once implementing them my line graph appearedlike this.
-
8/4/2019 Nuffield Bursary 19
9/131
I then re-set the K-means clustering tool and produced a heat graph to find out
the value of K-means clustering.
Dissapointingly all the K-means clustering values came up as #NA, which wasabsolutely unhelpful. I began to wonder whether K-means clustering wassupposed to have a value. After checking with my project supervisor they werenot. I then though that because data is already clustered the K-means clusteringmay just not change the data, so I decided to try some less clustered data. Iopened a data set that my project supervisor provided me with.
Column 1 Column 2 Column 3
1 0.56743519 0.60011442
1 -0.66558438 1.6899974
1 1.1253323 1.8156223
1 1.2876764 1.7119083
1 -0.14647135 2.2902498
1 2.1909155 1.6686005
1 2.1891642 2.1908381
-
8/4/2019 Nuffield Bursary 19
10/131
1 0.96236672 -0.20245711
1 1.3272924 0.98021044
1 1.1746391 0.8432827
1 0.81329142 -0.60408556
1 1.7257905 1.2573042
1 0.41168346 -0.056472928
1 3.1831858 2.4151415
1 0.86360412 0.1949096
1 1.1139313 1.528743
1 2.0667682 1.2193207
1 1.0592815 0.078098376
1 0.90435159 -1.1706745
1 0.16765054 0.94081218
1 1.2944108 -0.010633706
1 -0.33618186 1.614463
1 1.7143246 1.5077408
1 2.6235621 2.6924299
1 0.3082243 1.5912826
1 1.8579967 0.3564048
1 2.2540014 1.3803373
1 -0.59372958 -0.009115524
1 -0.44096443 0.98048933
1 1.5711476 0.95177921
2 4.31918E-05 -2.6101195
2 -0.31785945 -2.9120129
2 1.0950037 -3.6354652
2 -1.8739903 -3.5595733
2 0.42818327 -2.5563465
2 0.89563847 -3.9499038
2 0.73095734 -2.2188184
2 0.57785735 -2.4310394
2 0.040314032 -3.8217143
2 0.67708919 -3.2656069
2 0.56890021 -4.187777
2 -0.25564542 -5.2023207
-
8/4/2019 Nuffield Bursary 19
11/131
2 -0.37746896 -2.0136626
2 -0.29588711 -3.5186351
2 -1.4751345 -2.6726324
2 -0.23400405 -2.765943
2 0.11844484 -2.9785339
2 0.31480904 -4.0039445
2 1.4435082 -3.9471461
2 -0.35097474 -3.3744292
2 0.62323385 -4.1858862
2 0.79904862 -4.0559029
2 0.94088994 -1.5275201
2 -0.99209174 -2.9442562
2 0.21203515 -4.2173175
2 0.23788207 -3.0412271
2 -1.0077634 -4.1283439
2 -0.74204475 -4.3492775
2 1.082295 -3.2611016
2 -0.1314997 -2.0465346
Spotfire automatically plotted it in a scatter graph again.
When I tried to colour the clusters it came up with this as a scatter graph.
-
8/4/2019 Nuffield Bursary 19
12/131
This implies that there is only one cluster, which I believe isnt true as I thinktheres two reasonably separate clumps of data, which I have highlighted in thetwo black circles.
Amide
My project supervisor, Sven, introduced me to an imaging program; Amide.
-
8/4/2019 Nuffield Bursary 19
13/131
I began comparing images of mice pre-treatment and post treatment. Sventaught me how to load the SPECT image;
This was pretty straight forward. He then showed me how to load the CT image
which more complicated as it involved having to input voxel size, which directionthe image was scanned and other voxel dimensions. Here is the CT image;
-
8/4/2019 Nuffield Bursary 19
14/131
Sven also showed me other useful functions of Amide, including scrollingeffectively through the mouse, seeing the different images on the mice differentplanes, changing the contrast of the images helping you identify soft tissue orbones and overlapping the two images. The overlap of the CT and SPECT scan
produced the image below.
(Find uses of overlapping) . Sven then showed me how to highlight parts ofthe mouse, in particular the tumour and the whole body.
Labelling of the tumour;
-
8/4/2019 Nuffield Bursary 19
15/131
Labelling of the whole body;
-
8/4/2019 Nuffield Bursary 19
16/131
Amide also had some more features that my project supervisor told me to lookover and use. One is filtering; I clicked on the tools and used the tool FilterActive Data Set which first gave me three filters to choose between Gaussian(smoothing filter), median linear(preserves edges whilst removing speckled noisein 1D), median 3-D (preserves edges whilst removing speckled noise in 3D). Ichoose median 3-D, and then it asked to enter a kernel size which as the filter is3D will become the size of the side of a cube in which the median is taken from.
The median of that cube is then applied to all the pixels within that cube,smoothing the image. I started with 5 as the kernel size. As you can see from theimage below this remove some information from the image and smoothed it. Ithen went on to try an extreme filter which has a kernel size of 11, which moreextremely smoothed the image. The reason why someone would want to smoothan image would be to make objects more definite and remove noise.
-
8/4/2019 Nuffield Bursary 19
17/131
Sven then showed me how the mean/medium voxel uptake and the number ofvoxels could be calculated using the tool Calculating ROI statistics. Howeverevery time we tried to calculate ROI statistics the program would crash and errorbox would occur. My supervisor had to continue on with his own projects and so Ihad to figure out the error and how to overcome it. I did some research on theinternet for GLib-ERROR and on a computing forum someone suggested it wasbecause the computer was running out of RAM. So I installed an extra storage
device. The hard drive failed to help, so I tried individually calculating the ROIstatistics of the tumour and then the body. The program calculated the tumourstatistics, however the same error message came up when I tried to calculatethe ROI statistics of the body. I came to the conclusion because the body was alot bigger than the tumour the computer was having trouble processing thatmuch information. For the time being I decided to just focus on the tumours.
2 days laterAfter trying out different options on Amide I found that the program would crashwhen I calculated the ROI statistics of the whole body using a CT image as well.
The CT image contains a lot of data and so to prevent the program from crashingI decided to try to calculate ROI statistics of the body using just the SPECTimage. To my joy, it didnt crash. I decided to calculate the ROI statistics of thebody for same animal pre and post dosage to see how similar the size of bodycalculated was and thus how accurate my circling of the body is, as in theorythey should be the same size if you minus away the value of the tumour whichchanges in size due to the drug Duramycin.
Animal pre-dose body size (mm3)Whole body size:34712.30 Tumour size: 1221.55 Whole body excludingtumour: 33490.75
Animal post-dose body size(mm3)
Whole body size:45313.90 Tumour size:543.22 Whole body excludingtumour: 44770.67
-
8/4/2019 Nuffield Bursary 19
18/131
As you can see the difference between the mouses whole body excluding thetumour is very different (11279.92 mm3) pre and post which is due to myinaccurate circling of the body. My challenge is to receive some advice from myproject supervisor and practise using amide to get a more precise reading.
Animal Pre-Dose tumour statistics
Here I a mouse pre-dose statistics, its the same mouse whose images I havebeen using in the two previous pages to illustrate how Amide works. The table isexplained by my supervisor as follows; frame/gate are time frames andprocesses that we can disregard , duration is the time in seconds the mouse was
scanned for, midpoint is the midpoint of the time in seconds,median/mean/variation/stddev/ min/ max is the median/mean/variation/standard deviation/ minimum/maximum uptake of Duramycin pervoxel, size is the size of the tumour in millimetres, frac voxels is number ofvoxels in the tumour where a fraction of them (e.g. ) was within the tumour,voxels is the number of whole voxels in the tumour.
I decided to go through the same process to get the ROI statistics for the Animalpost-dose tumour to compare. The dose the animal received was the drugDuramycin. (Insert Research on Duramycin).
Animal Post-Dose tumour statistics
Fram
e
Durati
on (s)
Midp
t (s)
Gat
e
M
edian
M
ean
Var StdDe
v
Min
Max
Size
(mm^
3)
Frac.
Voxels
V
oxels
0
1.
000
0.50
0
0
286
28
3.839
61
06.09
78.
1415
-72
8
588
54
3.224
48680
3.58
5
0435
2
As you can see the frame, duration, midpt, gate and medium are the same whichis expected. The mean is greater so the uptake of the tumour cancer cells isgreater which may be good and they uptake more the Duramycin. Thevariation/standard deviation are smaller, which indicates more precise results.
The size of the tumour, number of fraction voxels and number voxels hasdecreased which indicates Duramycin is effective. Although the graphs display alot of raw information I wanted to explore using statistics programs to comparethe effect of Duramycin.
Im going to compare using Excel to plot a graph and Spotfire to plot a graph.Firstly I need to bits of data, and will be using the same data from the Duramycin
Fram
e
Durati
on (s)
Midp
t (s)
Gat
e
M
edian
Mean
Var StdDe
v
Min
Max
Size
(mm^
3)
Frac.
Voxels
Vo
xels
0
1.
000
0.50
0
0
268
27
2.368
95
237.6
30
8.606
-12
87
369
8
12
21.55
109468
1.11
112
4195
Fram
e
Durati
on (s)
Midp
t (s)
Gat
e
M
edian
Mean
Var StdDe
v
Min
Max
Size
(mm^
3)
Frac.
Voxels
Vo
xels
0
1.
000
0.50
0
0
268
27
2.368
95
237.6
30
8.606
-12
87
369
8
12
21.55
109468
1.11
112
4195
-
8/4/2019 Nuffield Bursary 19
19/131
trial. The two tables above show the details of one mouse in the Duramycin trial,however 3 more were used so I will construct two tables using the data from all 4mice. One table will compare the tumour size pre and post dose and the othertable will compare the number of whole voxels before and after.
Animals Tumour Size Pre/Post Duramycin
Animal Number Tumour Size preDuramycin (Size mm^3)
Tumour Size postDuramycin (Size mm^3)
1 1221.550 543.2242 1070.390 688.9053 805.974 516.1554 1149.430 365.952Animals Number of Voxels Pre/Post Duramycin
Animal Number Number of Voxels preDuramycin
Number of Voxels postDuramycin
1 1124195.000 504352.0002 987027.000 637940.0003 747283.000 480780.0004 1059559.000 341057.000
I started by using Excel to draw the graph for the table Animals Tumour SizePre/Post Duramycin and will assess it on 2 criteria; user friendliness and finalgraph outcome. I loaded my data into excel and then clicked on new line graph.I went on to simply enter my data and titles and within 3 minutes my data wasplotted on to the line graph. I then separately changed the axis name of the
graph and added grid lines. In terms of being user friendly, excel is very easy touse and straight forward. However additional time has to be spent addinggridlines/ axes titles, which I think would be better if they were automatic. Interms of graphically appearance the graph looks simple and each to gatherinformation from. Excel has a wide range of graph formats which can be usedfor different graphs in different situations.
-
8/4/2019 Nuffield Bursary 19
20/131
I then went on to create a graph in Spotfire. Firstly I had to load my data into anexcel document to be able to load it into Spotfire. I then opened Spotfire andloaded the excel document. It initially plotted a scatter graph with Animal on theX axis and Number of Voxels Pre Duramycin on the Y axis, which wasnt useful. Ithen clicked to form a new line graph and then added a new variable Number of
Voxels Post Duramycin on the Y axis. It then told me I had to sort the 2 Y axisvariables by Colour, Line or Trellis. As I had come across this yesterday I knew allyou had to do was drag the filters box onto the graph. I then changed the nameof the lines so that each line represented a different animal. Overall Spotfire isnot user friendly, to use it you need to be taught the ropes and a lot of trial anderror however the help guide is very useful. The look is very nice and clear, itsshows each individual animal and portrays a clear decline.
So after using both excel and Spotfire to create graphs I compared theirweaknesses and strengths. Excel is a lot more user friendly than Spotfire, as ituses less technical terms and has helpful buttons. However Spotfire isnt socomplex that its unusable, it just takes some time to learn how it works. Theappearance of both graphs are simple and straight forward, but I prefer theSpotfire graph and it shows a clear decline and also shows each individualanimals data result which is quite hard to see with the excel graph.
Amide is relatively user friendly as it has a simple layout and tools. However
finding the tools was quite hard as they were named technical terms. Onelimitation of Amide was the fact that you had to identify the tumours yourself byeye using a circle tool which was difficult to control, so if the tumour wasnt acircular shape it was hard to get an accurate size of the tumour and there washuman error involved. The calculating ROI statistics tool was really useful as itquantified the image/ tumours which is needed to compare images.
In terms of analysing biological data the combination of Amide with eitherExcel/Spotfire is a relatively simple but quite time consuming way of establishingtrends.
The R Project
-
8/4/2019 Nuffield Bursary 19
21/131
One of my objectives as to try and sample different statistics analysingprograms. So I decided to move on from Spotfire and try The R Project, which is afree statistics program (Insert stuff on the R Project) . After downloading
The R Project I realised it functioned using HTML which Im not familiar too. I readthe help document[5] but I couldnt understand the lingo or how to even begin. I
decided to wait and ask my supervisor for help.Ethical Issues
I found it really had to find an ethical issue within the band of statisticallyanalysing biological data so I decided to focus on the ethical issue of animaltesting which is regularly used at GE healthcare. Reasons against animal testingcome as followed; testing is painful and animals have rights. So can animals feelpain? Humans and Animals share similar mechanisms of pain detection, similarareas in the brain of pain detection and show similar pain behaviours, whichwould suggest that animals can feel pain.[7]However its notoriously hard todistinguish whether animals feel pain this is highlighted byDr J.D. Ranking quote
It is not as a rule possible to asses degrees of real pain in animals.[6] Somepeople also argue that animals have rights, based on idea that animals should betreated like humans- not property. On the other hand people argue that withoutanimal testing, treatments for multiple diseases, that would have caused painand deaths of humans, wouldnt be around[8]. Furthermore that scientists wouldknow so much less about breeding, evolution and behaviour as such informationis gained from animal studies. Personally I believe that animal testing is vital forthe science industry to progress. After researching I realised that it was not ablack and white answer but instead a lot of ethical issues within animal testing.For example euthanasia, this occurs to animals at the end of studies or if theyare undergoing unbearable suffering (depression, if theyre large animals noteating for 5 days, having an infection which is unresponsive to treatment).[9]Some people may believe that euthanasia is unnecessary murder and othersmay argue that if the animals are under great pain they should be put out oftheir misery. Another ethical issue is whether some animals should be used inanimal testing over other animals. It is almost unanimously agreed that usingendangered animals; pandas, polar bears, in animal testing isnt ethically righthowever which animal is ethically better to use in animal testing? The lawdoesnt regulate the use of invertebrates nearly as much as they regulate theuse of vertebrates which suggests that it is more ethical to use invertebrates. [6]
As you can see from this pie chart [10] mice are the most tested on animal dueto fact they share 99% DNA with humans, they are easy to keep and reproducequickly. To summarise animal testing is a broad topic containing lots of ethicaldilemmas and is essential for scientific progress but needs to be correctlyregulated to minimise suffering of animals.
http://en.wikipedia.org/wiki/File:Types_of_vertebrates_v2en.png -
8/4/2019 Nuffield Bursary 19
22/131
Hierchal Clustering
I had been interested in Hierchal clustering tool on Spotfire since look at theirdemo gallery and looking over their demo Clustering of Gene expression data.
The had used a heat map which is like a table with rows and columns, but
instead of numbers in the columns theres a colour to represent if that value islow, high or medium. A heat map is show below, and besides it is branchesseparating the clusters.
After reading the help menu, I realised that hierchal clustering can beimplemented using heat maps. The clustering is self-explanatory, the rows thatexhibit similar results are group together. The hierchal part I assumed meant toorder the rows from smallest to biggest. Here is a heat map in which hierchalclustering is implemented.
-
8/4/2019 Nuffield Bursary 19
23/131
As you can see there are 4 clusters, I presume they choose the order of clustersbased on the first 4 columns going from lowest to highest. To quantify the datafurther they set up 4 graphs (see below), each representing one cluster, andeach graph containing all the data of all the rows and columns in that cluster.
This way clusters that are similar can be identified.
All this information was relatively simple to figure out, but I wanted to determinewhat kind of data can be hierchal clustered and what data it would be useful tohierchal cluster. I started by thinking about a few requirements for the data;multiple rows of data which are independent from one another, data in columnsmust be varied so that clusters can be established and a hierarchy can beformed and that the data in the columns are continuous so they can be ordered
from smallest to biggest to establish a hierarchy. I looked over the data sets myproject supervisor had provided me with, and decided they werent suitable so Idecided to set up a dummy data set. I also decided as I was new to hierchalclustering to make 3 very distinct clusters; red, yellow and blue.
Animal Length of tail (cm) Weight (kg) Number of legs
1 0.1 1 100
2 5 12 4
3 20 34 6
4 6 13 4
5 21 33 6
6 0.3 2 110
7 22 35 6
8 21 34 6
9 7 14 4
10 5 14 4
11 0.2 3 106
-
8/4/2019 Nuffield Bursary 19
24/131
12 6 13 4
13 0.1 2 104
I first opened Spotfire, loaded my dummy data and produced an initial heat mapwhich looked like this.
I then clicked on the hierchal clustering and instructed it to cluster by columnand that the three columns it should cluster by are Length of tail, Weight andNumber of legs. The heat map produces came out as a great example of hierchalclustering.
-
8/4/2019 Nuffield Bursary 19
25/131
Although I was happy and excited with my heat map, I wanted to see if themethod would work on real data and not on fixed data. I decided to try this afterusing the parallel point map.
Parallel point map
Another tool, in Spotfire, I saw which helped highlight clusters was a graph calleda parallel point map. A parallel point map uses a line to represent an X variableand the line joins up values on the Y axis which represent different variables. Forexample if you were looking at animal 1 the line would join up the length of tailvalue, the weight value and the number of legs value. However instead of havingthe raw values, the parallel point tool takes the highest value for each Y variableand makes this equal to 100%, and the lowest value and make this equal to 0%.
The other Y values are converted to percentages depending on how big they are,for example if 20 were the highest value and 0 the lowest value, then a value of10 would be 50%. This is done for each line (X variable), and highlights whether
X variables have similar Y variable values.
I thought this would be a good method of identifying clusters and here is theresult of the parallel point map using the dummy data in Spotfire.
Along the right side it also showed the raw values for number of legs as I clickedon that X variable , however if you clicked on another column it would show theraw values for that X variable. I was happy with the results as it clearly showed
the clusters. However I wanted the clusters to be different colours, so by using atool I learnt in K-means clustering, I dragged the hierarchal clustering tool to thecolour section to colour the lines. Here was the outcome.
-
8/4/2019 Nuffield Bursary 19
26/131
As you can see the parallel coordinate plot now clearly shows the three clusters.
I opened a data document my supervisor had given me called Colon smalltranspose which contained a table showing the results of a test carried out onhealthy colons, and colon tumours, and 100s of columns containing thevariables that were tested. When I first created a heat map there were so manyvariables it would be impossible to cluster, as you can see from the image below.
So I decided to choose 3 variables to focus on. I randomly choose the first 3variables and the heat map came out like this.
-
8/4/2019 Nuffield Bursary 19
27/131
After I implemented hierchal clustering it looked like this.
This time as there werent distinct clusters, the number of clusters was a lotmore, and the number of rows per cluster was a lot less. This got me to thinkingthat perhaps, in the case at least, that hierchal clustering isnt useful becausethere isnt a causal relationship between the variables, because if there was theclusters would be more defined. This means the rows are clustered together byrandom chance and not by an actual link. This made me think of a heat mapwhich may be more effective, one where there is two columns; one sayingwhether the tumour is healthy or not, and one other random variable. This waythere should hopefully be two clusters of healthy and non-healthy colons, andthe other variable should hopefully be clustered in a similar way so that I can see
-
8/4/2019 Nuffield Bursary 19
28/131
that healthy colons have this value of the variable and non-healthy colons havethis value of the variable. In principle I should be able to say which values of thevariable indicate the colon is healthy and which values indicate the tumour isnthealthy.
I created this heat map in Spot fire and decided to use Gene ID which saidwhether the colon was healthy or not and then randomly H. sapiens ACTB mRNA for mutant beta-actin column as it was the next column along. The resultwas disappointing.
I realised that the gene description didnt cluster because I had ignored one ofmy first criteria; the data in the columns must be continuous. However I stillwanted to see whether healthy colons had certain values for H. sapiens ACTB m
RNA for mutant beta-actin and whether tumour colons have certain values forH. sapiens ACTB m RNA for mutant beta-actin. I figured out I could answer thisby clicking the gene description column, as spot fire naturally clustered thatcolumn and observing the colours in the other columns.
As you can see there wasnt a distinct colour (value forH. sapiens ACTB m RNAfor mutant beta-actin) for a healthy colon, but the much larger values (red) of H.sapiens ACTB m RNA for mutant beta-actin where all from colons with tumours.
-
8/4/2019 Nuffield Bursary 19
29/131
This suggests that there is a link between high values of H. sapiens ACTB m RNAfor mutant beta-actin and tumour colons.
I tried out a few other variables, and the variable H.sapiens m RNA for GCAP ll/uroguanylin precursor stood out as the heat map (see below) showed that the
majority of healthy colons has a low value (blue) for H.sapiens m RNA for GCAP ll/uroguanylin precursor and the majority of the colons with tumours had highvalues (red) of H.sapiens m RNA for GCAP ll/uroguanylin precursor. With thehelp of Spotfire and the hierarchal clustering tool relationships between variablescan be established, however they should be taken with a pinch of salt as itdoesnt indicate that the variables have a causal relationship (where on variableeffects the other variable) but rather a correlation (where a variable changes asanother variable changes).
3 days later.I realised that I could use hierchal clustering with two columns one of them beinggene description (so whether the tumour is healthy or has a colon) as I wouldsimply cluster the H.sapiens m RNA for GCAP ll/uroguanylin precursor columnand not the gene description column. The result came out like this.
-
8/4/2019 Nuffield Bursary 19
30/131
This even more clearly than the previous heat map shows that a lot of healthycolons have a small value (blue) of H.sapiens m RNA for GCAP ll/uroguanylinprecursor and that a lot of colons with tumours have a large value (red) ofH.sapiens m RNA for GCAP ll/uroguanylin precursor.
So after my research I have determined that the criteria needed to hierchalcluster is that-multiple columns so that rows can be clustered based on two values-the columns where hierchal clustering is being implemented must havecontinuous data, however one other column that doesnt contain continuous datacan also be placed alongside the other columns.-multiple rows of data that are independent from one another, so rows can beclustered
The kind of data which it is useful to use hierchal clustering on is on data withcolumns where you are trying to establish whether there is a correlationbetween. If hierchal clustering works and there are defined clusters, then youcan establish a link between columns. However if there is lots of clusterscontaining one or two rows they most likely have been clustered by chance, sotheres probably isnt a link between the two column variables.
Exploring Excel
Excel is a very simple program, which most people how used, so I wondered howuseful it is in analysing biological data. I have used it twice above; once making aline graph and once making a histogram. Im going to use it again, this timetrying out other graphs and figuring out which data is suited to which graph.
Pie Charts
From my understanding of pie charts, theyre used to show percentages assegments of a circle. So useful data to use in a pie chart would be if there was aquestion, and in the pie chart all the answers by frequency are shown. Forexample How do children travel to school, the answers may look as followed.Mode of transport Frequency
-
8/4/2019 Nuffield Bursary 19
31/131
Car 37Walk 23Public transport 28Cycle 9Other 3
This data could then be placed in excel, highlighted and then by pressing the piechart icon a pie chart is instantly produced.
Here is the simple pie chart created, and the process above highlights how easy
it is to create. However there are a few improvements I would like to make;1) Put a title on the graph so that a person looking at the graph couldunderstand the information
2) Order the segments from biggest to smallest, starting from North.3) Put the percentage frequency of the pie chart next to the corresponding
mode of transport
The first change was simple, click layout, chart title and then add chart title. Thesecond I believed would mean I would have to manually change the data. Thiswas not difficult as I only had 5 pieces of data, however for a larger data set thiswould prove problematic. Although you can order column of data from thebiggest to smallest, this would change only the frequency column not the mode
of transport column, and so the mode of transports would have the wrongfrequencies.
-
8/4/2019 Nuffield Bursary 19
32/131
As you can see the Walk frequencies and Public transport frequencies have beenflipped making the data incorrect. This was the first limitation when using theexcel pie chart tool. 3 was as easy to change as 1, all you needed to was click onthe format and changed it to included percentages. The finished pie chart lookedlike this.
I was happy with the outcome and Excel is very user friendly and quick. HoweverI think the pie chart should order your data from largest segment to smallestsegment.
I wondered how pie charts could be used to analyse biological data. I had a tableof group data which showed how many cells up took certain levels of radiationthat I had used when constructing the histogram. I wondered if I could use thisdata and construct a pie chart to show what amount of radiation was most tookup by cells. However I was concerned, that with 11 groups, it would be two much
information for the pie chart.
-
8/4/2019 Nuffield Bursary 19
33/131
Here is the data I will be using.
Uptake value post
-500 17
-400 82
-300 185
-200 317
-100 500
0 1143
100 19375
200 107230
300 101333
400 14116
500 188
I first ordered the data and then constructed a pie chart and here was theoutcome.
I think the problem with this pie chart is that it isnt very clear. Unless you had
some way of labelling the numbers in the column on the right as Amount ofradioactivity cells up took, its quite confusing. Another drawback is that some of
-
8/4/2019 Nuffield Bursary 19
34/131
the smaller percentages were rounded to 0%, and even the 1% percentage it ishard to distinguish the colour and thus the amount of radioactivity 1% of cells uptook. As the pie chart rounds to the nearest 1% it can make data less accurateand precise. Despite these flaws, the pie chart does show which amount ofradiotracer the cells mostly took up, 200 and 300. I decided to correct some of
the flaws concerning the low percentages to group all amounts which has 0 or1% as other. As excel doesnt have a tool for this I did it manually.
This was the new pie chart result.
This pie chart is less messy than the previous pie chart, but unlike the previouspie chart doesnt show the range of radioactive uptake values.
After doing these pie charts I came to two realisations;1) Pie charts are good in the fact that the variables can be continuous or
discrete, however the values have to be continuous.2) That some success of this pie chart is due to the large percentages of 200
and 300 radiotracer uptake value.
I wanted to test my second realisation by using similar data to before, howeverinstead of 11 groups, containing 32 groups. The previous pie chart representsthe percentage of cells that up took certain radiotracer levels after the tumourtreatment Duramycin was administered, and the data I am about to use is thepercentage of cells that up took certain radiotracer levels before the tumourtreatment Duramycin was administered. I will compare the pie charts afterwards.
I first I changed the data to size order and also group together all data 1% andunder which left me with 15 groups compared to final 5 groups in the previouspie chart. I also figured out that when using all continuous data, for the variableand value, that when ordering the values from biggest to smallest the variables
moved with them.
-
8/4/2019 Nuffield Bursary 19
35/131
I then calculated the pie chart and this was the outcome.
As you can see it is not as clear as previous pie chart, which contains the top 4percentages and the other group. If I was to implement the same rule for this piechart nearly 50% of cells would have up took other amounts of radiotracer and
too much information about the percentage of cell that up took radiotracer wouldbe loss.
I decided to see how easy the pie charts were to compare.
Its not easy to compare these two pie charts directly you have to draw
conclusions. The main uptake of radiotracer pre dose of Duramycin was between100-600, the main uptake of radiotracer post dose of Duramycin was 200-300.
-
8/4/2019 Nuffield Bursary 19
36/131
These two statements can be compared to draw two conclusion; the tumour predose Duramycin took up a greater range of radiotracer, 500 compared to 100post dose and the tumour pre dose took up more Duramycin, average of 100-600pre dose being 350 and the average of 200-300 post dose being 250. Howeverthis analysing shows that the graphs are not clear and therefore do not serve
their purpose of displaying this data.After comparing these two pie charts I came to two other realisations;1) Pie charts cannot be comparable when there are different numbers of
groups2) Pie charts are impractical to display data that has 10 or more variables
Although Pie charts had been bad at displaying this biological data, I set myself achallenge to find some biological data which would be well displayed in a piechart.
-
8/4/2019 Nuffield Bursary 19
37/131
Voxel histogram
Simply put my project aims to look at CT and SPECT images of animals pre andpost tumour treatment and using image analysis software to calculate thetumour cells radiotracer uptake value. Then using this information and statisticssoftware to display the data and determine two things; how the tumourtreatment affects the tumour and how the radiotracer is up taken by tumourcells.
Method
The image analysis process-upload the CT and SPECT into *Amide-make sure they are aligned and in the right orientation-change the aesthetics of them to make them easier to analyse e.g. filtering orchanging the colour range-add an ROI (region of interest) and shape is using the mouse and givingdimensions to circle the tumour-calculate the ROI statistics using a calculate ROI statistics tool-save the raw data values of this process
The image analysis process work by measuring each *voxels radiotracer uptakevalue and then lists every radiotracer value in a table, which is the raw data.
This leaves me with a table of radiotracer values (each representing one voxel)for one animals tumour pre/post tumour treatment, so I repeated the process ofall the animals tumours pre and post tumour treatment.
Once I had done this for the CT images I did it on the SPECT images for the sameset of animals.
I then wanted to calculate the frequency of groups of these radiotracer uptakesvalues of each animal pre and post tumour treatment in order to construct ahistogram, how I did this occurred in two different ways.
After completing these graphs I will compare them and explain the shapes.
Animal 1- manual
For the first animal I went with the manual approach and using a CT image.
I uploaded my raw data into an Excel work sheet. I had to decide what sizegroupings to use, since the maximum uptake was 3163.27 and the minimumuptake was -500 for animal 1, I decided to create 37 groups each 100 uptakeunits big.
I grouped the data in the following way; I ordered all the radiotracer values from
largest to smallest and then highlighted all the values within the designatedgroup. For example in the radiotracer values group from 0 from 100, I started
-
8/4/2019 Nuffield Bursary 19
38/131
with the smallest number 0 and then clicking on it scrolled up to the biggestnumber, 99. With all these values highlighted I looked at the right hand of thebar on the bottom of the excel spread sheet and recorded the count. This wasthen the frequency for the group. I repeated this for each group pre and posttumour treatment and filled in the pre and post columns in the table below. I
then calculated the percentage of voxels by dividing the frequency by the totalnumber of voxels in the tumour and then multiplying this by 100. I did thisseparately for pre and post as both had a different number of voxels within thetumour. Instead of doing this part manually for all frequencies I used and Excelformula in the first box in the % column and then dragged the formula down theentire column for pre and post%.
My format for the table consists of 38 rows each a representing a group, then 5rows; the first containing the uptake value, the second containing the frequencyof pre dose voxels that up took the corresponding radiotracer value, the thirdcontaining the percentage of pre dose voxels that up took the correspondingradiotracer value, the third containing the number of post dose voxels that up
took the corresponding radiotracer value, the fourth containing the percentageof post dose voxels that up took the corresponding radiotracer value.
Here is the table.
Uptake
value
pre pre% post post%
-500 0
0
17 0.006953
-400 0
0
82 0.03354
-300 0
0
185 0.075669
-200 0
0
317 0.12966
-100 0
0
500 0.204511
0 406 3.695276236 1143 0.467511
100 1053 9.584053882 19375 7.924789
200 1417 12.89706016 107230 43.85936
300 1546 14.07117503 101333 41.44736
400 1436 13.06999181 14116 5.773746
500 1237 11.25876035 188 0.076896
600 997 9.074360608 0 0
700 752 6.844452535 0 0
-
8/4/2019 Nuffield Bursary 19
39/131
800 482 4.387002822 0 0
900 440 4.004732866 0 0
1000 343 3.121871302 0 0
1100 257 2.33912806 0 0
1200 197 1.793028124 0 0
1300 144 1.310639847 0 0
1400 83 0.755438245 0 0
1500 53 0.482388277 0 0
1600 54 0.491489943 0 0
1700 28 0.254846637 0 0
1800 20 0.182033312 0 0
1900 10 0.091016656 0 0
2000 14 0.127423318 0 0
2100 6 0.054609994 0 0
2200 4 0.036406662 0 0
2300 2 0.018203331 0 0
2400 1 0.009101666 0 0
2500 2 0.018203331 0 0
2600 0 0 0 0
2700 1 0.009101666 0 0
2800 0 0 0 0
2900 0 0 0 0
3000 1 0.009101666 0 0
3100 1 0.009101666 0 0
As you can see the groups 100 to 500 take up a large percentage of frequencies(60%) for pre dose and the groups 200 and 300 take up a large percentage offrequencies (85%) for post dose. To make my data more accurate I decreasedthe size of the groupings in the section 100-600, to 20 uptake values. Here is theresult table after I had done so.
Uptake value pre pre% post post%
-500 0 0 17 0.006953
-
8/4/2019 Nuffield Bursary 19
40/131
-400 0 0 82 0.03354
-300 0 0 185 0.075669
-200 0 0 317 0.12966
-100 0 0 500 0.204511
0 406 3.695276236 1143 0.467511
100 171 1.556384818 3875 1.584958
120 188 1.711113134 3875 1.584958
140 205 1.865841449 3875 1.584958
160 233 2.120688086 3875 1.584958
180 256 2.330026395 3875 1.584958
200 280 2.548466369 13046 5.336093
220 265 2.411941385 17717 7.246632
240 292 2.657686357 22441 9.178849
260 305 2.776008009 25920 10.60183
280 275 2.502958041 28106 11.49595
300 313 2.848821334 27882 11.40433
320 303 2.757804678 25230 10.31961
340 293 2.666788022 20890 8.544457
360 303 2.757804678 16027 6.555386
380 334 3.039956312 11304 4.623578
400 298 2.71229635 2823.2 1.154749
420 299 2.721398016 2823.2 1.154749
440 274 2.493856376 2823.2 1.154749
460 278 2.530263038 2823.2 1.154749
480 287 2.612178029 2823.2 1.154749
500 256 2.330026395 37.6 0.015379
520 262 2.384636388 37.6 0.015379
540 233 2.120688086 37.6 0.015379
560 247 2.248111404 37.6 0.015379
580 239 2.17529808 37.6 0.015379
-
8/4/2019 Nuffield Bursary 19
41/131
600 997 9.074360608 0 0
700 752 6.844452535 0 0
800 482 4.387002822 0 0
900 440 4.004732866 0 0
1000 343 3.121871302 0 0
1100 257 2.33912806 0 0
1200 197 1.793028124 0 0
1300 144 1.310639847 0 0
1400 83 0.755438245 0 0
1500 53 0.482388277 0 0
1600 54 0.491489943 0 0
1700 28 0.254846637 0 0
1800 20 0.182033312 0 0
1900 10 0.091016656 0 0
2000 14 0.127423318 0 0
2100 6 0.054609994 0 0
2200 4 0.036406662 0 0
2300 2 0.018203331 0 0
2400 1 0.009101666 0 0
2500 2 0.018203331 0 0
2600 0 0 0 0
2700 1 0.009101666 0 0
2800 0 0 0 0
2900 0 0 0 0
3000 1 0.009101666 0 0
3100 1 0.009101666 0 0
I then plotted a histogram using this data set and here is the result.
-
8/4/2019 Nuffield Bursary 19
42/131
There was a problem with the shape of the graph as there were two spikes. Irealised these spikes were in places where the groups sizes changed from 20 to100 or vice versa, so I came to the conclusion for the graph to be accurate thatall groups must be equally spaced. I had to change my data table so that eachgroup was 20 radiotracer values big. Manually doing this in the same manner asbefore would be very time consuming so I divided each 100 big groupsfrequency by 5 so 1/5 of the frequency was put into a 5 20 big groups which arereplacing the 100 big group.
My data table then looked like this.
Uptake value pre pre% post post%
-500 0 0 3.4 0.001391
-480 0 0 3.4 0.001391
-460 0 0 3.4 0.001391
-440 0 0 3.4 0.001391
-420 0 0 3.4 0.001391
-400 0 0 16.4 0.006708
-380 0 0 16.4 0.006708
-360 0 0 16.4 0.006708
-340 0 0 16.4 0.006708
-320 0 0 16.4 0.006708
-300 0 0 37 0.015134
-280 0 0 37 0.015134
-260 0 0 37 0.015134
-
8/4/2019 Nuffield Bursary 19
43/131
-240 0 0 37 0.015134
-220 0 0 37 0.015134
-200 0 0 63.4 0.025932
-180 0 0 63.4 0.025932
-160 0 0 63.4 0.025932
-140 0 0 63.4 0.025932
-120 0 0 63.4 0.025932
-100 0 0 100 0.040902
-80 0 0 100 0.040902
-60 0 0 100 0.040902
-40 0 0 100 0.040902
-20 0 0 100 0.040902
0 81.2 0.739055247 228.6 0.093502
20 81.2 0.739055247 228.6 0.093502
40 81.2 0.739055247 228.6 0.093502
60 81.2 0.739055247 228.6 0.093502
80 81.2 0.739055247 228.6 0.093502
100 171 1.556384818 3875 1.584958
120 188 1.711113134 3875 1.584958
140 205 1.865841449 3875 1.584958
160 233 2.120688086 3875 1.584958
180 256 2.330026395 3875 1.584958
200 280 2.548466369 13046 5.336093
220 265 2.411941385 17717 7.246632
240 292 2.657686357 22441 9.178849
260 305 2.776008009 25920 10.60183
280 275 2.502958041 28106 11.49595
300 313 2.848821334 27882 11.40433
320 303 2.757804678 25230 10.31961
-
8/4/2019 Nuffield Bursary 19
44/131
340 293 2.666788022 20890 8.544457
360 303 2.757804678 16027 6.555386
380 334 3.039956312 11304 4.623578
400 298 2.71229635 2823.2 1.154749
420 299 2.721398016 2823.2 1.154749
440 274 2.493856376 2823.2 1.154749
460 278 2.530263038 2823.2 1.154749
480 287 2.612178029 2823.2 1.154749
500 256 2.330026395 37.6 0.015379
520 262 2.384636388 37.6 0.015379
540 233 2.120688086 37.6 0.015379
560 247 2.248111404 37.6 0.015379
580 239 2.17529808 37.6 0.015379
600 199.4 1.814872122 0 0
620 199.4 1.814872122 0 0
640 199.4 1.814872122 0 0
660 199.4 1.814872122 0 0
680 199.4 1.814872122 0 0
700 150.4 1.368890507 0 0
720 150.4 1.368890507 0 0
740 150.4 1.368890507 0 0
760 150.4 1.368890507 0 0
780 150.4 1.368890507 0 0
800 96.4 0.877400564 0 0
820 96.4 0.877400564 0 0
840 96.4 0.877400564 0 0
860 96.4 0.877400564 0 0
880 96.4 0.877400564 0 0
900 88 0.800946573 0 0
920 88 0.800946573 0 0
-
8/4/2019 Nuffield Bursary 19
45/131
940 88 0.800946573 0 0
960 88 0.800946573 0 0
980 88 0.800946573 0 0
1000 68.6 0.62437426 0 0
1020 68.6 0.62437426 0 0
1040 68.6 0.62437426 0 0
1060 68.6 0.62437426 0 0
1080 68.6 0.62437426 0 0
1100 51.4 0.467825612 0 0
1120 51.4 0.467825612 0 0
1140 51.4 0.467825612 0 0
1160 51.4 0.467825612 0 0
1180 51.4 0.467825612 0 0
1200 39.4 0.358605625 0 0
1220 39.4 0.358605625 0 0
1240 39.4 0.358605625 0 0
1260 39.4 0.358605625 0 0
1280 39.4 0.358605625 0 0
1300 28.8 0.262127969 0 0
1320 28.8 0.262127969 0 0
1340 28.8 0.262127969 0 0
1360 28.8 0.262127969 0 0
1380 28.8 0.262127969 0 0
1400 16.6 0.151087649 0 0
1420 16.6 0.151087649 0 0
1440 16.6 0.151087649 0 0
1460 16.6 0.151087649 0 0
1480 16.6 0.151087649 0 0
1500 10.6 0.096477655 0 0
1520 10.6 0.096477655 0 0
-
8/4/2019 Nuffield Bursary 19
46/131
1540 10.6 0.096477655 0 0
1560 10.6 0.096477655 0 0
1580 10.6 0.096477655 0 0
1600 10.8 0.098297989 0 0
1620 10.8 0.098297989 0 0
1640 10.8 0.098297989 0 0
1660 10.8 0.098297989 0 0
1680 10.8 0.098297989 0 0
1700 5.6 0.050969327 0 0
1720 5.6 0.050969327 0 0
1740 5.6 0.050969327 0 0
1760 5.6 0.050969327 0 0
1780 5.6 0.050969327 0 0
1800 4 0.036406662 0 0
1820 4 0.036406662 0 0
1840 4 0.036406662 0 0
1860 4 0.036406662 0 0
1880 4 0.036406662 0 0
1900 2 0.018203331 0 0
1920 2 0.018203331 0 0
1940 2 0.018203331 0 0
1960 2 0.018203331 0 0
1980 2 0.018203331 0 0
2000 2.8 0.025484664 0 0
2020 2.8 0.025484664 0 0
2040 2.8 0.025484664 0 0
2060 2.8 0.025484664 0 0
2080 2.8 0.025484664 0 0
2100 1.2 0.010921999 0 0
2120 1.2 0.010921999 0 0
-
8/4/2019 Nuffield Bursary 19
47/131
2140 1.2 0.010921999 0 0
2160 1.2 0.010921999 0 0
2180 1.2 0.010921999 0 0
2200 0.8 0.007281332 0 0
2220 0.8 0.007281332 0 0
2240 0.8 0.007281332 0 0
2260 0.8 0.007281332 0 0
2280 0.8 0.007281332 0 0
2300 0.4 0.003640666 0 0
2320 0.4 0.003640666 0 0
2340 0.4 0.003640666 0 0
2360 0.4 0.003640666 0 0
2380 0.4 0.003640666 0 0
2400 0.2 0.001820333 0 0
2420 0.2 0.001820333 0 0
2440 0.2 0.001820333 0 0
2460 0.2 0.001820333 0 0
2480 0.2 0.001820333 0 0
2500 0.4 0.003640666 0 0
2520 0.4 0.003640666 0 0
2540 0.4 0.003640666 0 0
2560 0.4 0.003640666 0 0
2580 0.4 0.003640666 0 0
2600 0 0 0 0
2620 0 0 0 0
2640 0 0 0 0
2660 0 0 0 0
2680 0 0 0 0
2700 0.2 0.001820333 0 0
2720 0.2 0.001820333 0 0
-
8/4/2019 Nuffield Bursary 19
48/131
2740 0.2 0.001820333 0 0
2760 0.2 0.001820333 0 0
2780 0.2 0.001820333 0 0
2800 0 0 0 0
2820 0 0 0 0
2840 0 0 0 0
2860 0 0 0 0
2880 0 0 0 0
2900 0 0 0 0
2920 0 0 0 0
2940 0 0 0 0
2960 0 0 0 0
2980 0 0 0 0
3000 0.2 0.001820333 0 0
3020 0.2 0.001820333 0 0
3040 0.2 0.001820333 0 0
3060 0.2 0.001820333 0 0
3080 0.2 0.001820333 0 0
3100 0.2 0.001820333 0 0
3120 0.2 0.001820333 0 0
3140 0.2 0.001820333 0 0
3160 0.2 0.001820333 0 0
3180 0.2 0.001820333 0 0
I then constructed a histogram from this data that looked like this.
-
8/4/2019 Nuffield Bursary 19
49/131
This time my histogram has no random spikes. I did two versions so that you cansee the shapes of both curves. Although the graph is good, it has a lot of emptyspace before -50 and after 1800, as the percentage of voxels that up tookradiotracer values outside this bracket is very small so I decided to trim thegraph and see what it looks like. I also wanted to add some grid lines so that Icould read values of it more easily.
Here is the finished graph.
-
8/4/2019 Nuffield Bursary 19
50/131
I was much happier with this graph as it was much clearer to see how Duramycinhad affected the tumour.
The numbers of voxels per tumour was very high (10987 for pre dose and244486 for post dose) and sorting them into the 185 groups manually was verylong winded and took me 2 hours to complete it. I knew that there must be aneasier way to do it to make my project feasible in the allocated time so I had tofind a tool to do it for me. However there wasnt a function on Excel so I startedresearching Excel toolbar packages, which add extra tool that can be used inExcel. I found one called Daniels extra-large toolbox, which I downloaded intoexcel and had the exact tool I wanted to group the data.
Animal 2 using tool
I used the same imaging process, saved the raw data and opened the document
in Excel.
I downloaded the toolbox, and placed the data into Excel however when I went tocalculate the data using the Multi histogram tool it came up with a TechnicalError. I read through the programming on the XL toolbox and realised that Ihadnt been using the tool correctly. I learnt the proper way of using the tool wasto select the data and then separately select a couple of columns, then eitheruse an auto button to calculate group size or enter your own group size. Imanually entered a group size of 20, as from previous experience this hadworked well. Although it was quicker than manually sorting the data the XL toolhad a couple of downsides as it took 20 minutes to order the data and slowedthe computer.
Here is the table, after I calculated percentages in the same process as animal 1.
-
8/4/2019 Nuffield Bursary 19
51/131
Uptake value pre pre% post post%
-1197.27721 2 0.000202628 14 0.002194557
-1177.277212 4 0.000405257 9 0.001410787
-1157.277212 4 0.000405257 19 0.002978327
-1137.277212 8 0.000810514 31 0.004859376
-1117.277212 12 0.001215771 47 0.007367441
-1097.277212 35 0.003545999 69 0.01081603
-1077.277212 42 0.004255198 96 0.01504839
-1057.277212 80 0.00810514 121 0.018967242
-1037.277212 125 0.012664281 150 0.023513109
-1017.277212 149 0.015095823 198 0.031037304
-997.2772117 232 0.023504906 216 0.033858877
-977.2772117 302 0.030596903 236 0.036993959
-957.2772117 376 0.038094157 256 0.04012904
-937.2772117 412 0.04174147 285 0.044674908
-917.2772117 501 0.050758438 277 0.043420875
-897.2772117 538 0.054507066 284 0.044518154
-877.2772117 597 0.060484606 297 0.046555956
-857.2772117 564 0.057141236 275 0.043107367
-837.2772117 602 0.060991178 285 0.044674908
-817.2772117 531 0.053797866 293 0.04592894
-797.2772117 491 0.049745296 273 0.042793859
-777.2772117 452 0.04579404 274 0.042950613
-757.2772117 419 0.04245067 275 0.043107367
-737.2772117 368 0.037283643 263 0.041226318
-717.2772117 325 0.032927131 246 0.038561499
-697.2772117 349 0.035358673 235 0.036837205
-677.2772117 338 0.034244216 251 0.03934527
-657.2772117 289 0.029279818 250 0.039188516
-637.2772117 282 0.028570618 258 0.040442548
-
8/4/2019 Nuffield Bursary 19
52/131
-617.2772117 295 0.029887703 253 0.039658778
-597.2772117 311 0.031508731 290 0.045458678
-577.2772117 287 0.029077189 270 0.042323597
-557.2772117 283 0.028671932 278 0.043577629
-537.2772117 295 0.029887703 287 0.044988416
-517.2772117 277 0.028064047 273 0.042793859
-497.2772117 284 0.028773247 300 0.047026219
-477.2772117 308 0.031204789 294 0.046085694
-457.2772117 315 0.031913988 296 0.046399202
-437.2772117 304 0.030799532 333 0.052199103
-417.2772117 308 0.031204789 322 0.050474808
-397.2772117 302 0.030596903 340 0.053296381
-377.2772117 317 0.032116617 309 0.048437005
-357.2772117 342 0.034649473 321 0.050318054
-337.2772117 310 0.031407417 351 0.055020676
-317.2772117 347 0.035156044 348 0.054550414
-297.2772117 370 0.037486272 359 0.056274708
-277.2772117 351 0.035561301 394 0.061761101
-257.2772117 378 0.038296786 389 0.06097733
-237.2772117 377 0.038195472 402 0.063015133
-217.2772117 412 0.04174147 425 0.066620476
-197.2772117 405 0.041032271 385 0.060350314
-177.2772117 423 0.042855927 430 0.067404247
-157.2772117 422 0.042754613 489 0.076652736
-137.2772117 509 0.051568952 549 0.08605798
-117.2772117 525 0.05318998 548 0.085901226
-97.27721173 514 0.052075524 634 0.099382075
-77.27721173 591 0.059876721 739 0.115841252
-57.27721173 597 0.060484606 936 0.146721802
-37.27721173 721 0.073047573 1449 0.227136636
-
8/4/2019 Nuffield Bursary 19
53/131
-17.27721173 794 0.080443513 2291 0.359123557
2.722788268 1129 0.114383786 3981 0.624037922
22.72278827 1665 0.168688224 6688 1.048371169
42.72278827 2681 0.2716235 10927 1.712851639
62.72278827 4448 0.450645777 16568 2.597101304
82.72278827 7543 0.764213376 24569 3.851290556
102.7227883 12649 1.281523928 34006 5.330578642
122.7227883 20689 2.096090486 44183 6.925864734
142.7227883 31231 3.164145293 53137 8.329440607
162.7227883 44619 4.520540451 61085 9.575321894
182.7227883 60381 6.117455634 64866 10.168009
202.7227883 75092 7.607889543 64962 10.18305739
222.7227883 87622 8.877357076 60748 9.522495775
242.7227883 96885 9.815830959 52052 8.15936245
262.7227883 100249 10.15665209 41860 6.561725047
282.7227883 96693 9.796378624 30715 4.814701023
302.7227883 87576 8.872696621 20604 3.229760699
322.7227883 73903 7.487426902 12510 1.960993319
342.7227883 57578 5.833471796 6870 1.076900408
362.7227883 40632 4.116600542 3623 0.567919968
382.7227883 27109 2.746527961 1686 0.264287349
402.7227883 16471 1.668746986 762 0.119446595
422.7227883 9051 0.916995263 319 0.050004546
442.7227883 4686 0.474758568 102 0.015988914
462.7227883 2171 0.219953233 38 0.005956654
482.7227883 1024 0.10374579 6 0.000940524
502.7227883 498 0.050454496 3 0.000470262
522.7227883 257 0.026037762 5 0.00078377
542.7227883 182 0.018439193 0 0
562.7227883 132 0.013373481 0 0
-
8/4/2019 Nuffield Bursary 19
54/131
582.7227883 115 0.011651139 0 0
602.7227883 113 0.01144851 0 0
622.7227883 119 0.012056396 0 0
642.7227883 107 0.010840625 0 0
662.7227883 96 0.009726168 0 0
682.7227883 92 0.009320911 0 0
702.7227883 84 0.008510397 0 0
722.7227883 89 0.009016968 0 0
742.7227883 78 0.007902511 0 0
762.7227883 62 0.006281483 0 0
782.7227883 50 0.005065712 0 0
802.7227883 46 0.004660455 0 0
822.7227883 36 0.003647313 0 0
842.7227883 46 0.004660455 0 0
862.7227883 37 0.003748627 0 0
882.7227883 32 0.003242056 0 0
902.7227883 24 0.002431542 0 0
922.7227883 23 0.002330228 0 0
942.7227883 18 0.001823656 0 0
962.7227883 19 0.001924971 0 0
982.7227883 23 0.002330228 0 0
1002.722788 13 0.001317085 0 0
1022.722788 17 0.001722342 0 0
1042.722788 16 0.001621028 0 0
1062.722788 6 0.000607885 0 0
1082.722788 11 0.001114457 0 0
1102.722788 8 0.000810514 0 0
1122.722788 6 0.000607885 0 0
1142.722788 9 0.000911828 0 0
1162.722788 4 0.000405257 0 0
-
8/4/2019 Nuffield Bursary 19
55/131
1182.722788 11 0.001114457 0 0
1202.722788 5 0.000506571 0 0
1222.722788 4 0.000405257 0 0
1242.722788 3 0.000303943 0 0
1262.722788 5 0.000506571 0 0
1282.722788 4 0.000405257 0 0
1302.722788 4 0.000405257 0 0
1322.722788 2 0.000202628 0 0
1342.722788 7 0.0007092 0 0
1362.722788 4 0.000405257 0 0
1382.722788 5 0.000506571 0 0
1402.722788 6 0.000607885 0 0
1422.722788 2 0.000202628 0 0
1442.722788 2 0.000202628 0 0
1462.722788 0 0 0 0
1482.722788 0 0 0 0
1502.722788 1 0.000101314 0 0
1522.722788 1 0.000101314 0 0
1542.722788 1 0.000101314 0 0
1562.722788 1 0.000101314 0 0
1582.722788 1 0.000101314 0 0
1602.722788 2 0.000202628 0 0
1622.722788 0 0 0 0
1642.722788 0 0 0 0
1662.722788 0 0 0 0
1682.722788 1 0.000101314 0 0
1702.722788 2 0.000202628 0 0
1722.722788 1 0.000101314 0 0
1742.722788 1 0.000101314 0 0
1762.722788 1 0.000101314 0 0
-
8/4/2019 Nuffield Bursary 19
56/131
1782.722788 0 0 0 0
1802.722788 2 0.000202628 0 0
1822.722788 1 0.000101314 0 0
I knew from previous experience not to include any radiotracer values withcorresponding voxel frequencies percentage of less than 0.1%, so I made my Xaxis range from -72.3 to 402.7, which although is only a 5th of all radiotracervalues it is 96% of all voxels.
Here was the graph.
After seeing the sharp line at 402.7 I realised I had made a mistake and that I cutoff my pre Duramycin values at
-
8/4/2019 Nuffield Bursary 19
57/131
I was happy with these two graphs, but I wanted the X axis to look neater, so Ichanged the radio uptake values to have only 2 decimal places compared to 7/8.Furthermore I realised although I had increased the percentages of voxels thatup took each radiotracer value up to 482.7 I hadnt increased the actually X axisradiotracer values up to 482.7, so that my graph went beyond the numbers onthe X axis, so I correct that.
This was the finished graph.
-
8/4/2019 Nuffield Bursary 19
58/131
I noticed a downfall with this graph; because the histograms pre and post areasoverlap it is hard to see the shape of each area. I thought this problem might besolved in two ways;
1) Stretch out the X axis as wide as possible so that the peaks can lie closerto their true values and may be more spread out. For example if twopoints 149 and 51 were both rounded to 100, on a not spread out enoughX axis they would appear to overlap when they truly dont. This hadworked with the previous histogram on animal 1, as I made the range ofvalues on the X axis smaller the peaks took less general values and sospread out.
2) Use a 3-D histogram so that the peaks are separated.
I went on to try out both these improvments on Excel.
-
8/4/2019 Nuffield Bursary 19
59/131
Improvement 1.
This worked to some extent; however the peaks truly overlap regardless of howaccurate the points are.
Improvement 2
This didnt work. I thought that the 3-D histogram would show a more side onview, however it didnt.
I came to the conclusion that with perfect histogram the sets of data wouldntoverlap, however with this histogram you can still see that the peak of thetumour post treatment is more to the left on the X axis than the peak of thetumour pre-treatment.
Along with the histograms I added line charts to give a clearer view.
Here is the line chart of animal 2.
Animal 3- using tool
I went on to try this same technique that I used on the animal 2 with animal 3 asthe histogram from the first animal was very different to the histogram from thesecond animal, and I wanted to see which was correct.
Here is the data table I collected using XL toolbox.
Uptake value pre pre% post post%
-
8/4/2019 Nuffield Bursary 19
60/131
-1235.06 0 0 2 0.000268
-1215.06 0 0 2 0.000268
-1195.06 0 0 11 0.001472
-1175.06 0 0 26 0.003479
-1155.06 0 0 31 0.004148
-1135.06 0 0 83 0.011107
-1115.06 0 0 182 0.024355
-1095.06 0 0 273 0.036532
-1075.06 0 0 369 0.049379
-1055.06 0 0 501 0.067043
-1035.06 0 0 598 0.080023
-1015.06 0 0 753 0.100765
-995.06 0 0 908 0.121507
-975.06 0 0 969 0.12967
-955.06 0 0 1035 0.138502
-935.06 0 0 1131 0.151348
-915.06 0 0 1194 0.159779
-895.06 0 0 1086 0.145326
-875.06 0 0 1110 0.148538
-855.06 0 0 1114 0.149073
-835.06 0 0 1022 0.136762
-815.06 0 0 975 0.130473
-795.06 0 0 893 0.1195
-775.06 0 0 852 0.114013
-755.06 0 0 726 0.097152
-735.06 0 0 647 0.08658
-715.06 0 0 635 0.084975
-695.06 0 0 535 0.071593
-675.06 0 0 542 0.072529
-
8/4/2019 Nuffield Bursary 19
61/131
-655.06 0 0 541 0.072396
-635.06 0 0 500 0.066909
-615.06 0 0 467 0.062493
-595.06 0 0 535 0.071593
-575.06 0 0 471 0.063028
-555.06 0 0 442 0.059148
-535.06 0 0 446 0.059683
-515.06 0 0 483 0.064634
-495.06 0 0 467 0.062493
-475.06 0 0 466 0.062359
-455.06 0 0 491 0.065705
-435.06 2 0.000396981 473 0.063296
-415.06 1 0.00019849 516 0.06905
-395.06 2 0.000396981 464 0.062092
-375.06 2 0.000396981 490 0.065571
-355.06 2 0.000396981 521 0.069719
-335.06 4 0.000793961 505 0.067578
-315.06 7 0.001389432 581 0.077748
-295.06 10 0.001984903 517 0.069184
-275.06 12 0.002381883 527 0.070522
-255.06 27 0.005359238 546 0.073065
-235.06 17 0.003374335 582 0.077882
-215.06 19 0.003771315 569 0.076143
-195.06 24 0.004763767 559 0.074804
-175.06 23 0.004565277 640 0.085644
-155.06 28 0.005557728 658 0.088052
-135.06 31 0.006153199 672 0.089926
-115.06 29 0.005756218 632 0.084573
-95.06 42 0.008336592 648 0.086714
-75.06 38 0.007542631 774 0.103575
-
8/4/2019 Nuffield Bursary 19
62/131
-55.06 36 0.00714565 771 0.103174
-35.06 49 0.009726024 818 0.109463
-15.06 67 0.013298849 831 0.111203
4.94 105 0.02084148 990 0.13248
24.94 151 0.029972033 1269 0.169815
44.94 264 0.052401435 1613 0.215849
64.94 497 0.098649671 2422 0.324107
84.94 956 0.18975671 3854 0.515735
104.94 1815 0.360259863 6511 0.87129
124.94 3619 0.718336334 10505 1.405759
144.94 6887 1.367002578 16151 2.161296
164.94 11385 2.259811871 24608 3.292996
184.94 17817 3.536501371 34130 4.567212
204.94 25613 5.083931616 45927 6.145864
224.94 34418 6.831638557 57502 7.694809
244.94 42803 8.495979579 66856 8.946544
264.94 49594 9.843927091 73388 9.820644
284.94 53877 10.69406097 74585 9.980824
304.94 54384 10.79469555 70774 9.470843
324.94 50455 10.01482722 63184 8.455164
344.94 43406 8.61566922 51572 6.901268
364.94 34625 6.872726046 38830 5.196157
384.94 25820 5.125019105 27109 3.627675
404.94 17934 3.559724734 17250 2.308362
424.94 11357 2.254254143 10097 1.351161
444.94 6727 1.335244133 5460 0.730647
464.94 3996 0.793167171 2632 0.352209
484.94 2050 0.40690508 1300 0.173964
504.94 984 0.195314438 519 0.069452
524.94 516 0.102420986 264 0.035328
-
8/4/2019 Nuffield Bursary 19
63/131
544.94 265 0.052599925 125 0.016727
564.94 145 0.028781091 70 0.009367
584.94 99 0.019650538 40 0.005353
604.94 74 0.014688281 28 0.003747
624.94 68 0.013497339 33 0.004416
644.94 62 0.012306398 30 0.004015
664.94 52 0.010321495 30 0.004015
684.94 56 0.011115456 23 0.003078
704.94 48 0.009527534 21 0.00281
724.94 49 0.009726024 25 0.003345
744.94 39 0.007741121 19 0.002543
764.94 36 0.00714565 15 0.002007
784.94 26 0.005160747 20 0.002676
804.94 20 0.003969806 19 0.002543
824.94 20 0.003969806 10 0.001338
844.94 28 0.005557728 14 0.001873
864.94 21 0.004168296 12 0.001606
884.94 17 0.003374335 19 0.002543
904.94 17 0.003374335 11 0.001472
924.94 19 0.003771315 15 0.002007
944.94 25 0.004962257 10 0.001338
964.94 12 0.002381883 11 0.001472
984.94 8 0.001587922 7 0.000937
1004.94 6 0.001190942 15 0.002007
1024.94 5 0.000992451 10 0.001338
1044.94 9 0.001786413 9 0.001204
1064.94 12 0.002381883 12 0.001606
1084.94 7 0.001389432 13 0.00174
1104.94 6 0.001190942 9 0.001204
1124.94 3 0.000595471 9 0.001204
-
8/4/2019 Nuffield Bursary 19
64/131
1144.94 4 0.000793961 15 0.002007
1164.94 2 0.000396981 10 0.001338
1184.94 1 0.00019849 10 0.001338
1204.94 3 0.000595471 9 0.001204
1224.94 4 0.000793961 10 0.001338
1244.94 0 0 5 0.000669
1264.94 3 0.000595471 14 0.001873
1284.94 0 0 3 0.000401
1304.94 1 0.00019849 12 0.001606
1324.94 1 0.00019849 14 0.001873
1344.94 3 0.000595471 5 0.000669
1364.94 0 0 9 0.001204
1384.94 0 0 9 0.001204
1404.94 0 0 5 0.000669
1424.94 0 0 3 0.000401
1444.94 0 0 6 0.000803
1464.94 0 0 9 0.001204
1484.94 0 0 8 0.001071
1504.94 0 0 6 0.000803
1524.94 0 0 7 0.000937
1544.94 0 0 6 0.000803
1564.94 0 0 5 0.000669
1584.94 0 0 13 0.00174
1604.94 0 0 7 0.000937
1624.94 0 0 8 0.001071
1644.94 0 0 5 0.000669
1664.94 0 0 12 0.001606
1684.94 0 0 4 0.000535
1704.94 0 0 2 0.000268
1724.94 0 0 3 0.000401
-
8/4/2019 Nuffield Bursary 19
65/131
1744.94 0 0 5 0.000669
1764.94 0 0 9 0.001204
1784.94 0 0 5 0.000669
1804.94 0 0 6 0.000803
1824.94 0 0 7 0.000937
1844.94 0 0 10 0.001338
1864.94 0 0 7 0.000937
1884.94 0 0 6 0.000803
1904.94 0 0 2 0.000268
1924.94 0 0 6 0.000803
1944.94 0 0 8 0.001071
1964.94 0 0 6 0.000803
1984.94 0 0 9 0.001204
2004.94 0 0 4 0.000535
2024.94 0 0 5 0.000669
2044.94 0 0 4 0.000535
2064.94 0 0 8 0.001071
2084.94 0 0 9 0.001204
2104.94 0 0 4 0.000535
2124.94 0 0 5 0.000669
2144.94 0 0 5 0.000669
2164.94 0 0 7 0.000937
2184.94 0 0 3 0.000401
2204.94 0 0 5 0.000669
2224.94 0 0 6 0.000803
2244.94 0 0 8 0.001071
2264.94 0 0 7 0.000937
2284.94 0 0 6 0.000803
2304.94 0 0 2 0.000268
2324.94 0 0 7 0.000937
-
8/4/2019 Nuffield Bursary 19
66/131
2344.94 0 0 3 0.000401
2364.94 0 0 2 0.000268
2384.94 0 0 2 0.000268
2404.94 0 0 4 0.000535
2424.94 0 0 4 0.000535
2444.94 0 0 3 0.000401
2464.94 0 0 4 0.000535
2484.94 0 0 2 0.000268
2504.94 0 0 1 0.000134
2524.94 0 0 4 0.000535
2544.94 0 0 4 0.000535
2564.94 0 0 3 0.000401
2584.94 0 0 1 0.000134
2604.94 0 0 2 0.000268
2624.94 0 0 0 0
2644.94 0 0 3 0.000401
2664.94 0 0 6 0.000803
2684.94 0 0 4 0.000535
2704.94 0 0 3 0.000401
2724.94 0 0 2 0.000268
2744.94 0 0 9 0.001204
2764.94 0 0 3 0.000401
2784.94 0 0 1 0.000134
2804.94 0 0 2 0.000268
2824.94 0 0 2 0.000268
I constructed a histogram from this data using the percentages greater than0.1%, this meant a range of uptake values of -1015.06 to 424.94.
Here is the resulting graph.
-
8/4/2019 Nuffield Bursary 19
67/131
As you can see the post tumour uptake values dont continuously increase at thestart of the graph, instead they fluctuate, so I decided to change my start pointon the X axis to -75.06 as to get a graph which my accurately represented thepeaks. I also extended the graph to 524.94 to get more pre dose of Duramycinvalues.
Here is the result.
As you can see the peaks are better represented, however I still though the first100 or so uptake values were not useful so I changed the start of X values to84.94.
Here is the final histogram.
-
8/4/2019 Nuffield Bursary 19
68/131
-
8/4/2019 Nuffield Bursary 19
69/131
As you see these graphs look like the histogram of the second animal opposed tothe first.
I had one more animal left, and to make sure the similarities in the histogramsfor animals 2 and 3 were more than a coincidence, I decided to undergo thesame process using this animal 4s data.
Animal 4- using tool
I used the same process as animal 2 and 3 to construct animal 5s table andgraph.
Here is the table.
Uptake pre pre% post post%
-70 1 0.000293207 1 0.000293066
-50 1 0.000293207 11 0.003223727
-30 11 0.003225277 27 0.007912784
-10 29 0.008503002 63 0.018463162
10 65 0.019058454 167 0.048942032
30 165 0.048379152 335 0.098177129
50 365 0.107020548 769 0.225367798
70 789 0.231340308 1647 0.482679796
90 1706 0.500211109 3231 0.94689643
-
8/4/2019 Nuffield Bursary 19
70/131
110 3333 0.977258867 5625 1.648496571
130 5794 1.698841246 9449 2.769181173
150 9663 2.833259054 14465 4.239200516
170 14709 4.312781479 20701 6.066760448
190 21049 6.171713736 27537 8.070160014
210 27805 8.152620098 33027 9.679092667
230 33353 9.779332426 36504 10.69808335
250 36621 10.73753284 37149 10.88711095
270 36971 10.84015528 34276 10.04513217
290 34145 10.01155236 29420 8.6220034
310 29174 8.554020454 23237 6.809975969
330 22869 6.705350441 16924 4.95984995
350 16615 4.871633984 11521 3.376414044
370 11316 3.317930193 7232 2.119453725
390 7067 2.072093732 4150 1.216224137
410 4009 1.175466786 2364 0.692808159
430 2283 0.66939
top related