qa in digital mammography: local activities and remote control h. bosmans et al

55
QA in digital mammography: local activities and remote control H. Bosmans et al.

Upload: eustacia-fisher

Post on 23-Dec-2015

217 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: QA in digital mammography: local activities and remote control H. Bosmans et al

QA in digital mammography: local activities and remote control

H. Bosmans et al.

Page 2: QA in digital mammography: local activities and remote control H. Bosmans et al

2

Belgium, 1996

• Role of radiographers in QC minimal• Enthusiasm of radiologists for QC: minimal• No physicists working in X-ray imaging

• EC-guided screening made the difference

Page 3: QA in digital mammography: local activities and remote control H. Bosmans et al

3

Recall film-screen-mammography…

Page 4: QA in digital mammography: local activities and remote control H. Bosmans et al

4

Page 5: QA in digital mammography: local activities and remote control H. Bosmans et al

5

Constancy in film-screen mammography

• AEC: long term reproducibility• AEC: object thickness and tube voltage

compensation• sensitometry: base and fog• sensitometry: speed• Sensitometry: Contrast & Gradient• Imge quality: artefacts• Image quality: spatial resolution, reference ROI• Image quality: threshold contrast visibility

Page 6: QA in digital mammography: local activities and remote control H. Bosmans et al

6

Centrally supervised performance tests

Page 7: QA in digital mammography: local activities and remote control H. Bosmans et al

7

Centrally supervised performance tests

Page 8: QA in digital mammography: local activities and remote control H. Bosmans et al

8

Page 9: QA in digital mammography: local activities and remote control H. Bosmans et al

9

Digital imaging: 10th birthday• It is another world, with other challenges• We stayed with centrally supervised QC

– Long term reproducibility– Detector homogeneity– Uncorrected defective DELs (DR)– Uncorrected defective DELs (DR)– Display & printer: Geometrical distortion (CRT)– Display & printer: Contrast visibility– Display & printer: Displaying artefacts

Page 10: QA in digital mammography: local activities and remote control H. Bosmans et al

10

Data from DICOM header & global scoreFrom 2 ‘FOR PROCESSING’ flood images:

• DICOM header is scrutinized for kV, mAs, anode/filter, detector temperature, detector ID, thickness & compr force, MGD, detector calibration date, …

• Follow up in time

• Comparison between DICOM headers possible

• Limiting values set

Page 11: QA in digital mammography: local activities and remote control H. Bosmans et al

11

Screenshot of our platform

Automated constancy check in digital mammography: implementation and first results of a multi-center study J. Jacobs , K. Lemmens , F. Shannoun , G. Marchal and H. Bosmans, RSNA 2007

Page 12: QA in digital mammography: local activities and remote control H. Bosmans et al

12

Centrally supervised performance tests

Page 13: QA in digital mammography: local activities and remote control H. Bosmans et al

13

Data from DICOM header & global score

Page 14: QA in digital mammography: local activities and remote control H. Bosmans et al

14

Data from the images

From 2 ‘FOR PROCESSIING’ flood images:In homogenous segment:• Noise power spectrum (1D, radial or 2D NPS)

In reference ROI:• Pixel value, SNR, st dev, variance

Global image analysis:• Thumbnail images of PV, SNR and variance; color coded if

%-value• Automatic artefact detection & pixel value copying

Page 15: QA in digital mammography: local activities and remote control H. Bosmans et al

15

Example: Siemens system

Page 16: QA in digital mammography: local activities and remote control H. Bosmans et al

» Mean PV; SNR; Std. Dev.» calculate values from small

ROIs

2 mm * 100 µm 20 x 20px (GE DR) * 70 µm 28 x 28px (Siemens DR, Hologic DR) * 50 µm 40 x 40px (Fuji CR, Agfa CR)

Mean PV; SNR; Std. Dev.; Variance; Min PV; Max PV;

Median PV; Kurtosis; Skewness THUMBNAILS

Data from the images

Page 17: QA in digital mammography: local activities and remote control H. Bosmans et al

17

Data from the images

Page 18: QA in digital mammography: local activities and remote control H. Bosmans et al

Type 1DR: find out the service technician came along

Page 19: QA in digital mammography: local activities and remote control H. Bosmans et al

Type 1

Page 20: QA in digital mammography: local activities and remote control H. Bosmans et al

20

Scanning system (Philips Microdose) : normal situation

Type 2

Page 21: QA in digital mammography: local activities and remote control H. Bosmans et al

21

Type 2

GE system: normal situation, be happy

Page 22: QA in digital mammography: local activities and remote control H. Bosmans et al

22

Type 2

Page 23: QA in digital mammography: local activities and remote control H. Bosmans et al

23

Type 2

Page 24: QA in digital mammography: local activities and remote control H. Bosmans et al

24

• CR system: Scan line artefact

Mean pixel value SNR Deviation SNR

Type 3

Page 25: QA in digital mammography: local activities and remote control H. Bosmans et al

25

Type 3

Page 26: QA in digital mammography: local activities and remote control H. Bosmans et al

26

Type 3

Page 27: QA in digital mammography: local activities and remote control H. Bosmans et al

27

CR: inhomogeneities

Type 4

Page 28: QA in digital mammography: local activities and remote control H. Bosmans et al

28

Type 4

Page 29: QA in digital mammography: local activities and remote control H. Bosmans et al

Agfa DM1000 DR

Type 4

Page 30: QA in digital mammography: local activities and remote control H. Bosmans et al

30

Latest example: artefact in the Ag filter

• Acquisition with 4cm of PMMA

• Acquisition with 7cm of PMMA, using Ag filter

Page 31: QA in digital mammography: local activities and remote control H. Bosmans et al

31

DR: calibration too quickly after a patient scan (ghost)

Type 5

Page 32: QA in digital mammography: local activities and remote control H. Bosmans et al

32

• A suboptimal start….

Type 5

Page 33: QA in digital mammography: local activities and remote control H. Bosmans et al

33

DR: Ghost artefact & small field used at calibration

Type 5

Page 34: QA in digital mammography: local activities and remote control H. Bosmans et al

Type 5

Page 35: QA in digital mammography: local activities and remote control H. Bosmans et al

Calibration phantom artefactCalibration needs 4 x rotation / flipping of phantom

Type 5

Page 36: QA in digital mammography: local activities and remote control H. Bosmans et al

36

Type 5: latest example

Page 37: QA in digital mammography: local activities and remote control H. Bosmans et al

37

Suggestions for improved mammo DQC

• CR: include all CR cassettes systematically in the DQC procedure• CR: Enforce completely filled in DICOM headers• CR&DR: Include all clinically used anode/filter combinations in

the DQC procedure• CR&DR: Work with NPS data, using averaging• Apply big data analysis techniques• Include QC (quality and dose) of the clinical image• Extra analysis of local variance ‘variations’Unchanged:2 homogenous acquisitions each day, with phantom rotated over 180° !!!!

Page 38: QA in digital mammography: local activities and remote control H. Bosmans et al

38

Daily QC could help prioritize the work!

(IWDM 2010, K. Michielsen) Hypothesis:“systems that show little or no deviation during DQC show basically unchanged results in the half-yearly quality control”• For 50 systems of 6 vendors (CR and DR)• 129 events: 2 half yearly tests and DQC• 74 events: 2 half yearly tests and DQC and same CDMAM

phantom• Data: the mean glandular dose (MGD) and signal-

difference to noise ratio (SDNR) for 2, 3, 4, 5, 6 and 7 cm of PMMA and small aluminum disk of 0.2 mm;

Page 39: QA in digital mammography: local activities and remote control H. Bosmans et al

39

• Declared unchanged if both SDNR & MGD did not change by more than 10%

• A system was declared as unchanged between two half-yearly QC tests if the contrast threshold did not increase by more than 15%.

• The DQC results were declared as ‘unchanged’ between these two points if the following criteria were met: – no change in anode/filter combination, – a maximum change of the average nominal kVp of 1.0, – less than 5% change in SNR, – less than 10% change in mAs– no change of detector ID.

Page 40: QA in digital mammography: local activities and remote control H. Bosmans et al

40

Page 41: QA in digital mammography: local activities and remote control H. Bosmans et al

41

Part 2

Page 42: QA in digital mammography: local activities and remote control H. Bosmans et al

42

Daily QC of monitors

• Obligation• To be done with a variable pattern

Page 43: QA in digital mammography: local activities and remote control H. Bosmans et al

12345

43

MoniQA - pattern

J. Jacobs, J Kotre, … , Med Phys. 2007 Jul;34(7):2744-58

Page 44: QA in digital mammography: local activities and remote control H. Bosmans et al

Luminance checkGeometric check

Resolution check

General artefacts

Page 45: QA in digital mammography: local activities and remote control H. Bosmans et al

45

Page 46: QA in digital mammography: local activities and remote control H. Bosmans et al

46

Page 47: QA in digital mammography: local activities and remote control H. Bosmans et al

47

Page 48: QA in digital mammography: local activities and remote control H. Bosmans et al

48

Page 49: QA in digital mammography: local activities and remote control H. Bosmans et al

49

• It can all be poor.. Yet we don’t see deviations in half yearly test. Is it a reading test for operators?

Page 50: QA in digital mammography: local activities and remote control H. Bosmans et al

50

Results from data analysis

1. Barco MDMG 5121 (N=89) 10. Eizo GS510 (N=108)

2. Barco MDMG 5221 (N=2) 11. Eizo GS520 (N=64)

3. Barco MDMG 5212 (N=1) 12. Eizo GS521 (N= 52)

4. Barco MDNG 5121 (N=164) 13. Eizo GX530 (N=2)

5. Barco MDNG 6121 (N=66) 14. Eizo GX1030 (N=3)

6. Barco MFGD 5421 (N=38) 15. Eizo G51-BLS (N=64)

7. Barco MFGD 5620 (N=6) 16. GE (N=2)

8. Barco MFGD 5621 (N=100) 17. Siemens SMD (N=62)

9. Barco E-5620 (N=6)

Page 51: QA in digital mammography: local activities and remote control H. Bosmans et al

51

• Mean score of all monitors:

Dag van de week Gemiddelde score Standaard deviatie Maandag 97.37 6.979 Dinsdag 97.02 8.511

Woensdag 97.16 7.527 Donderdag 97.24 7.527

Vrijdag 97.27 7.023 Weekend 97.87 4.032

Page 52: QA in digital mammography: local activities and remote control H. Bosmans et al

52

• Half yearly tests and daily QC tests are hardly correlating. ρ(652) = 0.107, p < 0.01.

• These tests may be testing other aspects of the imaging chain.

Page 53: QA in digital mammography: local activities and remote control H. Bosmans et al

53

Suggestions for improved monitor DQC

Make a very simple daily QC testKeep present test on weekly basis

– Not acceptable in euref protocol– Therefore not acceptable for application in

Belgium

Page 54: QA in digital mammography: local activities and remote control H. Bosmans et al

Conclusion

It is possible to organize qualityQuality control is not necessarily very time consuming !

Page 55: QA in digital mammography: local activities and remote control H. Bosmans et al

Acknowledgement

• The Leuven QA team & all our students• The Leuven mammography and LUCK• J Jacobs, Qaelum NV (www.qaelum.com)• The OPTIMAM project• The EUREF team