detecting liver lesions in ct

8/13/2019 Detecting Liver Lesions in CT

http://slidepdf.com/reader/full/detecting-liver-lesions-in-ct 1/4

Gastrointestinal Imaging

William W. Mayo-Smith, MD

Himanshu Gupta, MD

Mark S. Ridlen, MD Jeffre y M . Brod y, M D

Nancy C. Clements, PhD

Joh n J. Cro na n, M D

Detecting Hepatic Lesions:

The Added Utility of CT LiverWindow Settings1

PURPOSE: To prospectively evaluate the utility of adding computed tomographic(CT) liver windows to conventional soft-tissue windows for the detection of hepaticdisease.

MATERIALS AND METHODS: One of four radiologists experienced in abdominalimaging interpreted 1,175 consecutive abdominal CT scans from one institution.Hepatic images were first interpreted by using standard soft-tissue windows. Thenumber of lesions and confidence in lesion detection were recorded. The liver-window images were then interpreted in conjunction with the soft-tissue–windowimages, and thenumberof lesionsand confidencein detection were recorded again.

The proportion of patients in whom additional lesions were found by using liverwindows wasdetermined.

RESULTS: On soft-tissue–window and liver-window scansinterpreted together, 869(74%) patients had no hepatic lesions. Thirty-six (3.1%) patients had new lesionsseen with the addition of liver windows. Twelve of these 36 patients had no lesionsseen on soft-tissue–window scans. Twenty-six of the 36 patients with additionallesions seen had a history of neoplasm. There was a change in diagnosis in 1.7% of thepatientswith theaddition of liverwindowsand a changein recommendation forfollow-up in 0.85%.

CONCLUSION: Routine interpretation of liver-window scans for all abdominal CTscanshaslimited added utilityin detecting hepatic disease.

The use of dedicated computed tomographic (CT) liver win dows in addition to soft-tissue

windows is routine practiceat many institutions. Liver windows havea window level equal

to the attenuation level of hepatic parenchyma (50 HU without contrast material; 100 HU

after the intravenous administration of contrast material) and a narrower window width

(150 HU) than conventional soft-tissuewindows. In theory, thesesettingsoptimi zehepatic

lesion detection by optimizing gray-scale valuesto hepatic disease. By using a window level

centered on liver attenuation and a narrower window width, greater contrast can beattained in

the liver, thereby increasing hepatic lesion conspicuity. While interpretation of liver-window

scansis commonly performed, thereisarelativepaucityof datadocumentingan increased lesion

detection rate by using liver windows (1–4). Printing liver-window scans adds cost to the CT

examination, including film and processing costs, radiologist’s time, technologist’s time,

alternator space, and film storage. In practices where workstations are used for scan interpreta-

tion, routine interpretation of liver-window scans increases the radiologist’s interpretation time. The purpose of this study was to prospectively evaluate the usefulness of adding liver

windows to conventional soft-tissue windows for the detection of hepatic disease.

Liver-window scans were evaluated on the basis of (a ) the number of additional h epatic

lesion s detected, (b ) the level of confidence in lesion detection, and (c) whether the

diagnosisor recommendation for follow-up werechanged.

MATERIALS AND METHODS

Consecutive abdominal CT studies in 1,175 adult patients (649 women, 526 men; mean

age, 57 years; age range, 18–95 years) from one institution were prospectively analyzed

Index terms:

Computed tomography (CT), imagedisplayand recording

Computed tomography (CT), imageprocessing, 761.12111,761.12114, 761.12115

Liver neoplasms, CT, 761.12111,761.12114, 761.12115, 761.33

Liver neoplasms, secondary, 761.33

Radiology 1999; 210:601–604

1 From the Department of DiagnosticImaging, Brown University School of Medicine, Rhode Island Hospital, 593Eddy St, Providence, RI 02903. Fromthe 1997 RSNA scientific assembly.Received May 22, 1998; revision re-quested July 16; revision receivedSeptember 10; accepted October 6.Address reprint requests to W.W.M.S.

RSNA, 1999

See also the editorial by Brink (pp593–594) in thisissue.

Author contributions:

Guarantor of integrity of entire study,

W.W.M.S.; study concepts, W.W.M.S.,H.G.; study design, W.W.M.S., H.G.,M.S.R., J.M.B.,N.C.C., J.J.C.;definitionof intellectual content, W.W.M.S.; lit-eratureresearch, W.W.M.S., H.G.; clini-cal studies, W.W.M.S., H.G., M.S.R.,

J.M.B., J.J.C.; data acquisition,W.W.M.S., H.G., M.S.R., J.M.B., J.J.C.;data analysis, W.W.M.S., H.G., N.C.C.;statistical analysis, N.C.C.; manuscriptpreparation,W.W.M.S.;manuscript ed-iting and review, W.W.M.S., H.G.,M.S.R., J.M.B., N.C.C., J.J.C.

601



during an 8-month period. The CT

scans were obtained on four scanners

(HiSpeed Advantage or Hilight Advan-

tage [GE Medical Systems, Milwaukee,

Wis] or IQ Premier [Picker Interna-

tional, Cleveland, Ohio]). Therewere363

helical examinations, and 812 were non-

helical. For the helical examinations,

section collimation was 7 mm with apitch of 1. For the nonhelical ex-

aminations, collimation was 5–10 mm.

One thousand twenty-three examina-

tions were performed with the i ntrave-

nous administration of contrast ma-

terial, and 152 were obtained without

contrast material because it was not indi-

cated for the examin ation or because of

contrast material allergies. Contrast mate-

rial–enhanced helical scans were ob-

tained in the portal venous phase of

enhancement.

All scans were in terpreted prospec-

tively by one of four fellowship-trained

radiologists (W.W.M.S., M .S.R., J.M .B.,

J.J.C.) with expertise in abdominal imag-

ing. The liver was first evaluated by using

abdominal soft-tissue windows (win dow

level, 40 HU; window width, 400 HU).

Specific criteria evaluated were (a ) the

presence of a hepatic lesion graded on a

five-point ordinal degree of confidence

scale (with 1 representin g ‘‘lesion defi-

nitely not present’’ and 5 representing

‘‘lesion defin itely present’’), (b ) the num-

ber of hepatic lesions ranging from

zero to five and the number greater than

five, (c ) the presumed diagnosis, and

(d ) the recommendation for follow-up.

This information was entered on a data

sheet.

The CT scan was then reevaluated at

the same interpretation session with the

addition of liver windows(window level,50–100 HU; window width, 150 HU) to

the standard soft-tissue windows. The

four criteria listed above were reevalu-

ated, and this information was entered

on the second portion of the data sheet.

Initial data recorded for the soft-tissue

windows alone could not be changed

after the liver-window scans wereviewed.

One data sheet was filled out per pa-

tient. In addition, history, typeof CT scan

obtained (helical or nonhelical), use of

contrast material, and age and sex of the

patient were recorded for each patient.

The information wasentered into a com-

puter database by one author (H.G.).

Thedatawerethen analyzed for concor-

dance between interpretations of soft-

tissue–window scans alone versus those

of both soft-tissue–window and liver-

window scans. Agreement between the

ordinal measuresof degree of confidence

in lesion detection was estimated by

means of the weighted statistic. The

value is a measureof the degree to which

ratings with the two modalities exactly

agree, adjusted for the level of agreement

that would be expected if the two ratings

were in fact independent. A weighted

version of is a more useful measure

when thedataareordinal, aswith thefive

levels of confidence of lesion detection

used in this study. The weighted statis-

tic takes into account the fact that when

the data are ordinal, the level of disagree-

ment may be higher for two ratings that

are three categories apart compared with

two ratings that are onl y one category

apart.

A statistic equal to 1 indicates perfect

agreement, and a statistic equal to 0

indicates no moreagreement than would

be expected by chance alone (5,6). A P

value of .05 was considered to indicate a

statistically significant difference. A 2

test of in dependence was used to assess

the difference in lesion detection be-

tween readers.

a. b .

Figure 1. CT scans of a 73-year-old woman with pancreatic carcinoma (curved arrow), a liver metastasis, and malignant ascites. ( a) On thesoft-tissue–window scan, the metastasis in the posterior right lobe was missed prospectively. (b ) On thel iver-window scan, a lesion (straight arrow) isseen in the periphery of the right lobe.

Additional Lesions See n withLiver Windo ws

No. of LesionsSeen with Soft-

TissueWindows

No. of LesionsSeen with

Liver WindowsNo.of

Patients

0 1 121 2 4

2 3 20

602 • Radiology • March 1999 Mayo-Smith et al



RESULTS

The level of agreement between the two

interpretations was very high. The per-

centage of agreement observed between

interpretation of soft-tissue–window and

liver-window scans was 98% (1,155 of

1,175; 0.966). The use of liver win-

dows resulted in a changein diagnosis in

20(1.7%) of the 1,175patients. Therewas

a change in recommendation for fol-

low-up in only 10 (0.85%) of the 1,175

patients.

With the use of soft-tissue and liver

windows together, 869 patients had no

hepatic lesions. The same number of le-

sions were seen with soft-tissue and liver

windows in 270 patients. Thi rty-six pa-

tients had new hepatic l esions seen with

theaddition of liver windows(Table). The

mean size of the additional hepatic le-

sions detected was 8 mm (range, 5–20

mm), and th e majority were hypovascu-

lar. Of the 36 patients with additional

hepatic lesions seen, 26 had a known

primary neoplasm: colon (n 8), breast

(n 5), ovarian (n 2), renal (n 2),

pancreatic (n 1), esophageal (n 1),adrenal (n 1), lung (n 1), or cervical

(n 1) carcinoma; lymphoma (n 3); or

melanoma (n 1). The history for the

remaining 10 patients was Crohn disease

(n 1), pancreatitis (n 2), urinary tract

infection (n 1), renal calculus (n 1),

and screening of abdomen (n 5).

Twelve of the 36 patients had no le-

sions seen on soft-tissue–window scans

and one or more lesions seen with the

addition of l iver-window scans. Three of

these 12 patients had additional lesions

that weremasses. Onepatient had pancre-

atic carcinoma with encasement of thesuperior mesenteric vein (Fig 1), one pa-

tient had colon carcinoma with a lung

metastasis, and one patient had a mass

that was interpreted on liver-window

scans but that was not identified at subse-

quent in traoperative ultrasonography

(US), thatis, afalse-positiveresult. Four of

the 12 patients had very small lesions (5

mm or smaller) that were too small to

characterize (Fig 2). The small lesions

were stable at 1-year follow-up in two

patients, and the other two patients, who

werel ost to follow-up, had kidney stones

(n 1) and pancreatitis (n 1). Four of

the 12 patientshad focal fat diagnosed bymeans of negative follow-up examina-

tion results or confirmatory study.

Four of the 36 patients with additional

lesions hadonelesion seen on soft-tissue–

window scans and two or more lesions

seen with the addition of liver-window

scans. The h istory in these patients was

colon carcinoma (n 2), adrenal carci-

noma (n 1), and esophageal carcinoma

(n 1). Clinical management was altered

in one patient with a history of colon

carcinoma and arisinglevel of carcinoem-

bryonic antigen. Thi s patient underwent

exploratory laparotomy for attempted he-

patic resection but had multiplel esions at

intraoperative US that were missed on

both soft-tissue–window and liver-win-

dow scans. In the other three patients,

clini cal management did not change on

the basis of finding additional lesions

because th ese patients h ad extrahepatic

disease or were not candidates for resec-

tion.

In the cohort of patients with two to

five lesions seen on soft-tissue–window

a. b .

Figure 2. CT scans of a 54-year-old woman with ah istory of ovarian carcinoma with no residualdiseaseafter total abdominal hysterectomy, bilateral salpin go-oophorectomy, and chemotherapy.(a ) O n the soft-tissue–window scan, a small lesion in segment IV of the left lobe of the liver was

missed. (b ) On thel iver-window scan, the2-mm lesion (arrow) is seen. Thel esion wasstablei n size1 year later at follow-up CT. This presumably represents a small hepatic cyst seen better with liverwindows.

a. b .

Figure 3. CT scans of a 57-year-old woman with metastatic renal cell carcinoma to th e liver,brain, and lungs. (a ) On the soft-tissue–window scan, threemetastases were seen, and the fourthlesion was missed. (b ) On the liver-window scan, the fourth lesion (arrow) posterior to the portalvein is better seen.

Volume 210 • Number 3 Detecting Hepatic Lesions: Utility of CT Liver Window Settings • 603



scans, 20 had more lesions seen with the

addition of liver win dows (Fig 3). Twelve

of these patients had a history of carci-

noma: breast (n 5), colon (n 2), lung

(n 1), renal cell (n 1), and cervical

(n 1) carcinoma; lymphoma (n 1);

and melanoma (n 1).

The number of studiesi nterpreted and

additional lesionsdetected per reader were

as follows: Reader 1 interpreted 587 stud-ies and detected 27 additional lesions

(4.6%), reader 2 in terpreted 287 studies

and detected four additional lesions

(1.4%), reader 3 in terpreted 244 studies

and detected five additional lesions

(2.0%), and reader 4 interpreted 50 stud-

ies and detected no additional lesions

(0%). The difference in lesion detection

was higher in reader 1 than that in reader

2 when compared by using a 2 test of

independence (21 5.75; P .016), but

there was no significant difference be-

tween the other readers.

The histories of the patients enrolledwere as follows: screening of abdomen

(n 338); known primary carcinoma for

staging (n 425); trauma (n 142); rule

out abscess (n 131); rule out hemor-

rhage or aneurysm (n 68); pancreatitis

(n 45); and diverticulitis (n 26). No

additional lesions were seen by using

liver windows in patients with a history

of trauma or rule out abscess, hemor-

rhage, or aneurysm or diverticulitis.

DISCUSSION

This study has shown that liver windowshave limited added usefulness in the de-

tection of hepatic lesions. No hepatic

lesions were seen on both soft-tissue–

window and liver-window scans in 869

(74%) of the 1,175 patients, so the use of

liver windowsfor all abdominal CT scans

would be of no value in three-quarters of

our patients. Lesions were detected by

using soft-tissue–windows alone in 270

(88%) of 306 patients with hepatic le-

sions. Of the 36 patients with additional

lesions seen, 24 had one or more lesions

seen on soft-tissue–win dow scans alon e.

Only 12 patients with no lesions seen on

soft-tissue–win dow scans had lesionsseen

on l iver-window scans. This did not alter

patient treatment in the majority.

In the analysis of the referring history,

it is interesting to note that no patient

with a history of trauma or rule out

abscess, hemorrhage, or aneurysm or di-

verticulitis had additional benefit with

liver windows. Conversely, 26 of the 36

patients with additional lesions seen had

a known history of neoplasm. These data

suggest that li ver windows may be most

beneficial in patients undergoing CT forstaging of known carcinoma. This is a

potentially important cost-saving mea-

sure as the indications for abdominal C T

continue to expand.

Wh en assessing the added utili ty of

liver windows, it is important to look not

only at the additional number of lesions

seen but also at the clin ical context in

which this occurs. Thus, detecting addi-

tional lesions in a patient with multiple

metastases to other organs is less impor-

tant than when a single hepatic lesion is

missed by usingsoft-tissuewin dowsalone

in the absence of extrahepatic disease. This is particularly true in neoplastic con-

ditions that typically metastasize to mul-

tiple sites within an organ and to mul-

tiple organs, such as breast and lung

carcinoma. In the 26 patients with a

history of neoplasm and additional le-

sions seen on liver-window scans, 20 had

evidence of disease beyond the liver at

thetimeof CT.

There are several limitations to this

study. Al l scans were interpreted by on e

of four radiologists, so differencesi n inter-

pretation between readerswerenot strictly

accounted for. One reader h ad a slightly

higher detection ratethan theother three,but even the highest additional detection

rate of 4.6% is unlikely to be clinically

important. Another limitation of the

study is that because all readers knew the

purpose was to detect hepatic lesions,

they were diligent about looking for he-

patic diseaseby using soft-tissuewindows

alone. This may create bias in favor of

soft-tissue windows in this study. Con-

versely, interpretation of the liver-win-

dow scans after interpretation of the soft-

tissue–window scans may create bias in

favor of thel iver windows, as thereader is

reinterpretin g the same levels a second

time. These biases may offset one an-

other, but they are not easily controlled

or measured in this study. We attempted

to perform scan in terpretation as it most

commonly occursi n clinical practice. De-

spite these shortcomings, the data are

compelling that there is l imited useful-

ness to adding liver windows to all ab-

dominal CT examinations. Our data sug-

gest that theselectiveuseof liver windows

in the setting of known neoplasm could

provide cost savings without a compro-

misein patient care.With the constant pressure to reduce

examination costs, areas where savin gs

can be gained must be scrutinized, while

making sure that patient carei s not com-

promised. In radiologic practices that use

film, the cost of liver windows involves

not only the film and processing but the

radiologist’stime, technologist’stime,and

film movement and storage. These costs

can be substantial in a high-volume radi-

ology department. For practices that use

picturearchivingand communication sys-

tems for scan interpretation, physician

time increases the cost of routine use of liver windows. More selective use of liver

windows for patients with known malig-

nancy or a high clinical index of suspi-

cion of hepatic disease could result in

substantial savings.

Acknowledgm ents: The authors acknowledgethe assistance of all the CT technologists at ourinstitution who participated in this study. Inaddition, the authorsacknowledgeDebbieDes-

jardins for manuscript preparation.

References

1. FederleM P. Areli ver windowsnecessary inbody CT? (letter). AJR 1994; 163:1525.

2. Seltzer SE,J udyPF, Swensson RG,C han KH,Nawfel RD. Flattening of thecontrast detailcurve for l arge lesions on liver CT images.Med Phys 1994; 21:1547–1555.

3. Judy PF, Swensson RG, Nawfel RD, SeltzerSE. Contrast detail curves for liver CT. MedPhys 1992; 19:1167–1174.

4. Ishida M, Doi K, Loo LN, MetzCE, Lehr JL.Digital i mage processing: effect on detect-ability of simulated low contrast radio-graphic patterns. Radiol ogy 1984; 150:569–575.

5. Cohen J . A coefficient of agreement fornomi nal scales. Educ Psychol Meas 1960;20:37–46.

6. Cohen J. Weighted kappa: nominal scaleagreement with provision for scaled dis-agreement or partial credit. Psychol Bull1968;70:213–220.

604 • Radiology • March 1999 Mayo-Smith et al

detecting liver lesions in ct

Documents