Kevin S. Berbaum

Breast MRI lesion classification: improved performance of human readers with a backpropagation neural network computer-aided diagnosis (CAD) system.

To develop and test a computer‐aided diagnosis (CAD) system to improve the performance of radiologists in classifying lesions on breast MRI (BMRI).

Sonography of acute appendicitis in pregnancy

Background:Clinical evaluation of acute appendicitis is difficult in pregnant patients. Delay in diagnosis is associated with increased fetal mortality. The purpose of our study was to assess the value of sonography in the diagnosis of acute appendicitis in pregnant women.Methods:We obtained sonograms in 22 pregnant women suspected of acute appendicitis. All sonograms were performed using graded-compression to detect an enlarged appendix. The sonographic criteria for acute appendicitis were detection of a noncompressible blindended and tubular multilayered structure of maximal diameter greater than 6 mm.Results:The sonographic findings were correlated with surgical findings in seven cases and clinical follow-up in 15 cases. Acute appendicitis was diagnosed by sonography in three of 22 patients, and in all but one was confirmed by surgical and pathologic findings. In the remaining 19 patients, 15 improved on clinical follow-up; three were shown to have a normal appendix at surgery and one had focal acute inflammation at the tip of the appendix.Conclusions:Our experience suggests that graded-compression sonography is a useful procedure in pregnant patients suspected of acute appendicitis and has a similar accuracy as in nonpregnant women, especially in the first and second trimester.

CT analysis of metastatic neoplasms of the kidney. Comparison with primary renal cell carcinoma.

The CT findings in 32 patients with pathologically proven metastases to the kidney were compared to findings in 74 patients with renal cell carcinoma. Fourteen CT criteria were chosen to describe and characterize the lesions and 2 radiologists evaluated the CT images retrospectively according to these criteria. Renal metastases were characterized as small, multiple, bilateral, wedge-shaped, less exophytic, and located within the renal capsule. Renal cell carcinomas were single, unilateral, nonwedge-shaped, and exophytic, and easily transgressed the renal capsule. The sensitivity of CT to discriminate renal cell carcinoma from renal metastasis was 93.2% for renal cell carcinoma, and to discriminate renal metastasis from renal cell carcinoma was 75.0% for renal metastases by computer posterior probabilities. This study indicates that CT is useful for distinguishing these clinically important tumors. By using posterior probability, some unnecessary biopsies may be avoided.

Role of Faulty Visual Search in the Satisfaction of Search Effect in Chest Radiography

RATIONALE AND OBJECTIVES The authors tested the hypothesis that satisfaction of search effect, which is associated with the failure to detect native chest abnormalities in the presence of simulated nodules, is caused by reduced gaze on the native abnormalities. MATERIALS AND METHODS Gaze dwell time of 20 radiologists was recorded for the region around abnormalities on images. Ten radiographs were reviewed, nine of which contained native abnormalities. Each image was seen with and without a simulated nodule. RESULTS The decrease in the rate of true-positive findings in the detection of native abnormalities on images that contained simulated nodules confirmed the occurrence of a satisfaction of search effect. Gaze times on native abnormalities (up to the time of report of the abnormalities) were the same for images with nodules in which native abnormalities were missed (gaze time, 9.4 seconds) as they were for images without nodules in which native abnormalities were detected (gaze time, 9.5 seconds). Gaze time on missed native abnormalities was not affected by the presence (7.80 seconds) or absence (7.45 seconds) of nodules. CONCLUSION Reduction in gaze dwell time on the missed abnormalities is not the cause of satisfaction of search errors in chest radiographs.

Does a Normal Small-Bowel Enteroclysis Exclude Small-Bowel Disease? A Long-Term Follow-Up of Consecutive Normal Studies

The purpose of our study was to determine if a normal small-bowel enteroclysis excludes small-bowel disease in adult patients, using long-term follow-up as the major reference standard. We reviewed 193 consecutive small-bowel enteroclysis (SBE) studies completed during a period from January 1987 to February 1989, of which 83 were judged to be normal at the time of the study. Eight of these latter patients were excluded due to inadequate follow-up. The indications included detection of gastrointestinal bleeding, small-bowel obstruction, Crohns disease, nonspecific abdominal pain, chronic diarrhea, and a miscellaneous group. Each patient was followed for at least 3 years by chart review or until a definite diagnosis was established. Six of the 75 patients whose SBE was originally interpreted as normal were eventually judged to have small-bowel disease. The remaining 69 patients were judged to be free of small-bowel disease by autopsy, surgical laparotomy, endoscopie observation or biopsy, or long-term follow-up for at least 3 years. Therefore, a normal SBE correctly excluded small-bowel disease in 69 of our 75 patients (true negatives) and failed to diagnose disease in six patients (false negatives), for a specificity of.92±.03 (SE). In this experience, SBE was sufficiently specific in most patients to exclude small-bowel disease.

Small bowel enteroclysis survey

A survey questionnaire was sent to 125 radiology programs to determine the frequency and indications for the use of small bowel enteroclysis. Sixty-seven of the 84 respondents perform small bowel enteroclysis while 17 do not. The 3 most common indications for enteroclysis are detection of unexplained gastrointestinal bleeding, partial small bowel obstruction, and Crohns disease. We conclude that although small bowel enteroclysis is a well-established examination, there are wide differences in the frequency with which it is performed, the techniques of examination, and indications for its use.

CT Analysis of Metastatic Neoplasms of the Kidney

The CT findings in 32 patients with pathologically proven metastases to the kidney were compared to findings in 74 patients with renal cell carcinoma. Fourteen CT criteria were chosen to describe and characterize the lesions and 2 radiologists evaluated the CT images retrospectively according to these criteria. Renal metastases were characterized as small, multiple, bilateral, wedge-shaped, less exophytic, and located within the renal capsule. Renal cell carcinomas were single, unilateral, nonwedge-shaped, and exophytic, and easily transgressed the renal capsule. The sensitivity of CT to discriminate renal cell carcinoma from renal metastasis was 93.2% for renal cell carcinoma, and to discriminate renal metastasis from renal cell carcinoma was 75.0% for renal metastases by computer posterior probabilities. This study indicates that CT is useful for distinguishing these clinically important tumors. By using posterior probability, some unnecessary biopsies may be avoided.

Densitometric analysis of eccentric vascular stenoses: Comparison of CO2 and iodinated contrast media

Rationale and Objectives. The authors compared the accuracies of CO 2 and iodinated contrast material in the densitometric quantification of eccentric vascular stenoses. Methods. Five precision-machined eccentric phantom stenoses of 50%, 60%, 70%, 80%, and 90% cross-sectional area narrowing were integrated into a pulsatile ex vivo flow model, imaged with digital subtraction angiography (DSA), and analyzed with densitometry. Relationships between the actual and measured (densitometric) degree of cross-sectional area narrowing were evaluated by using linear regression analysis and paired Student t tests. Comparison measurements were obtained in en face and profile projections. In addition, the effect of iodinated contrast material concentration was evaluated over a range of dilutions (47–282 mg iodine per milliliter). Results. CO 2 yielded significantly more accurate results than did iodinated contrast material (282 mg iodine per milliliter) in the 50%, 60%, and 70% stenosis models when imaging was performed en face ( P 2 DSA when imaging in profile (slope = 0.91, intercept = 2.42% actual stenosis, r = .99). The accuracy of densitometric stenosis estimation was inversely related to the concentration of iodinated contrast material. Conclusion. CO 2 DSA densitometry, under the conditions of these experiments, yields quantitative measures of relative cross-sectional area narrowing that are comparable with, and under some circumstances surpass, those obtained with iodinated contrast material-based DSA. In this model, CO 2 was more useful than iodinated contrast material in 50%–70% stenosis when imaging in the least-optimal plane of stenosis quantification, the en face projection.

Diagnostic accuracy of 3D color volume-rendered CT images for peroneal tendon dislocation in patients with acute calcaneal fractures

Background Use of three-dimensional (3D) color volume-rendered (VR) images has been reported to be more time-efficient compared to that of cross-sectional computed tomography (CT) images for the diagnosis of peroneal tendon dislocation. However, the diagnostic performance of this technique has not been studied. Purpose To test diagnostic accuracy of 3D color VR CT images of ankle for peroneal tendon dislocation in patients with acute calcaneal fractures. Material and Methods The study consisted of 121 ankle CT studies from 105 consecutive patients (85 men, 20 women; mean age, 42 years; age range, 16–75 years) with acute calcaneal fractures. Peroneal tendon dislocation was diagnosed on multiplanar CT images by consensus of two experienced musculoskeletal radiologists, which served as the reference standard. Three other musculoskeletal radiologists independently reviewed 3D images alone on a workstation. The readers determined whether or not there was peroneal tendon dislocation using three degrees of certainty (definite, probable, and possible). Diagnostic performance of 3D images for peroneal tendon dislocation was evaluated by calculating the sensitivities, specificities, and area under the receiver-operating characteristic (ROC) curves. Results Forty-eight (40%) out of 121 studies showed peroneal tendon dislocation based on the expert readings using multiplanar reformatted images. Sensitivities/specificities of 3D images measured 0.92/0.81, 0.88/0.90, and 0.81/0.92 for three readers, respectively. The area under the proper binormal ROC curve based on all three readers (0.93, 0.94, and 0.92) measured 0.93 with a 95% confidence interval of 0.89–0.98. Conclusion Diagnostic accuracy of 3D images is comparable to, but not as good as that of MPR images for the diagnosis of peroneal tendon dislocation in patients with acute calcaneal fractures.

Effect of fatigue on reading computed tomography examination of the multiply injured patient

Abstract. Our goal was to ascertain how fatigue affects performance in reading computed tomography (CT) examinations of patients with multiple injuries. CT images with multiple fractures from a previous study of satisfaction of search (SOS) were read by radiologists after a day of clinical work. Performance in this study with fatigued readers was compared to a previous study in which readers were not fatigued. Detection accuracy for obvious injuries was not affected by fatigue, but accuracy for subtle fractures was reduced (P=0.016). An SOS effect on decision thresholds was evident mirroring recent studies. Without fatigue, readers spent more time interpreting and reporting findings as the number of the injuries increased. When fatigued, readers did not increase reading time as fracture number increased. Without fractures, reading time for not-fatigued and fatigued readers was the same (P=0.493) but was significant (P=0.016) with an added subtle fracture. The difference increased with a major injury (P=0.003) and increased further with both a major injury and subtle fracture (P=0.0007). Fatigue and multiple abnormalities have independent effects on detection performance but do interact in determining search time.