HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text Full Text (PDF) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Petrick, N. Articles by Hadjiiski, L. M. Search for Related Content PubMed PubMed Citation Articles by Petrick, N. Articles by Hadjiiski, L. M.

Breast Cancer Detection: Evaluation of a Mass-Detection Algorithm for Computer-aided Diagnosis—Experience in 263 Patients¹

¹ From the Department of Radiology, University of Michigan Medical Center, CGC B2102, Box 0904, 1500 E Medical Center Dr, Ann Arbor, MI 48109-0904. From the 2001 RSNA scientific assembly. Received June 18, 2001; revision requested August 8; revision received November 7; accepted January 7, 2002. Supported by USPHS grant CA 48129 and research grant DAMD 17-96-1-6254 from the U.S. Army Medical Research and Materiel Command. N.P. supported by the Whitaker Foundation and USPHS grant CA 79943. B.S. supported by Career Development Award DAMD 17-96-1-6012. L.M.H. supported by USAMRMC grant DAMD 17-98-1-8211. Address correspondence to N.P. (e-mail: petrick@umich.edu).

View larger version (40K): [in a new window]   Figure 1. Histogram summarizes the subtlety of the lesions observed on the 138 mammographic pairs and the 142 mammographic pairs obtained from the group 1 and group 2 databases, respectively, as ranked by the radiologist reviewing the cases. Each mass on each mammogram was rated independently by the radiologist. For comparison purposes, the plot is of the percentage of masses falling within each category. The total number of masses in each group can be found in Table 1.

View larger version (49K): [in a new window]   Figure 2. Block diagram of the mass-detection system evaluated in this study. DWCE = density-weighted contrast enhancement.

View larger version (44K): [in a new window]   Figure 3. FROC performance curves for group 1 calculated on the basis of scoring of individual masses. Per case and per mammogram performance curves for detection of both malignant and benign masses are depicted. The curves show the true-positive (TP) fraction achievable for a large range of mass marker rates. It is evident that the performance of the algorithm in the group 1 cases was better in detection of malignant versus benign masses, with an approximately constant difference in the true-positive fractions between the two throughout the entire CAD marker range plotted.

View larger version (45K): [in a new window]   Figure 4. FROC performance curves for group 2 calculated on the basis of scoring of individual masses. Per case and per mammogram performance curves for the detection of both malignant and benign masses are depicted. The curves show the true-positive (TP) fraction achievable for a large range of mass marker rates. The performance of the algorithm in group 2 cases is again better in detection of malignant versus benign masses, but the difference between the two is not constant as a function of CAD marker rate. It is evident that overall performance of the algorithm in group 2 benign masses is much worse than its performance in group 1 benign masses, while the performance difference between its assessment of group 1 malignant masses and its assessment of group 2 malignant masses is small.

View larger version (46K): [in a new window]   Figure 5. FROC performance curves for group 1 and group 2 calculated on the basis of grouped mass scoring. Per case and per mammogram performance curves for detection of malignant masses are depicted. These curves show how the true-positive (TP) fraction varies as a function of marker rate for group scoring, which is expected to be the most clinically relevant measure of algorithm performance. The algorithm showed consistent malignant mass-detection performance for both independent test sets over a wide range of marker rates.

View larger version (44K): [in a new window]   Figure 6. Combined group 1 and group 2 FROC performance curves in spiculated and nonspiculated masses calculated on the basis of scoring of individual masses. The benign spiculated mass curve is not shown because of the small number of cases in this category. The curves indicate that the CAD algorithm was more effective in detecting spiculated masses than it was in detecting nonspiculated ones. TP = true-positive.