Abstract #0594

Automatic Search in Breast MRI Dataset For Detection of Suspicious Lesions Using Mask R-CNN

Yang Zhang¹, Kai-Ting Chang¹, Siwa Chan², Peter Chang¹, Daniel Chow¹, Jeon-Hor Chen^1,3, and Min-Ying Lydia Su¹

¹Department of Radiological Sciences, University of California, Irvine, CA, United States, ²Department of Medical Imaging, Taichung Tzu-Chi Hospital, Taichung, Taiwan, ³Department of Radiology, E-Da Hospital and I-Shou University, Kaohsiung, Taiwan

A Mask R-CNN algorithm was implemented to search the entire dataset of breast MRI to identify suspicious lesions for further diagnosis. A total of 102 patients with confirmed cancer were analyzed. There were a total of 2,314 positive cases (i.e. imaging slices containing lesion); and 8,512 slices without lesion as negative cases. The search results show 1,943 true positives; 6,149 true negatives; 2,363 false positives; and 371 false negatives, with sensitivity 0.83, specificity 0.72, and the overall detection accuracy 0.75. The Dice Similarity Coefficient of the tumor segmented in the detection box compared to ground truth is 0.84.

This abstract and the presentation materials are available to members only; a login is required.

Join Here