Room: Exhibit Hall | Forum 1
Purpose: Many clinical studies have shown that the diagnosis ability of radiologists for discovering lesions can be enhanced by combining cranio-caudal (CC) and medio-lateral oblique (MLO) view images, compared with utilizing one of them. Therefore, correlating corresponding position on two-view images has been an important step in breast cancer diagnosis. In this work, we proposed a method for registering same lesions on two-view mammograms.
Methods: The method could find a curve, which represents the corresponding position of a lesion on another view mammogram, if the position of the same lesion are known on one view mammogram. Our correlating method consists of five steps: 1) Restore the three-dimensional breast from two-view mammograms; 2) Select three feature points in three-dimension to present the space location of the known lesion; 3) Rotate the three-dimensional breast to simulate the direction change between CC view and MLO view; 4) Map three feature points onto another view image; 5) Interpolate between mapped feature points by quadratic equations to obtain a continuous curve.
Results: To evaluate the validity of the method, a data set of 266 pair of two-view images including 400 pair of lesions were established. Each pair of lesions with correspondences have been depicted on CC-view and MLO-view images by experienced radiologists. The experiments showed that the mean average error between the calculated curves and the lesion centers was about 3.4082Â±2.8265mm. For 95% detection sensitivity, the confidence interval of our method was Â±8.7741mm.
Conclusion: Our method can effectively correlate position of the same lesions on two-view mammograms by calculating a curve. It helps radiologists to discover lesions more accurately and rapidly when reading mammograms.
Funding Support, Disclosures, and Conflict of Interest: National Natural Science Foundation of China (81301940 and 81428019), National Key Research and Development Program (2016YFA0202003), Guangdong Natural Science Foundation of China (2016A030310388 and 2017A030313692), and Southern Medical University Startup fund (LX2016N003).
Not Applicable / None Entered.