CAD system uses deep learning to detect, segment, classify masses from mammograms
Researchers have developed a fully integrated computer-aided diagnosis (CAD) system that detects, segments and classifies masses from mammograms using deep learning and a deep convolutional neural network (CNN), according to a new study published by the International Journal of Medical Informatics.
The authors used a proven regional deep learning model, You-Only-Look-Once (YOLO), for detecting potential masses in the breast tissues of patient data taken from a publicly available dataset. Why YOLO? They provided a number of reasons in their analysis.
“First, YOLO has a robust ability to detect the masses directly from entire mammograms,” wrote author Mugahed A. Al-antari, PhD, of Kyung Hee University in Yongin, South Korea, and colleagues. “Second, detected bounding boxes via YOLO accurately align the masses, thereby, a low rate of false positives is achieved compared with other studies. Third, it can even detect challenging cases where the masses exist either over pectoral muscles or inside dense regions. Fourth, the running time of the testing and required memory are extremely low compared to other more complex deep learning models.”
Masses were then segmented using a new deep network model, full resolution convolutional network (FrCN).
“FrCN consists of two main consecutive encoder and decoder networks,” the authors wrote. “The encoder network involves thirteen convolutional layers. However, unlike the previous deep models, the max-pooling and sub-sampling layers are removed from the encoder network to preserve the full spatial resolution of the original input as well as the details of the objects. This is a key modification to avoid any information loss during feature map generation for accurate pixel-to-pixel mass segmentation.”
Finally, masses were classified as benign or malignant using a CNN.
Overall, the researchers reported that their fully integrated CAD system was a success. YOLO-based deep learning achieved “the best detection performance compared with the latest deep learning models,” FrCN provided high-quality segmentation and the overall accuracy of the CCN’s mass classification was more than 95 percent.
“The proposed CAD system based deep learning through detection, segmentation, and classification could be used for clinical applications to assist radiologists,” the authors concluded.
Michael has more than 18 years of experience as a professional writer and editor. He has written at length about cardiology, radiology, artificial intelligence and other key healthcare topics.