A Semantic Segmentation Network Based Real-Time Computer-Aided Diagnosis System for Hydatidiform Mole Hydrops Lesion Recognition in Microscopic View
This addresses the challenge of missed or misdiagnosed hydatidiform mole lesions for pathologists, but it is incremental as it applies existing deep learning techniques to a specific medical domain.
The researchers tackled the problem of identifying hydatidiform mole hydrops lesions in microscopic images by developing a real-time computer-aided diagnosis system using a semantic segmentation network, achieving accurate labeling and real-time response in evaluations.
As a disease with malignant potential, hydatidiform mole (HM) is one of the most common gestational trophoblastic diseases. For pathologists, the HM section of hydrops lesions is an important basis for diagnosis. In pathology departments, the diverse microscopic manifestations of HM lesions and the limited view under the microscope mean that physicians with extensive diagnostic experience are required to prevent missed diagnosis and misdiagnosis. Feature extraction can significantly improve the accuracy and speed of the diagnostic process. As a remarkable diagnosis assisting technology, computer-aided diagnosis (CAD) has been widely used in clinical practice. We constructed a deep-learning-based CAD system to identify HM hydrops lesions in the microscopic view in real-time. The system consists of three modules; the image mosaic module and edge extension module process the image to improve the outcome of the hydrops lesion recognition module, which adopts a semantic segmentation network, our novel compound loss function, and a stepwise training function in order to achieve the best performance in identifying hydrops lesions. We evaluated our system using an HM hydrops dataset. Experiments show that our system is able to respond in real-time and correctly display the entire microscopic view with accurately labeled HM hydrops lesions.