Abstract:
|
Practicing smart healthcare and using machine learning techniques to segment, track and classify diseases in medical images, thereby relieving clinicians from the heavy burden of image diagnosis and improving healthcare quality, have evolved into a trend. In particular, rapidly growing deep learning techniques such as transfer learning, YOLO, and U-Net may be combined with traditional machine learning methods and applied to medical images at varied data volume levels to broaden the spectrum of diagnostic techniques for medical images. This talk discusses practical clinical issues concerning atrial fibrillation, vertebral compression fracture, glaucoma, and intracranial metastasis as examples to explain how important machine learning techniques may be integrated with clinical medical data and professional clinicians’ accumulated experience. In the long run, the development of machine learning techniques to medical image analysis and diagnosis methods is expected to enhance the establishment of real time, accurate, and comprehensive clinical diagnoses.
|