From Multidisciplinary Computional Anatomy
- Patch-based Machine Learning and Deep Learning in Medical Image Processing, Analysis and Diagnosis
- Kenji Suzuki, Ph.D.(Associate Professor, Department of Electrical and Computer Engineering & Medical Imaging Research Center, Illinois Institute of Technology)
- Image processing and analysis, and computer-aids in diagnosis are indispensable in medical imaging. Machine leaning (ML) has become one of the most active areas of research in the medical imaging field, because “learning from examples or data” is crucial to handling a large amount of data (“Big data”) coming from medical imaging systems. Recently, as the available computational power increased dramatically, patch/pixel-based ML emerged in the medical imaging field, which uses pixel values in image patches directly, instead of features calculated from segmented objects as input information. Patch/pixel-based ML is a versatile, powerful framework that can acquire image-processing and analysis functions, including noise reduction, lesion and organ enhancement, pattern separation, segmentation, and classification, through training with image examples. On the other hand, in the computer vision field, deep learning that has a deep architecture drew enthusiastic attentions. Deep learning learns high-level computer-vision tasks from patches in images, which has a close relationship with patch/pixel-based ML. In this talk, patch/pixel-based ML in medical image processing and computer-aided diagnosis (CAD) of lesions in medical images is overviewed, including separation of bones from soft tissue in chest radiographs, radiation dose reduction in CT, lung nodule detection in chest radiography and CT, polyp detection in CT colonography, and detection of liver tumors in hepatic CT and MRI. Relationships between patch/pixel-based ML and deep learning (such as deep neural networks) are also discussed.
- 教授 佐藤 嘉伸
- yoshi @ is.naist.jp