Histopathologic Cancer Detection Based on Deep Multiple Instance Learning

摘要

Advances in imaging technology have led to a consistent improvement in image resolution. Histopathology image, one type of high-resolution image, can provide enormous information, which has a profound impact on the field of medical image analysis. However, the histopathology image is too complex to analyze. In this paper, deep multiple instance learning that integrates the deep convolutional neural network and the multiple instance learning is adopted to solve the histopathologic cancer detection problems. The experiments illustrate that deep multiple instance learning method can obtain a higher accuracy than other methods.