Based on the neural network image recognition algorithm in deep learning, the system is applied to the fields of radiology, pathology and dermatology, in which patients such as X-ray film, tumor cells and skin damage are identified, and a medical image interpretation service project for mobile Internet is built. The algorithm uses hybrid convolutional network (PCL) to learn the underlying translation invariant features, and iteratively applies convolutional neural network (CNN) to learn the different nonlinear features of the input image set in a hierarchical manner. The algorithm's gallery and query examples include sets of face or object images with different perspectives, backgrounds, facial expressions, resolutions, and illuminations. Cooperate with medical and health organizations and hospitals at all levels to escort people's health. The system is convenient, fast and intelligent. It can read films for users in real time. Daily health knowledge and health reminders can provide protection for people's health to a certain extent. Promote the development of social medical care, reduce workload and costs, improve work efficiency, help structured reports, liberate potential social productivity, save invalid time for registration and waiting for doctors to see films, and can do more meaningful and valuable things and liberate potential Social productivity, save social resources, and create more social value.