Training-obtaining method of image detection learning network, image detection device, and medium

Abstract

The invention relates to a training-obtaining method of an image detection learning network, an image detection device, and a medium. The training method comprises following steps: constructing a first learning network and a second learning network, wherein the parameter number of the second learning network is less than the parameter number of the first learning network; utilizing a medical imagetraining data set to train the first learning network by a processor so as to obtain a trained first learning network; based on the learning result of the first learning network, utilizing the medical image training data set to train the second learning network by the processor; and utilizing the trained second learning network to detect medical images. The obtained second learning network can take both calculation precision and calculation efficiency into account in medical image detection.

Description

[0001] cross reference [0002] This application claims priority to U.S. Provisional Application No. 62 / 650,268, filed March 29, 2018, the entire contents of which are hereby incorporated by reference. technical field [0003] The present disclosure generally relates to image processing and analysis. More specifically, the present disclosure relates to a training method and an acquisition method of a learning network for medical image detection, as well as a medical image detection device and a non-transitory computer-readable medium on which a corresponding program is stored. Background technique [0004] With the development of medical technology, image acquisition devices can acquire medical images with richer details and higher resolution, which can provide physicians with more comprehensive image information. Early identification of lesions such as malignancies, invasive cancers, etc. in higher resolution medical images can lead to timely treatment and significantly re...

AI Technical Summary

Problems solved by technology

However, accurate delineation of tumor regions and core identification is time-consuming and error-prone for pathologists

Although computer-aided diagnosis (CAD) methods have been introduced in an attempt to assist pathologists in their diagnostic tasks, however, medical images such as WSI can have image resolutions as high as 200,000 × 100,000 pixels, and considering the computational load, traditional CAD methods can only handle Small area of ​​WSI

[0005] Recently, although deep learning methods (such as but not limited to using GoogleNet as a detector, using recurrent neural networks, or using model integration (that is, several Inception V3 models), etc.) have been introduced to detect large-scale medical images such as WSI analysis, but the existing methods are difficult to balance calculation accuracy and calculation efficiency (calculation load, calculation speed and cost)

Images