A lesion image detection device, method and computer-readable storage medium

Abstract

The present invention provides a lesion image detection device. The lesion image detection device includes: an image acquisition module, which acquires a medical image to be detected; a detection module, which detects a lesion on the medical image according to a trained neural network model; wherein, The neural network model includes M-level sequentially cascaded neural network structures. During training, the negative samples input to the second-level to M-th-level neural network structures include false positive samples of the upper-level neural network structure, and M is greater than A positive integer of 1.

Description

technical field [0001] The present invention mainly relates to the field of medical images, in particular to a lesion detection device, method and computer-readable storage medium. Background technique [0002] Cancer is one of the most harmful diseases to human health today. Early diagnosis and treatment of many cancers can prevent and treat them, or increase the 5-year survival rate of patients. With the popularity of medical imaging equipment and the improvement of national health awareness, more and more people choose medical imaging (such as low-dose CT) for early cancer screening. Early cancers often appear as small lesions on imaging, and it takes a lot of time for doctors to screen them. With the popularization of medical imaging equipment, medical imaging data has increased exponentially, while the number of radiologists has grown slowly, which has led to a substantial increase in the workload of doctors, and the demand for computer-aided detection (Computer-Aided ...

AI Technical Summary

Problems solved by technology

The existing deep learning CAD system has significantly improved the lesion detection rate, but there are still the following problems: 1) The current CAD system is developed for a single lesion, which is not universal and cannot be applied to other Lesion detection problems on lesions or other modalities; 2) The current CAD system can only detect lesions of a single scale

Using a single-scale deep learning network to detect such lesions, the trained network often only has strong sensitivity to lesions of a certain scale, and cannot be applied to the detection of lesions of different sizes.

3) While the current CAD system improves the detection rate of lesions, there are a large number of false positives, which reduces the user experience of clinical use

Images