Multi-modal image analysis method and system for cancer diagnosis

Abstract

The invention provides a multi-modal image analysis method and system for cancer diagnosis, and relates to the technical field of image data analysis. The method comprises the steps: extracting a suspected lesion region through a feature pyramid network, carrying out the classification through a false positive attenuation network based on context features, and extracting a true positive ROI. According to the invention, the classification neural network based on the residual network is adopted, the pathological type is stably and accurately predicted, doctors are assisted in improving the screening efficiency, missed diagnosis and misdiagnosis are effectively reduced, and medical image-based cancer diagnosis and reasonable treatment of primary hospitals are facilitated.

Description

technical field [0001] The present invention relates to the technical field of image data analysis, in particular to a multimodal image analysis method and system for cancer diagnosis. Background technique [0002] Early cancer screening and treatment has become a major health issue worldwide, and imaging examinations are the main medical means for early detection and treatment of breast cancer, thyroid cancer and other cancers. Currently commonly used imaging techniques include mammography, ultrasound, MRI, etc., which can not only realize the location of lesions but also observe and evaluate the entire tumor tissue, such as location, number, size, shape, edge shape, density, internal echo, signal or degree of enhancement. Doctors use the observation of medical imaging pictures to determine the location and judgment of cancer foci and further evaluate the pathological types of tumors as much as possible. However, in the field of medical imaging, there is a large gap betwee...

AI Technical Summary

Problems solved by technology

However, the current radiomics analysis of breast cancer is combined with a single high-frequency ultrasound or X-ray molybdenum palladium or magnetic resonance (MRI) imaging examination mode, and the accuracy of breast cancer pathological classification is still low. Poor outcome, often requires further invasive histopathological examination

However, biopsy or postoperative pathological detection only takes part of the tumor sample tissue, ignoring the possible heterogeneity of the tumor, especially the larger mass tissue, which has certain limitations.

Images