Nuclear medicine multi-mode lesion image detection method and system

Abstract

The invention discloses a nuclear medicine multi-mode lesion image detection method and system. The method and system can effectively carry out the detection and integration processing of a lesion target image through the information of different modes, do not need extra multi-mode preprocessing work, and can comprehensively and completely reserve the information of each mode. After multi-scale feature fusion, the information of different modes is integrated, detail information of the different modes can be effectively reserved, and subsequent target detection by using the information of the different modes is facilitated. The principle part of a target detection head adopts the Centernet design concept in the prior art, a focus target is described as a central point, other characteristics such as the size of the target are obtained through direct regression in a feature map, and thus, the method is simple in principle and high in compatibility, does not need complex post-processing and achieves end-to-end detection.

Description

technical field [0001] The invention relates to the field of medical image processing, in particular to a nuclear medicine multimodal focus image detection method and system. Background technique [0002] Most of the current medical lesion detection methods are limited to one modality, such as CT, PET, MRI, SPECT, etc. alone. Restricted by its imaging principle, a single modality is often not enough for accurate processing of lesion images and follow-up personnel's judgment of lesions. Integrate different medical imaging equipment such as CT, PET, MRI, SPECT or different molecular probes, etc., integrate the advantages of various molecular imaging technologies, and develop a multi-modal fusion molecular imaging technology that can simultaneously provide the anatomical structure level of biological processes The combination of different modal information of information at the level of metabolism, function, physiology and pathology, and molecular and cellular levels makes use...

AI Technical Summary

Problems solved by technology

However, the imaging characteristics and data sizes of different devices and modalities are often inconsistent. Conventional methods usually rely on modal preprocessing such as scale change, registration, etc. to unify the modal data, and the operation method is too complicated.

Images