Rib fracture detection model training system and method, rib fracture detection system and method

Abstract

The invention relates to a rib fracture detection model training system and method, a rib fracture detection system and method, and belongs to the technical field of artificial intelligence. The method comprises the following steps: acquiring a plurality of rib-containing CT images, and marking rib fracture areas on the CT images by experts; extracting a skeleton region in the CT image as a target region, and sampling the CT image in the target region to obtain a 3D local CT image; adopting a 3D DeepLab network architecture module, using a plurality of positive samples and a plurality of negative samples to train the model, constructing a novel rib fracture detection model based on high-resolution 3D ResNet-HR, using the obtained detection model to carry out rib fracture detection, and obtaining good expected performance. The obtained detection model is used for a rib fracture detection system, compared with manual observation, the speed is increased by more than 15 times, and the technical problem that an existing rib fracture detection technology is poor in sensitivity and specificity is well solved.

Description

technical field [0001] The invention relates to a technique for detecting fractures, in particular to a training system and a training method for a model based on deep learning to automatically detect fractures in rib CT images, a system and a detection method for using the model for detection, and belongs to artificial Smart technology field. Background technique [0002] Diagnosis of rib fractures is an important and complex clinical practice, both forensic and daily task scenarios of multiple businesses (such as insurance claims) need to be accurately diagnosed. However, there is little prior art investigating this labor-intensive task of automating machine learning techniques. [0003] While it is well known that rib fractures are associated with significant morbidity and mortality, thoracic trauma accounts for 10% to 15% of all traumatic injuries. Conventional chest CT is the main choice for thoracic trauma examination, which has the advantage of revealing occult frac...

AI Technical Summary

Problems solved by technology

[0007] The present invention proposes a rib fracture detection model training system, method, detection system and detection method, which are used to solve the existing method of classifying whether there is rib fracture detection based on the 2D semi-automatic rib unfolding tool and the projection of the two-dimensional CNN Technical problems with poor sensitivity and specificity in

Images