A fully automatic pelvic tumor segmentation method and system, storage medium and terminal

Abstract

The present invention provides a fully automatic pelvic tumor segmentation method and system, a storage medium and a terminal, comprising the following steps: performing preprocessing and data amplification on images in a meningioma image set to obtain meningioma training data; based on the meningioma training Data training U‑net full convolutional neural network; preprocessing and data amplification of pelvic tumor images drawn with tumor boundary gold standard in the pelvic tumor image set to obtain pelvic tumor training data; training data based on the pelvic tumor training data The obtained U‑net full convolutional neural network is fine-tuned to obtain the optimal U‑net full convolutional neural network; the collected pelvic tumor images are input into the optimal U‑net full convolutional neural network to obtain pelvic tumor segmentation result. The fully automatic pelvic tumor segmentation method and system, storage medium and terminal of the present invention realize fully automatic pelvic tumor segmentation through the U-net fully convolutional neural network of transfer learning, thereby greatly improving the efficiency of medical-industrial interaction and the accuracy of pelvic tumor segmentation Spend.

Description

technical field [0001] The invention relates to the technical field of medical image processing, in particular to a fully automatic pelvic tumor segmentation method and system, a storage medium and a terminal. Background technique [0002] Tumors of the musculoskeletal system, regardless of primary or secondary bone tumors, often exhibit marked heterogeneity. Especially for complex pelvic tumors, the lesions involve multiple pelvic regions and the surgical resection range is large, which makes the automatic segmentation of pelvic tumors always challenging. [0003] At present, tumor boundary image determination is the link with the lowest efficiency and the most impact on surgical results in medical-engineering interaction engineering. Existing pelvic tumor segmentation methods are usually manually labeled and segmented by experienced physicians, which have the following shortcomings: [0004] (1) It is time-consuming and labor-intensive, and it is difficult to obtain a la...

AI Technical Summary

Problems solved by technology

[0004] (1) Time-consuming and labor-intensive, it is difficult to quickly obtain a large number of segmentation results;

[0005] (2) The segmentation results are affected by the physician's subjectivity and experience, and the repeatability is low

Images