Intra-abdominal adipose tissue segmentation method based on deep learning

Abstract

The invention relates to an intra-abdominal adipose tissue segmentation method based on deep learning, and is used for segmenting visceral adipose tissue and subcutaneus adipose tissue at the same time. The method comprises the following steps: recognizing abdominal adipose tissue pixel points; performing deep learning on the recognized abdominal adipose tissue pixel points, so as to obtain substantive characteristics of the abdominal adipose tissue pixel points; inputting the substantive characteristics of the abdominal adipose tissue pixel points into a classification algorithm, so as to obtain a preliminary segmentation result; converting the preliminary segmentation result to a polar coordinate, and further amending abdominal adipose tissue pixel points which are misclassified, so as to obtain an abdominal adipose tissue segmentation map; computing the ratios of various adipose tissues in the abdominal adipose tissue segmentation map according to gradients, and visually displaying a computed result through a volume rendering technology. Compared with the prior art, the method has the advantages that the application is timely, retraining is not required, the error rate is low, the precision degree is high, and fully automatic segmentation is realized.

Description

technical field [0001] The invention relates to the field of medicine, in particular to a deep learning-based intra-abdominal fat segmentation method. Background technique [0002] Segmentation of intra-abdominal fat into different parts is of great value in the medical field. Compared with the more intuitive body mass index (BMI), the volume of visceral adipose tissue (VAT) and subcutaneous adipose tissue (SAT) can more accurately and effectively predict and analyze obesity-related diseases. [0003] Most of the existing intra-abdominal fat segmentation methods are mainly manual segmentation, that is, doctors or experienced personnel manually process the original image to mark the range of intra-abdominal fat and visceral fat. At present, semi-automatic segmentation software has appeared on the market, but its main work is still done manually. Although this kind of software can improve the efficiency of segmentation to a certain extent, it still takes 6-8 minutes for doct...

AI Technical Summary

Problems solved by technology

First of all, these algorithms need to be retrained for each new picture, which consumes a lot of time; secondly, these algorithms need to split the SAT and VAT separately, and need to process the picture twice, which increases the probability of additional errors

Images