Liver tumor early recurrence prediction method based on 3D CNN and LSTM

Abstract

The invention provides a liver tumor early recurrence prediction method based on 3D CNN and LSTM. The method comprises the following steps: acquiring a CT image and clinical information of an HCC patient, and counting postoperative ER information of the HCC patient; carrying out image segmentation on the CT image, and then dividing obtained data samples into a training set and a test set; training by using the training set to obtain a 3D CNN model of a mapping relation between the CT image and the ER information; extracting iconography features of the data sample corresponding to the liver tumor area, performing dimension reduction on the extracted iconography features by using an LASSO logistic algorithm, and selecting out iconography features useful for ER prediction; carrying out statistical analysis on the clinical information, carrying out ER clinical factor univariate analysis by utilizing chi-square test, and selecting a clinical factor corresponding to a test level P less than 0.05; and training to obtain an optimal LSTM model by using features obtained by a 3D CNN model, imaging features useful for ER prediction and clinical factors corresponding to a test level P less than 0.05, and predicting the ER of an HCC patient by using the LSTM model.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a method for predicting early recurrence of liver tumors based on 3DCNN and LSTM. Background technique [0002] Hepatocellular carcinoma (HCC) is a common malignant tumor that endangers human physical and mental health. In theory, patients with early HCC are ideal candidates for liver transplantation, partial hepatectomy and radiofrequency ablation. Even for patients with advanced liver cancer, surgical resection is beneficial to prolong survival. The time interval from surgical resection to recurrence is an independent prognostic factor affecting survival and an important cause of death. Many studies have found that the peak recurrence of HCC after radical resection is within 1 year, so early recurrence (Early Recurrence, ER) can be defined Within 1 year, the mortality rate after HCC recurrence is very high. If ER can be predicted before surgery, the risk index of HCC ...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the problem that the existing ER prediction method for HCC patients depends heavily on the segmentation quality of tumor images and affects the prediction results, the present invention provides a method for early recurrence prediction of liver tumors based on 3DCNN and LSTM, which can reduce the image segmentation quality. Dependence, and can automatically extract features from images, improving the accuracy of liver tumor ER prediction

Images