European Radiology: The value of deep convolutional neural networks in predicting preoperative microvascular infiltration and clinical prognosis of HCC patients

Hepatocellular liver carcinoma (HCC) is curative resection surgery is the first choice early HCC patients
.

However, the recurrence rate after 2 years is very high, reaching 50%

Hepatocellular liver carcinoma (HCC) is curative resection surgery is the first choice early HCC patients

Imaging can help predict the MVI of HCC

Recently, published in the E uropean a journal Radiology study this explores DCNN use of contrast in a large number of candidates for HCC surgery enhanced CT scan (CECT) identify MVI and predict the clinical outcome of the performance , for accurate evaluation before surgery and clinical quality programs The formulation of this provides a reference basis for imaging studies .
 

This retrospective study included 1116 HCC patients who underwent preoperative CECT and radical hepatectomy

In the training and validation cohorts, the proportion of MVI-positive patients was 38.

FIG a tris Phase DCNN probability constructed DCNN nomogram .
b DCNN collinear FIG ROC analysis MVI prediction curve in the training and validation cohort .
DCNN, deep convolutional neural network; MVI, microvessel invasion .
 

FIG a tris Phase DCNN probability constructed DCNN nomogram .

b DCNN collinear FIG ROC analysis MVI prediction curve in the training and validation cohort .
DCNN, deep convolutional neural network; MVI, microvessel invasion .
FIG a tris Phase DCNN probability constructed DCNN nomogram .
b DCNN collinear FIG ROC analysis MVI prediction curve in the training and validation cohort .
DCNN, deep convolutional neural network; MVI, microvessel invasion .
 

In this study, a joint nomogram was established by combining clinical factors and radiological characteristics based on preoperative CECT , which can be used to predict MVI before surgery
.

At the same time , the combined nomogram can also predict DFS and OS , providing a non-invasive imaging reference basis for the formulation of clinical personalized treatment plans

In this study, a joint nomogram was established by combining clinical factors and radiological characteristics based on preoperative CECT , which can be used to predict MVI before surgery

Original source :

Xinming Li , Zhendong Qi , Haiyan Du ,et al.

Deep convolutional neural network for preoperative prediction of microvascular invasion and clinical outcomes in patients with HCCs.
DOI: 10.
1007/s00330-021-08198-w

Of Li Xinming Zhendong Qi Haiyan Du 10.
1007 / W-s00330-021-08198 10.
1007 / s00330-021-08198-W in this message