Multimodal Deep Learning Signature for Evaluation of Response to Bevacizumab in Patient With Colorectal Cancer Liver Metastasis

Multimodal Deep Learning Signature for Evaluation of Response to Bevacizumab in Patient With Colorectal Cancer Liver Metastasis

Establishment and validation of the deep learning signature of bevacizumab efficacy in initially unresectable CRLM patients

Initially unresectable CRLM patients were included in this study. The tumor response was assessed by local MDT group. The signature will classified patients into responder or non-responder group. We will administer mFOLFOX6+bevacizumab regimen to responders, and FOLFOXIRI regimen to non-responders.

Inclusion Criteria:

Age 18-75 years; Histologically proven colorectal adenocarcinoma; Simultaneous liver-limited metastases; Initially unresectable liver metastases determined by a local MDT; Life expectancy of > 3 months; RAS mutation and BRAF V600E wild-type; ECOG 0-1; Available PET/CT imaging before treatment; Available colonoscopy biopsy specimens before treatment

Exclusion Criteria:

Previous systemic treatment for metastatic disease; Previous surgery for metastatic disease; Extrahepatic metastases; Unresectable primary tumor; Major cardiovascular events (myocardial infarction, severe/unstable angina, congestive heart failure, CVA) within 12 months before randomisation; Acute or subacute intestinal obstruction; Second primary malignancy within the past 5 years; Drug or alcohol abuse; No legal capacity or limited legal capacity; Pregnant or lactating women; Uncontrolled hypertension, or unsatisfactory blood pressure control with ≥3 antihypertensive drugs; Peripheral neuropathy;