Improving Survival Predictions in Colorectal Cancer with ResNet
Author Information
Author(s): Chiu Sung-Hua, Li Hsiao-Chi, Chang Wei-Chou, Wu Chao-Cheng, Lin Hsuan-Hwai, Lo Cheng-Hsiang, Chang Ping-Ying
Primary Institution: Tri-Service General Hospital, National Defense Medical Center, Taipei, Taiwan
Hypothesis
Can the residual convolutional neural network (ResNet) improve survival predictions in colorectal cancer patients with unresectable liver metastases treated with bevacizumab-based chemotherapy?
Conclusion
The use of ResNet can enhance survival predictions for colorectal cancer patients with liver metastases undergoing treatment.
Supporting Evidence
- ResNet-MR showed higher discriminatory ability for mortality than RD-MR.
- Patients classified as responders by ResNet had longer overall survival than non-responders.
- ResNet identified additional responders who were initially classified as non-responders.
Takeaway
This study shows that a computer program can help doctors better predict how long patients with liver cancer will live after treatment.
Methodology
A retrospective review of liver-only mCRC patients treated with BBC was performed, evaluating CT images and survival outcomes.
Potential Biases
Subjective bias may arise from radiologists' classifications of CT images.
Limitations
The study is retrospective, single-center, and may have subjective bias in image classification.
Participant Demographics
84 patients (53 males, 31 females) with a median age of 60 years.
Statistical Information
P-Value
0.0167 and 0.0225
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website