590P - A deep-learning model to predict the completeness of cytoreductive surgery in colorectal cancer with peritoneal metastasis

Background

Colorectal cancer (CRC) CRC with peritoneal metastasis (PM) obtain poor prognosis. Peritoneal Cancer Index (PCI) is used to evaluate the PM extent and selection of Cytoreductive surgery (CRS) combined with hyperthermic intraperitoneal chemotherapy (HIPEC). However, PCI score assessed before surgery is not precise.We have developed a novel AI framework of DeAF (Decoupling feature alignment and fusion) to aid the selection and predict completeness of CRS in PM.

Methods

185 CRC patients with PM recruited from four tertiary hospitals were enrolled. The internal patient cohort was subsequently stratified into training and validation cohorts, comprising 84 and 30 individuals, respectively. Deep learning was used to train the DeAF model of Simsam algorithms by contrast CT images and then cooperate with clinicopathological parameters to increase the performance. The performance was evaluated by accuracy, sensitivity, specificity, and AUC by ROC in the internal and external validation cohorts.

Results

The AI model demonstrated a robust ability to predict the completeness of CRS with AUC of 0.9 in internal validation cohort. The model can predict whether patients are suitable for CRS. It can differentiate patients who would benefit from CRS. The model also showed high predictive performance of CRS completeness in external validation cohorts with AUC values of 0.906, 0.960, and 0.933, respectively (rounded to three decimal places).

Conclusions

The novel DeAF framework can aid surgeons to select proper patients for CRS and predict the completeness of CRS. The model can change surgical decision-making and provide potential benefits for patients with PM.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

The National Natural Science Foundation of China.

Disclosure

All authors have declared no conflicts of interest.