Artificial intelligence-based model to predict recurrence after local excision in T1 rectal cancer

Area of Science:

• Oncology
• Artificial Intelligence
• Medical Imaging

Background:

• Current guidelines recommend additional surgery for T1 colorectal cancer (CRC) with high-risk features after local excision (LE), despite low recurrence rates.
• Surgery for low rectal cancer (RC) can impair anal function and reduce quality of life.
• There is a need to reduce unnecessary surgeries in these patients.

Purpose of the Study:

• To develop and validate an artificial intelligence (AI) model for predicting recurrence risk after LE in T1 RC.
• To reduce unnecessary surgical interventions in patients with T1 RC.

Main Methods:

• An artificial neural network (ANN) was developed using pathological images from 496 T1 RC specimens from Guangdong Provincial People's Hospital (GDPH).
• The ANN model was validated using an independent external dataset of 150 images from Sun Yat-sen Memorial Hospital (SYSMH).

Main Results:

• The ANN model demonstrated high predictive accuracy, with an area under the receiver operating characteristic curve (AUC) of 0.979 in the training cohort and 0.978 in the validation cohort.
• The AI-based model identified over 34.9% of patients who could avoid unnecessary additional surgeries compared to current guidelines.

Conclusions:

• A novel ANN model is proposed for predicting recurrence risk in T1 RC patients post-LE.
• This AI tool can guide clinical decisions, potentially reducing unnecessary invasive surgeries and improving patient outcomes.