Evaluation of the Risk of Recurrence in Patients with Local Advanced Rectal Tumours by Different Radiomic Analysis...

Area of Science:

• Radiomics and artificial intelligence in oncology.
• Machine learning applications in medical imaging.

Background:

• Radiomics generates large datasets from medical images, requiring advanced analytical techniques.
• Artificial intelligence, particularly machine learning and deep learning, is crucial for exploiting radiomic data.
• Previous radiomics studies in colorectal cancer focused on treatment response, metastasis, and survival, with limited exploration of recurrence-free survival.

Purpose of the Study:

• To evaluate and compare conventional machine learning and deep learning models for predicting recurrence-free survival in locally advanced rectal tumors.
• To assess the impact of 2D vs. 3D image analysis and tumor vs. tumor-plus-peritumoral environment analysis in conventional models.
• To develop and test a deep learning model using MRI textural analysis for recidivism risk.

Main Methods:

• Compared six conventional machine learning models (2D vs. 3D, tumor vs. tumor-plus-peritumoral analysis) using MRI textural analysis.
• Developed a 16-layer convolutional neural network (deep learning model).
• Utilized a 2D MRI image database including native images and bounding boxes for the deep learning model.

Main Results:

• The study evaluated the performance of various radiomic models in predicting recurrence-free survival.
• Deep learning models showed potential effectiveness in analyzing MRI textural data for recidivism risk.
• Comparison between conventional and deep learning approaches was central to the study's findings.

Conclusions:

• Radiomics, powered by artificial intelligence, shows promise for predicting recurrence-free survival in rectal cancer patients.
• Deep learning models offer a powerful approach for analyzing complex MRI textural data.
• Further research can refine these models for improved clinical decision-making in rectal cancer management.