A Convolutional Neural Network based system for Colorectal cancer segmentation on MRI images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Accurate segmentation of colorectal cancer is crucial for effective treatment planning and personalized medicine.
• Current segmentation methods can be time-consuming and subjective, necessitating automated solutions.

Purpose of the Study:

• To develop and evaluate a novel Convolutional Neural Network (CNN) based system for the automatic segmentation of colorectal cancer.
• To assess the performance of the proposed system using established metrics and compare it against a radiologist-revised standard.

Main Methods:

• The system employs a multi-step approach including pre-processing for normalization and tumoral area highlighting.
• Classification is performed using an ensemble of three CNNs analyzing different MR sequences, with results combined via majority voting.
• Post-processing steps are included to reduce false positive segmentations.

Main Results:

• The system achieved a Dice Similarity Coefficient (DSC) of 0.60, Precision (Pr) of 0.76, and Recall (Re) of 0.55 on the testing set.
• Leave-one-out validation yielded a median DSC of 0.58, Pr of 0.74, and Re of 0.54.
• These results indicate a robust performance in segmenting colorectal cancer regions.

Conclusions:

• The developed CNN-based system demonstrates promising capabilities for automatic colorectal cancer segmentation.
• Further validation on larger datasets is warranted to fully establish its clinical utility.
• The system has the potential to significantly enhance personalized medicine approaches in oncology.