Feedforward Artificial Neural Network-Based Colorectal Cancer Detection Using Hyperspectral Imaging: A Step towards...

Area of Science:

• Medical Imaging
• Oncology
• Artificial Intelligence

Background:

• Colorectal cancer (CRC) diagnosis relies on visual assessment, AI, surgery, and histopathology.
• Hyperspectral imaging (HSI) is a promising non-invasive optical technology for medical applications.

Purpose of the Study:

• To evaluate the efficacy of combining HSI with AI algorithms for discriminating CRC.
• To assess HSI's capability in detecting biological changes related to tumor staging and therapy.

Main Methods:

• 54 CRC patients underwent HSI of the tumor from the mucosal side.
• A four-layer perceptron neural network classified images into cancer (CA), adenomatous margin (AD), and healthy mucosa (HM).
• Leave-one-patient-out cross-validation was employed for performance assessment.

Main Results:

• The neural network achieved 86% sensitivity and 95% specificity in classifying CA or AD.
• Significant differences in perfusion parameters, such as oxygen saturation, correlated with tumor staging and neoadjuvant therapy.
• HSI detected biological changes in tissue induced by chemotherapy.

Conclusions:

• HSI combined with automatic classification effectively differentiates CRC from healthy mucosa.
• HSI is a valuable tool for detecting treatment-induced biological changes in colorectal tissues.