Tissue classification of oncologic esophageal resectates based on hyperspectral data

Area of Science:

• Oncology
• Medical Imaging
• Computational Pathology

Background:

• Esophageal carcinoma is a significant global health concern, necessitating accurate post-operative pathological examination.
• Esophageal resection with gastric pull-up is a key curative treatment, requiring precise margin assessment.
• Intraoperative analysis of resected tissue could expedite diagnosis and enhance patient safety.

Purpose of the Study:

• To evaluate the efficacy of hyperspectral imaging (HSI) for detecting esophageal carcinoma in resected tissues.
• To assess the performance of supervised classification algorithms in differentiating malignant from healthy tissue.
• To explore the potential for automated intraoperative margin assessment.

Main Methods:

• Hyperspectral imaging (HSI) was used to record data from esophago-gastric resectates of 11 patients.
• Four supervised classification algorithms were evaluated: random forest, support vector machines (SVM), multilayer perceptron, and k-nearest neighbors.
• Algorithms were trained to differentiate malignant from healthy tissue based on HSI data.

Main Results:

• Support vector machines (SVM) achieved the highest performance with 63% sensitivity and 69% specificity for cancerous tissue detection.
• Cross-validation revealed patient-specific performance variations.
• Automated data classification and visualization were achieved in under 1 second.

Conclusions:

• Several classification algorithms show potential for automated esophageal carcinoma detection in resected tissue.
• Increasing dataset size and combining methods may improve diagnostic performance.
• Software implementation for intraoperative tumor boundary visualization could aid surgeons in oncologic procedures.