Algorithm-based intraoperative diagnosis of liver tumors using infrared spectroscopy

Area of Science:

• Biomedical Spectroscopy
• Medical Diagnostics
• Machine Learning in Medicine

Background:

• Intraoperative classification of liver tumors (hepatocellular carcinoma, cholangiocellular carcinoma, metastases) is challenging due to infiltrative growth.
• Current methods like frozen section analysis are time-consuming and subjective.
• There is a need for rapid, objective alternatives for real-time surgical guidance.

Purpose of the Study:

• To assess fiber-based attenuated total reflection infrared (ATR IR) spectroscopy combined with machine learning for intraoperative liver tumor classification.
• To evaluate the method's ability to differentiate normal liver tissue from various tumor subtypes based on biochemical signatures.

Main Methods:

• Analysis of fresh liver tissue from 69 surgical patients using a fiber-based Ge ATR IR spectroscopy probe.
• Application of supervised machine learning algorithms for classification.
• Validation using cross-validation and independent test sets.

Main Results:

• High accuracy in classifying normal tissue versus tumor subtypes (0.90 overall accuracy).
• Distinguished normal and tumor tissues with 0.89 sensitivity and 0.92 specificity, primarily based on glycogen content and tissue compactness.
• Accurately classified tumor subtypes (HCC, CCC, metastases) with an average accuracy of 0.90, identifying unique biochemical signatures for each.

Conclusions:

• Fiber-based ATR IR spectroscopy with machine learning offers a rapid, objective, and highly accurate tool for intraoperative liver tumor classification.
• This label-free biochemical approach can enhance surgical precision and potentially reduce recurrence risks.
• The method shows promise for application across various solid tumor types.