Innovative label-free lymphoma diagnosis using infrared spectroscopy and machine learning on tissue sections.

Area of Science:

• Biomedical Spectroscopy
• Computational Pathology
• Machine Learning in Diagnostics

Background:

• Lymphoma diagnosis presents challenges due to histological and clinical diversity.
• Current diagnostic methods often require specialized facilities and expertise.
• There is a need for accessible, cost-effective diagnostic tools for routine laboratories.

Purpose of the Study:

• To evaluate the efficacy of Attenuated Total Reflection-Fourier Transform Infrared (ATR-FTIR) spectroscopy coupled with machine learning for lymphoma detection and subtyping.
• To determine if ATR-FTIR can differentiate between lymphoma and non-malignant lymphoid tissues.
• To assess the potential of this technique as a non-destructive, user-friendly diagnostic tool.

Main Methods:

• Analysis of lymphoid tissue sections from 295 lymphoma patients and 389 controls using ATR-FTIR spectroscopy.
• Spectral data was divided into 70:30 train-test sets.
• Partial Least Squares Discriminant Analysis (PLS-DA) models were developed to classify samples and subtypes.

Main Results:

• Significant spectral differences between lymphoma and non-malignant tissues were observed in the 1800-900 cm⁻¹ region, linked to biochemical components.
• The PLS-DA model achieved an AUC of 0.882 on the independent test set for differentiating lymphoma from non-malignant tissue.
• Distinct spectral patterns and clustering were identified among different lymphoma subtypes.

Conclusions:

• ATR-FTIR spectroscopy combined with machine learning is a promising tool for lymphoma diagnosis.
• The technique offers a non-destructive, rapid, and inexpensive approach.
• This method has the potential for implementation in non-specialized laboratories, improving accessibility to lymphoma diagnostics.