Multimodal Lung Cancer Subtyping Using Deep Learning Neural Networks on Whole Slide Tissue Images and MALDI MSI.

Area of Science:

• Computational pathology
• Artificial intelligence in oncology
• Mass spectrometry imaging

Background:

• Accurate subtyping of non-small cell lung cancer (NSCLC) is critical for treatment planning.
• Distinguishing adenocarcinoma (ADC) from squamous cell carcinoma (SqCC) can be challenging in routine clinical practice.
• AI offers potential for enhancing tumor detection and classification.

Purpose of the Study:

• To develop and validate a two-modality AI algorithm for detecting and classifying NSCLC tumor subtypes.
• To integrate data from matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI MSI) and whole slide images (WSIs).
• To assess the algorithm's performance in a clinical context.

Main Methods:

• A two-modality AI algorithm was developed, combining MALDI MSI data with H&E-stained WSIs.
• Tumor areas were detected via WSI segmentation, followed by classification using MALDI MSI data.
• The algorithm was trained on tissue microarrays (TMAs) from 232 patients and validated on whole tissue sections.

Main Results:

• The AI algorithm achieved 94.7% accuracy at the spectrum level for tumor classification.
• It correctly classified 15 out of 16 test sections.
• An additional quality control step led to 100% accuracy on 14 of 16 tested sections.

Conclusions:

• The AI-based method demonstrates high accuracy in detecting and classifying NSCLC subtypes.
• Integrating AI with MALDI MSI data represents a significant step towards clinical application.
• The approach has the potential to aid pathologists and improve diagnostic efficiency.