Quantitative Mass Spectrometry Imaging Using Multivariate Curve Resolution and Deep Learning: A Case Study.

Area of Science:

• Analytical Chemistry
• Biomedical Imaging
• Computational Biology

Background:

• Mass spectrometry imaging (MSI) is crucial for identifying compounds and mapping their distribution in tissues.
• Quantitative analysis in MSI often requires extensive sample preparation and can be challenging.
• Chlordecone, a carcinogenic pesticide, poses a health risk, necessitating accurate detection methods in biological samples.

Purpose of the Study:

• To develop a novel, sample-preparation-free quantitative approach for MSI using deep learning (DL) and multivariate curve resolution (MCR).
• To quantitatively determine chlordecone levels in mouse liver tissues.
• To validate the proposed method against established techniques like MALDI-MSI and GC-MS.

Main Methods:

• A convolutional neural network (CNN) was trained using pixels as samples, with chlordecone amounts determined by multivariate curve resolution-alternating least-squares (MCR-ALS).
• The CNN model addressed data scarcity by treating each pixel as an individual sample.
• The method was tested on mouse liver tissues exposed to chlordecone for varying durations (1, 5, 10 days).

Main Results:

• The developed DL-MCR approach achieved high prediction accuracy (R² = 0.93–0.96) for chlordecone quantification.
• Performance surpassed that of Support Vector Machine (SVM) and Partial Least Squares (PLS) models.
• Quantitative results for chlordecone in mouse liver tissues were comparable to traditional MALDI-MSI and GC-MS methods.

Conclusions:

• The proposed integrated MCR-DL method offers a robust and accurate solution for quantitative MSI.
• This technique eliminates the need for sample preparation, streamlining the analysis process.
• The method demonstrates significant potential for analyzing carcinogenic compounds in biological tissues.