MSACL 2025 Abstract

MassVision: Enhancing Clinical Translation of Mass Spectrometry Imaging Through a Comprehensive Analytical Platform

Amoon Jamzad, Martin Kaufmann, Jade Warren, Natasha Iaboni, Christopher J. B. Nicol, John F. Rudan, Kevin Y. M. Ren, David Hurlbut, Sonal Varma, Gabor Fichtinger, Parvin Mousavi
Queen’s University, Kingston, ON, Canada

Amoon Jamzad, PhD (Presenter) Queen's University

Presenter Bio: Dr. Amoon Jamzad is an AI research scientist and engineer focused on machine learning for biomedical and clinical data. His work centers on translating complex data, particularly mass spectrometry imaging, into clinically useful insights through computational methods and practical software tools. He is currently a Postdoctoral Fellow and Adjunct Assistant Professor at School of Computing, Queen’s University. He is the developer of MassVision, an open-source platform for MSI analysis, and his work has appeared in venues including Analytical Chemistry, MICCAI, and IJCARS.

Relevant Financial Disclosures (within past 24 months, reported on Apr 22, 2026) No relevant financial relationship(s) to disclose.

INTRODUCTION:
Mass spectrometry imaging (MSI) has emerged as a powerful tool that combines molecular specificity with spatial resolution, enabling detailed characterization of biological tissue samples. Its clinical applications are particularly transformative in oncology for distinguishing tissue types and identifying molecular biomarkers associated with disease, thus improving diagnostic accuracy. Despite these advantages, the use of MSI in clinical settings is challenging due to the complexity and high-dimensional nature of its data. Existing tools often do not provide a complete solution – from data visualization to training of tissue-type classification models – for exploring and analyzing MSI data. As a result, researchers have to use a combination of multiple analytical tools or develop their own highly specialized methods, which requires coding and software development expertise. This makes the analysis workflow more complex, reducing efficiency and making standardization difficult, especially for high-throughput clinical applications.

OBJECTIVE:
We introduce MassVision, an all-in-one, user-friendly, and free software solution for exploration and AI-driven analysis of MSI data.

METHODS:
MassVision was developed as part of the widely-used open-source platform called 3D Slicer. This is to ensure that the software is reliable, easy to use, works on different operating systems, and can handle large scale datasets. MassVision offers several important functionalities including data visualizations (spectral vs. spatial, targeted vs. untargeted), pattern identification (pixel clustering in local and global contrasts), dataset generation (multi-modal co-localization, histopathology-guided annotation), data preprocessing (multi-dataset merge, spectral and spatial filtering), and AI model training and validation (several supervised classification methods, data stratification and balancing). MassVision also allows researchers to directly apply their trained tissue classification models to new whole-slide MSI data, making it suitable for clinical use.

MassVision was validated using both an in-house MSI dataset (DESI of colorectal cancer tissues), and several publicly available datasets across diverse modalities (MALDI and DESI) and tissues (prostate, colorectal, brain xenograft). The implemented functionalities of MassVision were evaluated in different use cases including untargeted slide exploration, identification of biologically informative features, and AI-based tissue-type classification. The performance of MassVision in these tasks was assessed both qualitatively (visual comparisons to gold-standard histopathology), as well as quantitatively (confusion matrices and accuracy metrics).

RESULTS:
MassVision significantly improved the preparation of datasets with pathology-validated labels. It achieved more than fivefold improvement in speed and data yield on a sample slide compared to the conventional approach. Furthermore, the novel visualization methods implemented in MassVision effectively highlighted biologically significant patterns specific to different pathology labels, or heterogeneity within a single pathology label. Additionally, MassVision demonstrated successful training, validation, and deployment of tissue-type classification on colorectal data. Customizing the analysis by balancing the training datasets has been shown to significantly enhance the performance, qualitatively and quantitatively. Lastly, MassVision proved to be reliable when used across different modalities and organs. It successfully identified tissue-specific features indifferent public datasets that had previously been reported. These results highlighting the versatility and clinical utility of MassVision.

CONCLUSIONS:
MassVision addresses important challenges that previously limited the practical use of MSI in clinical settings. It provides an all-in-one solution, simplifies complex data analysis, handles large datasets efficiently, and speeds up the workflow. Designed for users without advanced programming skills, MassVision integrates visualization, image annotation, and AI-based tissue classification into a single, easy-to-use platform. As a result, MassVision facilitates the transition of MSI from basic research into clinical diagnostics. This promotes precise, label-free tissue analysis, which can ultimately improves patient outcomes.

REFERENCES:
1. Jamzad, A., et al., MassVision: An open-source end-to-end platform for AI-driven mass spectrometry image analysis. bioRxiv, 2025; https://doi.org/10.1101/2025.01.29.635489.
2. Jamzad, A., MassVision User Guide, 2025; https://SlicerMassVision.readthedocs.io/.