The MALMO project combines artificial intelligence, computational pathology, 3D whole-slide image reconstruction and mechanistic modelling to better understand melanoma progression, vascular organization and treatment resistance.
Research Themes
Project Overview
The MALMO project investigates how artificial intelligence and mechanistic modelling can be combined to improve our understanding of melanoma progression and treatment resistance. A key biological motivation is that resistance may emerge from metabolic rewiring: cancer cells adapt to nutrient- and oxygen-depleted tumor microenvironments, enabling survival under hostile conditions.
In this context, MALMO focuses on whole-slide images, digitized pathology slides and vascular structures as key indicators of oxygen and nutrient supply. By extracting and reconstructing vascular networks from histological data, the project aims to provide quantitative features that can support predictive mechanistic models of oxygen diffusion and tumor microenvironment heterogeneity.
Computational Pipeline
The project develops a 2D and 3D image-analysis pipeline for the extraction, registration and reconstruction of vascular structures from melanoma whole-slide images. Blood vessels are isolated as a biologically meaningful biomarker, since they form the network through which oxygen and nutrients are delivered to tissue.
The pipeline has two complementary goals. First, it contributes to the automation and standardization of pathology workflows, which are often time-consuming and expert-dependent. Second, it extracts quantitative vascular features that can be integrated into mechanistic models of melanoma tissue organization, oxygen diffusion and metabolic adaptation.
Scientific Impact
MALMO is positioned at the interface between computational pathology, cancer biology, mathematical modelling and artificial intelligence. It illustrates a broader research direction: using AI not only for image-based prediction, but also as a bridge between histological structure, biological mechanisms and clinically relevant hypotheses.
By combining image-derived vascular features with mechanistic modelling, the project aims to contribute to a more interpretable and biologically grounded understanding of melanoma heterogeneity and treatment resistance.
Resources
WSI Registration
Optimized whole-slide image registration code for 3D reconstruction.
Open Repository3D Reconstruction Demonstration
The video below illustrates the 3D reconstruction of vascular structures from whole-slide histopathology images, supporting downstream modelling of tissue architecture and oxygen diffusion in melanoma.
