The digital pathology firm PathAI has released PathExplore Fibrosis, an AI-based tool that analyzes fibrosis and collagen structures from H&E-stained whole-slide tissue images. The software quantifies fibrotic areas and collagen fibers from standard pathology slides, replacing specialized staining techniques and microscopy equipment. The tool processes large datasets of tissue images, designed to work with existing laboratory imaging systems. (The company notes that PathExplore Fibrosis is for research use only. It is not for use in diagnostic procedures.)
In the tumor microenvironment (TME), fibrosis and collagen affect how tumors respond to treatment. These structures influence tumor stiffness, immune cell infiltration, and the architectural organization of tumors. Those are all factors that can determine treatment success and cancer spread. Researchers use this data to evaluate drug efficacy, identify biomarkers for cancer prognosis, and develop new treatments targeting fibrotic pathways. The platform provides detailed analysis of the morphology and spatial distribution of these structures, helping researchers understand how the TME contributes to cancer progression and treatment resistance.
To learn more about the news, we caught up with Ben Glass, VP of Product and Translational Research at PathAI, who discusses how the technology moved from an internal research project to a widely available research tool. In our interview, Glass details the platform’s machine learning capabilities and explains its applications in biomarker discovery and drug development.
[PathAI]
Ben Glass
PathExplore Fibrosis has several impactful applications in oncology, specifically in the area of biomarker discovery. One major area is biomarkers predictive of treatment response—the fibrosis and stromal elements of the tumor microenvironment can influence how well a patient responds to certain therapies, and our tool can help identify biomarkers that correlate with patient benefit from specific treatments. It can also be useful in the discovery of prognostic biomarkers; the extent, quality, and spatial distribution of fibrosis may provide valuable insights for patient stratification and treatment planning. In drug development, PathExplore Fibrosis allows researchers to evaluate the impact of new therapies on fibrosis, potentially accelerating the development of more effective treatments.
Finally, PathExplore Fibrosis will enable access to collagen and fibrosis research to many more scientists than previously possible. Current methods for fibrosis measurement require complex staining techniques or expensive imaging setups. Here, we are able to measure fibrosis present in whole-slide images of H&E-stained tissue, which are widely available as part of standard lab workflows. Even retrospective analysis of archival image data is now possible.
Can you provide some examples of the new insights the platform will offer for the study of tumor biology and therapeutic response?
PathExplore fibrosis has the potential of unlocking novel insights and strategies from translational research all the way to clinical development.
To name a few:
- Predicting Treatment Response: Fibrosis can influence the efficacy of various cancer therapies. By characterizing the quantity, nature, and spatial orientation of fibrosis in the tumor microenvironment, PathExplore Fibrosis can discover novel biomarkers to identify patients who are more or less likely to respond to specific therapies. An example of this will be presented in poster form at the upcoming Society for Immunotherapy of Cancer (SITC) Annual Meeting, November 6-10, 2024.
- Prognostic Biomarker: The extent and spatial distribution of fibrosis can provide valuable prognostic information, potentially aiding in patient stratification and treatment planning.
- Drug Development: As combination therapies targeting different aspects of tumor pathogenesis are emerging, PathExplore Fibrosis can be used to interrogate the association of fibrosis, TME composition, and response to the treatment to accelerate the identification of effective strategies for drug development and combination therapies.
In addition, can you provide more information on the ML techniques involved? Are they bespoke methods?
The accuracy of PathExplore Fibrosis stems from its unique ground truth data. We trained the model on an extensive dataset of images using our own quantitative multi-modal anisotropic imaging (QMAI) technology that we built in house. This allowed us to capture high-quality data from H&E-stained slides, focusing on collagen and fibrosis regions with incredible precision. The result is a product that can reliably identify and quantify fibrosis with a level of detail that hasn’t been possible until now.
Can you share more on the backstory? It began as an internal research project?
PathExplore Fibrosis started with our work on quantitative multi-modal anisotropic imaging (QMAI). We developed this imaging method to quantify collagen fibers directly from H&E-stained tissue samples, and we used these as the basis for training our machine-learning model. The idea was to create a tool that could accurately assess fibrosis in H&E whole-slide images. After extensive validation and refinement, we transformed this technology into an off-the-shelf product that researchers and clinicians can use right out of the box. It’s been an exciting journey from concept to a fully realized product.
Filed Under: clinical trials, machine learning and AI
