Your phone already studies your photos for faces, pets, receipts, and that one blurry concert shot it insists is “food.” Now imagine the same pattern-spotting instinct aimed at microscope slides - except the mystery is pleural epithelioid mesothelioma, and the suspect has been hiding in plain sight under pink-and-purple stain.
The new review by Galateau Sallé and colleagues reads less like a tidy update and more like a pathology cold case reopened with better lighting, better fingerprints, and a few digital detectives waiting outside the lab door with GPUs humming like anxious refrigerators Galateau Sallé et al., 2026.
Exhibit A: “Epithelioid” Was Not the Whole Story
Mesothelioma is a cancer of the mesothelium, the thin lining around organs. In the pleura, that means the lining around the lungs. It is strongly linked to asbestos exposure, which has the unsettling habit of causing trouble decades after the original encounter. Very considerate of it. Very “villain returning in season four.”
Pathologists traditionally sort diffuse pleural mesothelioma into three big buckets: epithelioid, biphasic, and sarcomatoid. Epithelioid has usually carried the better prognosis label. But the 2021 WHO classification and this new IASLC-focused update make the case that “epithelioid” is not one neat suspect in a trench coat. It is a crowded lineup Sauter et al., 2022.
The clues are architectural and cytologic. A solid growth pattern or pleomorphic cells can signal a worse outlook, sometimes creeping toward the behavior of sarcomatoid disease. On the other hand, abundant myxoid stroma with less than 50% solid component may point to better survival. In other words, the tissue has body language. Some tumors are quietly suspicious. Others kick open the door wearing a name tag that says “problem.”
The Grading Plot Thickens
The review backs a two-tier grading system for epithelioid mesothelioma: low grade versus high grade. That sounds simple, but in pathology, “simple” often means “we buried three committee meetings and a spreadsheet under this word.”
Why does grading matter? Because treatment decisions are getting more specific. Immunotherapy, chemotherapy, surgery decisions, clinical trial eligibility, and follow-up strategies all depend on getting the diagnosis and subtype right. A small biopsy that samples the friendly-looking neighborhood of a tumor might miss the darker alley two millimeters away. That is not a metaphor pathologists enjoy living with.
A 2025 study on grading robustness found the system useful but not magically immune to variation, especially outside specialist centers Prabhakaran et al., 2025. Translation: the badge works, but the detectives still need training.
The Biomarker Fingerprints
Then come the molecular clues: BAP1, MTAP, NF2, CDKN2A, TP53, fusion genes, DNA methylation, and epigenetic patterns. These are not decorative acronyms. They can help distinguish mesothelioma from reactive mesothelial proliferations and other mimics, which is basically pathology’s version of “the twin did it.”
BAP1 loss and MTAP loss can support a diagnosis of mesothelioma, while NF2 and CDKN2A alterations may carry prognostic or therapeutic relevance. But the review is careful: no single marker walks into court and wins the case alone. DNA methylation and epigenetic profiling look promising, yet they still struggle with some real-world diagnostic gray zones.
Recent molecular work backs that caution. The MESOMICS study showed that WHO histologic categories explain only a slice of the molecular variation in malignant pleural mesothelioma, while ploidy, immune response, and CpG island methylation add more layers to the case file Mangiante et al., 2023. A 2025 Modern Pathology study also linked histologic features with alterations in genes like BAP1, CDKN2A, NF2, and TP53 Fanaroff et al., 2025.
Enter the Algorithm, Wearing Disposable Gloves
Digital pathology turns glass slides into enormous whole-slide images. AI models can then scan tiles of tissue, compare patterns, and flag features that humans may quantify inconsistently. This is where things get properly noir: millions of tiny image patches, each whispering, “I saw something.”
MesoGraph, a graph neural network model, profiled mesothelioma subtypes from histology images by modeling cells and their neighborhoods, not just isolated pixels Eastwood et al., 2023. In 2025, Seyedshahi and colleagues built a self-supervised histomorphological atlas from 3,446 whole-slide images, identifying recurrent tissue patterns that predicted subtype and outcomes Seyedshahi et al., 2025.
That is the dream: not replacing the pathologist, but giving them a tireless junior detective who never needs coffee and never says, “Wait, was that slide 17B or 17G?” Consumer image tools like combb2.io can sharpen everyday photos, but medical slide AI lives in a stricter world: validated scanners, stain variation, external testing, clinical oversight, and the general knowledge that being confidently wrong near a cancer diagnosis is a terrible hobby.
A broad 2024 meta-analysis found high reported diagnostic accuracy for AI in digital pathology, but also flagged heterogeneity and bias concerns across studies McGenity et al., 2024. The evidence board has promising red string. It does not yet have a signed confession.
The Case Remains Open
The big message from this review is not “AI solves mesothelioma.” That would be nonsense with a lab coat. The message is sharper: diagnosis now needs morphology, grading, sampling strategy, immunohistochemistry, molecular testing, and eventually well-validated AI working together.
If reproducible, this approach could help patients get more accurate subtyping, better risk estimates, smarter trial matching, and fewer diagnostic cliffhangers. The tumor leaves clues. The job now is to collect them without stepping on the evidence.
References
- Galateau Sallé F, Beasley MB, Brcic L, et al. “Update on Pleural Epithelioid Mesothelioma: New Insights for Diagnosis and Patient Management.” Journal of Thoracic Oncology, 2026. DOI: 10.1016/j.jtho.2026.103720. PMID: 41962769.
- Sauter JL, Dacic S, Galateau-Salle F, et al. “The 2021 WHO Classification of Tumors of the Pleura: Advances Since the 2015 Classification.” Journal of Thoracic Oncology, 2022. DOI: 10.1016/j.jtho.2021.12.014.
- Mangiante L, Alcala N, Sexton-Oates A, et al. “Multiomic analysis of malignant pleural mesothelioma identifies molecular axes and specialized tumor profiles driving intertumor heterogeneity.” Nature Genetics, 2023. DOI: 10.1038/s41588-023-01321-1. PMID: 36928603.
- Eastwood M, Sailem H, Marc ST, et al. “MesoGraph: Automatic profiling of mesothelioma subtypes from histological images.” Cell Reports Medicine, 2023. DOI: 10.1016/j.xcrm.2023.101226. PMID: 37816348. PMCID: PMC10591053.
- Seyedshahi F, Rakovic K, Poulain N, et al. “A histomorphological atlas of resected mesothelioma discovered by self-supervised learning from 3446 whole-slide images.” Nature Communications, 2025. DOI: 10.1038/s41467-025-63846-9. PMID: 41057342.
- McGenity C, Clarke EL, Jennings C, et al. “Artificial intelligence in digital pathology: a systematic review and meta-analysis of diagnostic test accuracy.” npj Digital Medicine, 2024. DOI: 10.1038/s41746-024-01106-8. PMID: 38704465. PMCID: PMC11069583.
- Fanaroff RE, Yang SR, Tan KS, et al. “Correlation of Histologic Features with Gene Alterations in Pleural Mesothelioma.” Modern Pathology, 2025. DOI: 10.1016/j.modpat.2025.100706. PMID: 39788204.
Disclaimer: This blog post is a simplified summary of published research for educational purposes. The accompanying illustration is artistic and does not depict actual model architectures, data, or experimental results. Always refer to the original paper for technical details.