AI tool decodes brain tumor genome during surgery

New AI approaches overcome these challenges.

This tool, called CHARM (Cryosection Histopathology Assessment and Review Machine), Available for free to other researchers. According to the research team, it needs to be clinically validated through testing in real-world settings and approved by the FDA before it can be introduced into hospitals.

Decoding the molecular code of cancer

Recent advances in genomics have enabled pathologists to distinguish between molecular features and the behaviors that such features predict, not only within specific types of brain tumors, but across different types of brain tumors. For example, gliomas are the most malignant brain tumors, The most common form of brain tumor – has three major subvariants with different molecular markers and different proliferation and spreading propensities.

The ability of new tools to expedite molecular diagnostics may be of particular value in areas with limited access to technology to perform rapid cancer gene sequencing.

Beyond decisions made during surgery, knowing a tumor’s molecular type provides clues about its aggressiveness, behavior, and potential response to different treatments. Such knowledge may aid in postoperative decisions.

In addition, new tools enable symptomatic intraoperative diagnostics. World Health Organization recently updated classification system Glioma diagnosis and grading of severity require that such diagnosis be based on the tumor’s genomic profile.

training charm

CHARM was developed using 2,334 brain tumor samples from 1,524 glioma patients from three different patient populations. When tested on a series of never-before-seen brain samples, the tool identifies tumors with specific molecular mutations with 93% accuracy, leading to distinct prognoses and distinct responses to treatment. successfully classified three major types of glioma with distinct characteristics.

Going a step further, the tool was able to capture visual features of tissue surrounding malignant cells. This allowed us to spot distinct areas of high cell density and high cell death within the sample, both indicative of a more aggressive type of glioma.

This tool will pinpoint clinically significant molecular alterations in a subset of low-grade gliomas, a subtype of glioma that is less malignant and therefore less likely to invade surrounding tissue. I was able to Each of these alterations indicates a different propensity for proliferation, spread, and therapeutic response.

The tool further correlated cell appearance (nuclear shape, presence of pericellular edema) with the molecular profile of the tumor. This means that the algorithm can accurately identify the association between cell appearance and tumor molecular type.

The ability to assess a broader context around an image makes the model more accurate, said Yu, and brings it closer to how human pathologists visually assess tumor samples.

The researchers say the model was trained and tested on glioma samples, but could be successfully retrained to identify other brain tumor subtypes.

Scientists are already designing AI models to profile other types of cancer (colon, lung, breast), but gliomas suffer from molecular complexity and large variability in tumor cell shape and appearance. , remains particularly difficult.

Yu said the CHARM tool needs to be retrained regularly to reflect new disease taxonomies that emerge from new knowledge.

“Like human clinicians who need to engage in ongoing education and training, AI tools need to stay up to date with the latest knowledge to maintain peak performance.”

Author, Funding, Disclosure

Collaborators included McLean P. Nasrallah, Junhan Zhao, Chen Che Tsai, David Meredith, Eliana Marostica, Keith L. Rigon, and Jeffrey A. Golden.

This work was supported in part by the National Institute of General Medical Sciences grant R35GM142879, the Google Research Scholar Award, the Blavatnik Center for Computational Biomedicine Award, the Partners Innovation Discovery Grant, and the Schlager Family Award for early-stage digital health innovation. rice field.