Developments in AI for Cancer Care: January 2025 Roundup

📰 Ongoing research at University of Illinois Urbana-Champaign, and led by surgical oncologist Dr. Claudius Conrad, aims to revolutionize surgical precision in liver, pancreas, and bile duct tumor removal. The work also aims to optimize surgical planning and real-time decision making. Dr. Conrad’s team is leveraging Fujifilm’s Vincent system, which uses AI algorithms to convert standard CT scans into ultra-high-definition 3D models that give surgeons detailed views of tumors, blood vessels, and surrounding tissues. Read more here.

📰 In a study led by medical oncologist Dr. Ravi B. Parikh, a team of researchers from Emory University and University of Pennsylvania developed a new AI-powered platform called TrialTranslator, which could help doctors identify individual patients or subgroups that may or may not respond well to a given therapy, based on clinical trial results. Doctors could make better informed treatment decisions as a result with patients. Read more here.

📰 In a study led by Diego Chowell, PhD, and Luc Morris, MD, a team of researchers from Mount Sinai and Memorial Sloan Kettering Cancer Center developed an AI-powered machine learning model called SCORPIO, which could materially advance precision oncology by predicting how well cancer patients will respond to immune checkpoint inhibitors. In the study, the model’s predictions were more accurate than widely validated biomarkers like tumor mutation burden (TMB). It is also potentially more cost-effective because it leverages data from routine blood tests. Read more here.

📰 Researchers at Newcastle University have developed an AI and machine learning-powered prognostic tool called DeepMerkel, which could help guide personalized clinical decision making and predict outcomes in Merkel cell carcinoma and other aggressive skin cancers based on personal and tumor-specific features. In a data pool of nearly 11,000 patients, the tool was able to accurately identify high-risk patients at earlier MCC stages. Read more here.

📰 Researchers at Stanford Medicine have developed an AI model called MUSK (multimodal transformer with unified mask modeling). Trained on millions of medical images of standard pathology slides and pathology-related text, the tool was able to identify people with non-small cell lung cancer that benefitted from immunotherapy, better than using PD-L1 expression. It also improved upon existing AI models in identifying people with melanoma most likely to experience disease recurrence within five years. Read more here.

📰 Stanford Medicine introduced a new AI tool that helps physicians efficiently translate medical test results to patients in accessible language. This move is part of a broader trend across healthcare to employ AI in reducing administrative burden for doctors. PCPs across the Stanford health system are already using the tool, with plans underway to roll it out to specialists later in 2025. Read more here.

📰 AI-guided mammography screening could lead to detection of more breast cancer cases without worsening the number of unnecessary recalls for additional assessment, which is a burden on both patients and radiologists. These findings are based on the Germany-based PRAIM study of over 460,000 women who underwent screening mammography within the past few years. Read more here.

📰 Researchers at the University of Pennsylvania Perelman School of Medicine have developed an AI tool that can detect from medical imaging cellular characteristics of cancer that are otherwise very challenging to pick up. The tool, called Multi-modal Spatial Omics, or MISO, marks an advancement in the field of spatial multi-omics and has the potential to provide cell-level insights that help improve understanding of specific tumors and guide treatment plans accordingly. Researchers have seen early promise in bladder cancer, gastric cancer, and colorectal cancer. Read more here.

📰 A retrospective study led by Seema Dadhania at the Imperial College London Department of Surgery & Cancer highlighted that an AI prediction model, called C the Signs, could help with earlier detection of colorectal cancer. The AI model analyzed 20 years worth of patient data and was able to identify 29% of cancer patients as high-risk up to 5 years prior to their diagnoses. While the model showed markedly lower specificity than standard screening methods, it holds promise for detecting very early stage disease and facilitating timely interventions in CRC, which is seeing an increase in incidence in younger populations. Read more here.

📰 Research at McGill University led by Dr. Matthew Dankner and Dr. Reza Forghani showed that an AI model can detect brain metastasis with high accuracy, up to 85%. The model was trained on MRI scans from over 130 patients. The study highlights the potential role of machine learning in developing an MRI-based biomarker for determining brain metastasis invasion pattern (BMIP), which in clinical practice could help with catching the spread of cancer to the brain sooner. Read more here.

🚀 Perthera, a precision oncology technology company, announced the introduction of PDACai, the first clinically validated AI predictor for whether a pancreatic adenocarcinoma (PDAC) patient will respond better to one of the two common front-line chemotherapy regimens for PDAC. Read more here.

🚀 The DeepScribe ambient AI clinical documentation tool is coming to providers using Flatiron Health’s OncoEMR platform. Documentation and administrative tasks continue to be a key area of AI innovation to improve provider experience. Read more here.