Written by : Jayati Dubey
January 31, 2025
DeepMerkel is designed to analyze patient-specific and tumor-specific factors to predict treatment outcomes.
A team of researchers led by Newcastle University has developed an AI-powered system to predict the severity and progression of Merkel cell carcinoma (MCC), a rare but aggressive form of skin cancer.
The AI system, called DeepMerkel, combines machine learning with clinical expertise to generate personalized predictions, helping doctors and patients make more informed treatment decisions.
The findings, published in Nature Digital Medicine and the Journal of the American Academy of Dermatology, demonstrate how AI can revolutionize prognostication and treatment planning for MCC and potentially other aggressive skin cancers.
DeepMerkel is designed to analyze patient-specific and tumor-specific factors to predict treatment outcomes.
By integrating AI with clinical data, the system enables a more precise and individualized approach to managing MCC.
Dr Tom Andrew, a Plastic Surgeon and CRUK-funded PhD student at Newcastle University, and the study’s first author, emphasized the importance of this development.
"DeepMerkel allows us to predict the course and severity of Merkel cell carcinoma, enabling personalized treatment so that patients receive optimal management," he said.
The research highlights how AI can uncover subtle patterns in clinical data, improving the accuracy of patient-specific predictions.
This is particularly significant given that MCC cases have doubled in the last two decades, disproportionately affecting older adults.
The study was conducted in collaboration with leading experts, including Professor Penny Lovat, an expert in dermato-oncology at Newcastle University, and Dr Aidan Rose, Senior Clinical Lecturer at Newcastle University and Consultant Plastic Surgeon at Newcastle Hospitals NHS Foundation Trust.
Dr Rose highlighted the role of AI in refining patient prognosis, said, "Accurately predicting patient outcomes is crucial for guiding clinical decision-making, especially for aggressive skin cancers. AI developments like DeepMerkel allow us to provide personalized survival predictions and inform medical teams about the most effective treatment options."
The researchers used advanced statistical and machine learning techniques to develop DeepMerkel, including explainability analysis and deep learning feature selection.
They incorporated a modified XGBoost framework to enhance the system’s predictive capabilities.
Their findings in Nature Digital Medicine highlighted how DeepMerkel could identify mortality risk factors, providing valuable insights into disease progression.
In the Journal of the American Academy of Dermatology, the researchers detailed their analysis of nearly 11,000 patients from two countries, showing that DeepMerkel could identify high-risk patients at earlier stages of MCC.
This early detection enables healthcare providers to make better-informed decisions about radical treatments and intensive disease monitoring.
The research team envisions DeepMerkel as a tool to enhance patient-clinician collaboration, ensuring individuals have access to personalized information that can guide their treatment choices.
Dr Andrew emphasized the need for continued investment, and said, "Our next step is to further develop DeepMerkel so it can present clinicians with treatment pathway options. With further integration into routine clinical practice, we hope to expand its use to other tumor types."
The researchers are optimistic that AI-driven systems like DeepMerkel will reshape cancer prognostication and improve precision medicine, making treatments more tailored, effective, and patient-centered.
Stay tuned for more such updates on Digital Health News.