Research indicates that artificial intelligence (AI) can help assess the progression and seriousness of aggressive skin cancers like Merkel cell carcinoma (MCC), improving clinical decisions by offering personalized outcome predictions based on specific treatment for both patients and healthcare providers.
Artificial intelligence can assess the progression and seriousness of aggressive skin cancers, such as Merkel cell carcinoma (MCC), improving clinical decisions by offering personalized treatment outcome predictions for patients and their healthcare providers.
A global team, spearheaded by Newcastle University researchers in the UK, has fused machine learning with medical expertise to create a web-based platform called “DeepMerkel.” This innovative system predicts treatment outcomes for MCC based on individual and tumor-specific characteristics.
They suggest that this tool could also be utilized for other aggressive skin cancers, allowing for precise prognostication, enhancing informed clinical decisions, and improving patient options.
MCC
MCC is a rare yet highly aggressive form of skin cancer, notoriously challenging to treat. It typically impacts older individuals with weakened immune systems, often diagnosed at advanced stages, which correlates with poor survival rates.
Dr. Tom Andrew, a Plastic Surgeon and PhD candidate funded by Cancer Research UK at Newcastle University and the lead author, stated, “DeepMerkel is enabling us to predict the progression and intensity of Merkel cell carcinoma, allowing us to tailor treatments so that patients receive optimal care.”
“Utilizing AI has helped us identify intricate new patterns and trends in the data, which enables us to provide more accurate predictions on an individual basis.”
“This is crucial as the rate of MCC diagnoses has doubled in the two decades leading up to 2020. While still rare, it is an aggressive cancer that increasingly affects older populations.”
The study was conducted alongside Penny Lovat, a Professor of Dermato-oncology at Newcastle University, and Dr. Aidan Rose, a Senior Clinical Lecturer at Newcastle University and Consultant Plastic Surgeon at Newcastle Hospitals NHS Foundation Trust.
Dr. Rose remarked, “Accurately predicting patient outcomes is essential for guiding clinical decisions, especially in complex cases involving aggressive skin cancer. This leads to challenging, sometimes life-altering treatment choices. Our advancements with AI allow us to offer personalized survival forecasts and inform a patient’s medical team about optimal treatment strategies.”
In two related publications in Nature Digital Medicine and the Journal of the American Academy of Dermatology, the team outlines how they utilized sophisticated statistical analysis and machine learning to create the web-based prognostic tool, DeepMerkel.
Method
In Nature Digital Medicine, the research team explains how they applied explainability analysis alongside patient data to unveil new insights regarding mortality risk factors for MCC. They also combined deep learning feature selection with an enhanced XGBoost framework to develop the DeepMerkel prognostic tool.
By analyzing data from nearly 11,000 patients across two countries, the researchers illustrated in the Journal of the American Academy of Dermatology how DeepMerkel successfully identified high-risk patients at earlier stages of cancer, enabling healthcare professionals to make more informed choices regarding when to employ radical treatment methods and intensive monitoring.
Patients First
The team envisions that DeepMerkel will empower patients with better information to collaborate with their medical teams in determining the most suitable treatment options for them individually.
Dr. Andrew added, “With additional funding, our next exciting step is to further enhance DeepMerkel, allowing the system to suggest options that assist clinicians in identifying the best treatment pathways available.”
The forthcoming goal is to incorporate the DeepMerkel platform into standard clinical practice and expand its application to other tumor types.
