A recent open-access study published in Frontiers in Oncology highlights a novel, non-invasive approach for classifying pancreatic neoplasms using machine learning and plasma-derived extracellular vesicles (EVs). The research, led by teams from the National and Kapodistrian University of Athens and the University of Ioannina, was supported by the PHOREVER project and aligns directly with its goals to improve diagnostic technologies through photonics-enhanced platforms.
🔬 Why it matters:
Pancreatic cancer remains one of the deadliest forms of cancer, often detected too late for curative treatment. Traditional biomarkers like CA19-9 offer limited accuracy, especially in early stages. This study introduces a flow cytometry-based pipeline that uses EVs from blood plasma—tiny membrane-bound particles secreted by cells—as diagnostic indicators. These EVs, when combined with biochemical and hematological data, were analyzed using advanced machine learning algorithms to stratify patients into three classes: benign lesions, exocrine tumors, and endocrine neoplasms.
📊 Key findings:
-
The models achieved over 96% accuracy in multiclass classification tasks.
-
The most impactful features were EV subpopulations marked by CD45, CD63 and EphA2, along with select blood-based parameters like glucose and bilirubin.
-
Explainable AI tools (SHAP) confirmed the contribution of EV-based features to the model’s diagnostic power.
-
External validation with new patient samples confirmed the pipeline’s robustness and clinical relevance.
💡 What this means for PHOREVER:
The study validates the clinical potential of EV-based diagnostics enhanced by photonics and intelligent data processing—core pillars of the PHOREVER vision. It also strengthens the role of flow cytometry and AI as combined tools for early-stage cancer detection.
📚 Read the full article here.
Stay tuned as PHOREVER continues to push the boundaries of non-invasive diagnostics!