I recently had the opportunity to speak at the PharML workshop during ECML-PKDD 2026 in Naples, Italy. As one of the leading and most prestigious academic conferences in machine learning and data mining in Europe, it was an incredible honor to present our research, “Learning from VAE Errors to support ECG-based Differential Diagnosis of Myocardial Scar,” through both a Power Pitch and a poster session.
During the presentation, I shared how our team utilized DTW-based reconstruction errors from a lightweight VAE model—trained exclusively on normal ECGs—to detect scar-related anomalies. A central theme of my talk was the clinical importance of ensuring the model correctly identifies the positive patients, helping to prioritize accessible, low-cost ECG screening over expensive imaging methods.
It was a fantastic experience taking the stage at such a high-caliber event, exchanging ideas with fellow researchers, and discussing the future of interpretable AI in cardiology.
Read the full paper: http://arxiv.org/abs/2609.05294
