Peripheral arterial and lung biopsy impedance studies

Three study programmes: SEPARATE (in vivo peripheral arterial clot trial, 17 patients), E-SEPARATE (ex vivo thrombus validation versus histology, 15 patients), and INSPECT (first in human lung tissue classification during bronchoscopic biopsy, 26 patients). I led ML for intravascular impedance tissue classification across these full cycle clinical trial programmes. I automated the pipeline from data ingestion through feature extraction to inference, with uncertainty quantification at labelling and inference for high-risk clinical AI use.

Built an impedance feature set reused across peripheral arterial and lung oncology clinical trials without rebuilding pipelines per study.

Psychiatric treatment-response from multimodal biomarkers

Owned biosignal and imaging feature extraction (EEG, ECG, galvanic skin response) for a transprognostic multimodal biomarker algorithm predicting treatment response across major depressive disorder (MDD), attention-deficit/hyperactivity disorder (ADHD), obsessive-compulsive disorder (OCD), and post-traumatic stress disorder (PTSD). Managed automated biosignal data collection and analysis (EEG, ECG, galvanic skin response) for 12,000+ subjects across clinical trial cohorts.

The validated algorithm achieved external validation on unseen clinical cohorts (TRIPOD Type 4) and ranked #1 in an international competition.

Validation across studies

The impedance and EEG work above shares one validation thread: protocol-aligned pipelines from ingestion through feature extraction and inference, with uncertainty quantification at labelling and inference for high-risk clinical use. The same impedance feature engineering supported peripheral arterial trials (n = 17 in vivo, n = 15 ex vivo) and the INSPECT first in human lung study (n = 26). Multimodal psychiatric models met external validation on pooled clinical trial cohorts (TRIPOD Type 4). Delivery ran with investigator teams in Belgium, Australia, and France.