Healthcare AI Penn State · CPHK Pemba · ESWA (IF 7.5) Active Multi-Site Clinical Consortium
PlacentaVision, Maternal Biomarkers & Endoscopic Boundary Clinical AI
Non-invasive placental ultrasound weight and perfusion modeling (PlacentaVision), explainable preterm birth serum biomarker risk models in Pemba, and BAFNet boundary-aware polyp segmentation (ESWA 2025).
Scientific Context & Approach
Maternal morbidity and preterm delivery disproportionately burden rural and island clinics that lack tertiary imaging specialists or invasive pathology workflows.
Through multi-institutional clinical collaborations spanning Penn State University, CPHK Pemba, and clinical computer vision partners, SAAIL Lab builds calibrated models for bedside placental morphometry, SHAP-explainable neonatal risk stratification, and contour-aware endoscopic segmentation.
Core Technical Objectives
- Predict placental weight, perfusion status, and perinatal risk from routine bedside imaging under the PlacentaVision consortium.
- Deliver SHAP-explainable preterm birth and neonatal mortality risk scoring validated with Pemba health cohorts.
- Advance ultra-efficient boundary-aware segmentation networks such as BAFNet (0.914 EndoScene Dice in 5.98M parameters).
Verified Deliverables & Outputs
- BAFNet endoscopic segmentation published in Expert Systems with Applications (2025, IF 7.5)
- SHAP-explainable maternal preterm biomarker pipelines across Zanzibar & Pemba cohorts
- Joint placental imaging morphometry & 4 co-supervised public health PhD studies