PhD Defence Sheyasi Pathak | Machine Learning in Healthcare: From Petri-dish AI to Reality-centric AI

We are thrilled to share that Dr. Shreyasi Pathak successfully defended her PhD on 9 September 2025.

Her thesis, titled "Machine Learning in Healthcare: From Petri-dish AI to Reality-centric AI", tackles one of the most pressing challenges in medical AI: bridging the gap between research prototypes and clinical practice.

While deep learning has shown immense promise in healthcare, many models remain “petri-dish AI”—highly experimental, yet unfit for real-world workflows. Shreyasi’s work reimagines this landscape by laying the foundation for reality-centric AI in healthcare—models that are:
1️⃣ Built on real-world data
2️⃣ Designed for real-world hospital settings
3️⃣ Delivering outputs that are clinically meaningful

Her thesis investigates these pillars across three critical use cases:

• 🩺 Breast cancer diagnosis using mammography
• 💤 Sleep stage prediction from electrophysiological signals
• 🏥 Post-operative mortality prediction in elderly hip fracture patients

Highlights include:
✅ Privacy-preserving data sharing through a novel model-to-data platform
✅ Advances in handling weak labels, multimodal data, and interpretability
✅ New metrics for evaluating prototype-based explanations in clinical AI

Shreyasi’s research paves the way for AI that is not just innovative but also usable—a crucial step toward improving patient care and supporting clinicians worldwide.

We couldn’t be prouder of this incredible achievement—congratulations, Dr. Pathak! 👏