Research Interests

Exploring multimodal medical research

Research Focus

1. Multimodal Data Integration in Oncology - MINDS and HONeYBEE Frameworks

Developing AI platforms that unify fragmented cancer data (radiology, pathology, genomics, proteomics, clinical records) into comprehensive patient profiles for treatment decisions. HONeYBEE (published in npj Digital Medicine, IF: 15.1) transforms complex medical data into AI-ready formats, addressing the critical challenge where 80% of cancer data remains siloed, potentially improving diagnostic accuracy by 30-40%.

2. Foundation Models and Deep Learning for Medical AI

Adapting state-of-the-art architectures like Transformers and Graph Neural Networks to identify subtle patterns in cancer biology across thousands of cases. This research accelerates drug discovery timelines and identifies novel therapeutic targets by enabling AI to understand complex tumor relationships that human experts might miss.

3. Multi-Omics Integration for Precision Medicine - SeNMo Platform

Created SeNMo (Self-normalizing Multi-omics), an AI system that analyzes complete molecular tumor profiles to predict patient survival and treatment response with 85% accuracy. The platform enables personalized treatment recommendations by overcoming incompatible data format limitations across different molecular tests.

4. AI-Powered Medical Imaging and Digital Pathology

Building automated systems that analyze gigapixel pathology slides and radiological scans to detect microscopic cancer characteristics, grade aggressiveness, and predict metastatic risk. This technology reduces pathologist workload by 60% while maintaining 95% diagnostic accuracy, addressing critical expertise shortages in underserved areas.

5. Clinical NLP and Automated Data Extraction - CLEVER System

Developing CLEVER, an AI system using consensus-based reasoning to extract clinical information from unstructured pathology reports with 92% accuracy. This tool transforms years of text reports into structured research databases, eliminating months of manual chart review and accelerating clinical research across thousands of patients.

Key Research Platforms

HONeYBEE

Published in npj Digital Medicine (IF: 15.1)
Enabling scalable multimodal AI in oncology through foundation model-driven embeddings. Transforms complex medical data into AI-ready formats, addressing the critical challenge where 80% of cancer data remains siloed.

View Publication

SeNMo

Published in IJMS
Self-normalizing Multi-omics neural network for pan-cancer prognostication with 85% accuracy. Enables personalized treatment recommendations by analyzing complete molecular tumor profiles.

View Publication

CLEVER

Under Review
Consensus-based reasoning with locally deployed LLMs for structured data extraction from surgical pathology reports. Achieves 92% accuracy in extracting clinical information from unstructured text.

View Preprint

EAGLE

Under Review
Efficient Alignment of Generalized Latent Embeddings for multimodal survival prediction with interpretable attribution analysis. Advances cross-modality correlation analysis for treatment outcome prediction.

View Preprint