NCD Detection Framework

Novel framework for noncommunicable disease detection via prompt engineering and domain knowledge integration (AEJ 2025)

Alexandria Engineering Journal (2025)

A novel framework combining prompt engineering and domain knowledge integration for early detection of noncommunicable diseases (NCDs). By leveraging LLM-guided feature engineering, we achieved improved accuracy in NCD prediction tasks.

Key Achievements

  • LLM-Guided Feature Engineering: Integrated domain knowledge through prompt engineering
  • Improved Accuracy: Outperformed traditional ML baselines on NCD prediction
  • Early Detection Focus: Designed for proactive healthcare interventions
  • Clinical Validation: Tested on real-world clinical datasets

Methodology

Approach:

  1. Domain knowledge extraction via LLM prompting
  2. Automated feature engineering guided by medical expertise
  3. Ensemble learning with engineered features
  4. Clinical outcome prediction for NCDs

Diseases Covered: Cardiovascular disease, diabetes, chronic respiratory diseases, cancer

Framework: GPT-based prompting, scikit-learn, PyTorch

Clinical Impact

This work demonstrates how large language models can enhance traditional machine learning for healthcare by incorporating medical domain knowledge through natural language prompting, making sophisticated clinical prediction more accessible.

Status: Published in Alexandria Engineering Journal (2025), Volume 133, pp. 586-614 Authors: Mohammad Junayed Hasan, Suhra Noor, Sifat Momen (*Equal contribution)

References