How I used artificial intelligence to build a disease prediction tool—and why this matters for the future of healthcare in underserved communities.
Introduction:
Chronic diseases like diabetes continue to strain healthcare systems across the globe—especially in developing regions where early diagnosis is often unavailable. During my undergraduate studies, I developed a predictive model that uses artificial neural networks to identify individuals at risk of developing diabetes. This article explores how I built the tool, the technology behind it, and why AI is becoming critical in advancing global public health.
Why Diabetes?
Diabetes is a silent threat. In many African countries, including Zimbabwe where I come from, patients often discover their condition only after complications have already developed. Preventive diagnosis, especially using data-driven methods, can change that narrative.
The Vision:
My goal was to develop a system that could:
- Accept patient data as input (age, BMI, glucose levels, etc.)
- Use machine learning to classify risk levels
- Support healthcare workers in making early interventions
The Approach:
I chose a Backpropagation Neural Network (BPNN) architecture due to its effectiveness in pattern recognition tasks.
Key steps:
Data Collection: I used a preprocessed dataset from the Pima Indians Diabetes Database (a standard benchmark in health AI research).
Feature Selection: Attributes included glucose concentration, age, BMI, pregnancies, and insulin levels.
Training: The neural network was trained using a supervised learning approach.
I experimented with learning rates, hidden layers, and activation functions to maximize accuracy.
Accuracy & Validation: After tuning, I achieved a predictive accuracy of approximately 82%—an encouraging result for a first implementation.
Technology Stack:
Programming Language: Java (Java EE)
ML Framework: Custom implementation (no libraries like TensorFlow)
Front-end: HTML for basic UI
IDE: NetBeans
Challenges I Faced:
Limited access to computational resources
Balancing training data to avoid bias
Interpreting results in a clinically meaningful way
Why This Matters:
While this was a university project, it reflected a deeper concern: how can we bring advanced technologies to the frontlines of public health?
With funding I have the opportunity to lead in ethical AI innovation—and solutions like this, if scaled and refined, can:
- Empower community health centers
- Reduce burden on hospitals
- Detect diseases early in high-risk populations
What’s Next:
I am continuing to explore AI in image analysis, with a focus on improving disease detection through medical imaging. My goal is to contribute to research that delivers real-world healthcare impact—especially for communities that have historically been underserved.
Final Thoughts:
AI is more than just algorithms—it's a tool to bridge the gap between potential and access. As engineers and scientists, our job is to make that bridge walkable.
Let’s Connect:
If you're working in AI for health or are passionate about making tech more inclusive, I’d love to hear from you. Connect with me on LinkedIn or check out my GitHub here.
