https://drive.google.com/file/d/1HZWHd1j49IwhbGYe04x_S8kMERsFslb8/view?usp=sharing
EquiHER: AI for Gender-Equal Healthcare
Built With
Python, TensorFlow/Keras, NumPy, Pandas, Scikit-learn, Matplotlib, Google Colab, GitHub
About the Project
Inspiration
EquiHER started from a real problem in healthcare: women are often diagnosed later or incorrectly because symptoms can present differently, and many medical datasets have historically been male-centered. I wanted to build something that supports fairer diagnosis, not just “better prediction.”
What it does
EquiHER is a diagnostic risk flagging tool. It uses a neural network trained on synthetic medical data to identify patients who may be at higher risk of being misdiagnosed (with extra attention to patterns that can affect women).
It analyzes 15 clinical variables and reaches about 84-87% accuracy.
How I built it
- Created a synthetic dataset with medically realistic ranges
- Cleaned and normalized the features
- Trained a neural network (15 inputs → 128 → 64 → 32 → output)
- Evaluated accuracy and tracked loss curves
- Built a simple interface concept for entering patient values and returning a risk flag
The model (with LaTeX)
The model is trained using cross-entropy loss:
$$ L = - \sum_{i=1}^{n} y_i \log(\hat{y}_i) $$
And updated by gradient descent:
$$ \theta_{t+1} = \theta_t - \eta \nabla L(\theta) $$
Challenges
- Real medical data is hard to access, so I had to design synthetic data carefully.
- Preventing overfitting while keeping accuracy strong.
- Making the idea clinician-friendly (the goal is support, not replacement).
What I learned
I learned that in healthcare, model performance, it’s also about fairness, trust, and how the tool is used.
Impact
EquiHER supports:
- SDG 3.8 (Universal Health Coverage) by improving quality through better diagnostic support
- SDG 5 (Gender Equality) by reducing gender-based healthcare bias
Log in or sign up for Devpost to join the conversation.