AI-Powered Insulin Dosage Adjustment Bot

AI-powered insulin dosage adjustment system using Artificial Neural Networks and Natural Language Processing

The Outcome

My AI-powered insulin dosage adjustment system leverages Artificial Neural Networks and Natural Language Processing to optimize patient care. Using machine learning algorithms, it analyzes patient data—such as blood sugar levels, carb intake, and activity—to recommend accurate insulin doses. The system integrates speech recognition for hands-free operation, allowing users to provide voice input and receive spoken dosage recommendations. Built with Python, Keras, NumPy, and Pandas, it ensures precision, efficiency, and real-time interaction, enhancing accessibility for diabetes management.

Technologies Used

I built this project in a Jupyter Notebook using Python and external libraries. I imported my Excel data, standardized it, and trained my model and integrated some google APIs for a speech-to-text feature

Visit GitHub Repository

01

Artificial Neural Network

Developed using Keras with a Sequential model, employing Dense and Dropout layers for improved generalization.

02

Natural Language Processing

Implemented speech recognition using speech recognition with Google’s Speech API, and text-to-speech conversion via pyttsx3, enabling seamless voice interaction.

03

Data Preprocessing and Model Evaluation

Processed dataset using Pandas for data manipulation, NumPy for numerical operations, and scikit-learn for standardization, training/test splitting, and performance evaluation.

04

Languages and External Libraries

Built with Python, utilizing Jupyter Notebook for development.

Process Abstraction

Discovery & Research

Step One

General Questionnaire

For hyper/hypoglycemia diagnosis

Step Two

Speech and Text Options

NLP

Step Three

Data Training and Standardizing

ANN

Step Four

Understanding User Input

Bodyweight, physical activity, carb intake...etc.

Step Five

Final Diabetes Insulin Dosage