Advanced Pregnancy Risk Assessment Using AI Model Chain by Kiruthika Subramani

In this session, I will explore the innovative integration of ML Model and the Gemini Large Language Model (LLM) to develop a cutting-edge predictive model aimed at enhancing pregnancy risk management. This model is designed to not only assess the severity of pregnancy-related risks but also to translate these findings into regional languages, thereby making healthcare services more accessible and accurate for diverse populations.

I will begin by providing a comprehensive overview of the model’s architecture, detailing how robust machine learning algorithms are combined with the advanced natural language processing capabilities of the Gemini LLM. This integration allows for precise risk assessment and effective communication of complex medical information in a way that is easily understandable by patients and healthcare providers alike.

Attendees will gain valuable insights into the practical application of this model in real-world scenarios. I will present case studies and examples demonstrating how the model has been used to predict and manage pregnancy risks, ultimately improving patient outcomes. Additionally, I will discuss the challenges faced during the development and deployment of the model, and how these were overcome.

Furthermore, I will highlight the benefits of combining machine learning with natural language processing in the healthcare sector. This includes improved accuracy in risk assessment, enhanced patient communication, and the potential for broader implementation across various healthcare settings. By the end of the session, attendees will have a thorough understanding of how this innovative approach can transform pregnancy risk management and contribute to better healthcare outcomes overall.

https://youtu.be/vN9QcnpKb10
DevFest Montreal 2024

Transcript

3. About Me • Upcoming Gen AI Engineer Intern at Certkor.ai • Masters Student, Mila, Canada • IBM Champion Data & AI | IBM Z | Today's Architects • Google Developer Expert in ML | Cloud Champion Innovator • 204 Tech Talk Shows across the Globe | Author of 2 Books
10. AI Model Chain • An AI model chain refers to a sequence of interconnected models that work together to achieve a complex task. It involves multiple steps, each building upon the previous one to process data, make decisions, and generate outputs
11. 1. Sentiment Analysis and Recommendation System • Task Sentiment Analysis : Model BERT • Task Recommendation System : Model Collaborative Filtering or Content-Based Filtering 2. Image Classification and Object Detection • Task Image Classification : Model ResNet • Task Object Detection : Model YOLO (You Only Look Once)
12. 3. Medical Diagnosis and Treatment Recommendation • Task Medical Diagnosis : Model CNN or LSTM • Task Treatment Recommendation : Model Reinforcement Learning 4. Fraud Detection and Prevention • Task Anomaly Detection : Model Autoencoder • Task Fraud Classification : Model Random Forest or SVM • Task Fraud Prevention : Model Reinforcement Learning
13. AI Model Chain for the pregnancy Risk Assessment
16. Let’s understand our data Key Features in the Data • Age: Age of the individual. • SystolicBP: Systolic Blood Pressure (mm Hg). • DiastolicBP: Diastolic Blood Pressure (mm Hg). • BS: Blood Sugar level ( mmol/L). • BodyTemp: Body Temperature (°F). • HeartRate: Heart Rate (beats per minute). • RiskLevel: The risk level of maternal health, categorized as high risk, mid risk, or low risk.
17. What we are going to do We developed an AI-powered maternal health support system that leverages advanced machine learning models to provide personalized care and risk assessment. By analyzing user-provided health data and emotional input, the system offers tailored recommendations, early risk detection, and multilingual support.