R&D at the Heart of Innovation
At AkiliAI, we are dedicated to advancing the intersection of artificial intelligence and healthcare. Our research endeavors focus on developing innovative solutions that enhance patient care, streamline clinical workflows, and contribute to the broader medical community. Below is an overview of our ongoing and past research projects.
Ongoing Research
Evaluating the Use of Large Language Models (LLMs) for Medical Review Classification
Addressing the Limitations of Traditional NLP and Sentiment Analysis in Healthcare
Background
Traditional Natural Language Processing (NLP) and sentiment analysis models often struggle with the complexity and specificity of medical language, leading to suboptimal performance in healthcare applications. Recent advancements in Large Language Models (LLMs) present an opportunity to overcome these challenges.
Objective
This research project aims to evaluate the effectiveness of LLMs in classifying and analyzing medical reviews, focusing on improving accuracy and reliability over conventional NLP methods.
Methodology
- Data Collection: Gather a comprehensive dataset of medical reviews and clinical notes.
- Model Selection and Training: Fine-tune state-of-the-art LLMs on the collected dataset to accurately capture medical terminologies and context.
- Performance Evaluation: Compare the performance of LLMs against traditional NLP models in tasks such as sentiment analysis, topic classification, and information extraction.
- Analysis of Limitations: Identify and document any limitations or challenges encountered when applying LLMs to medical text, such as handling domain-specific jargon or context.
Expected Outcomes
- Improved Classification Accuracy: Demonstrate that LLMs can more accurately classify medical reviews compared to traditional models.
- Enhanced Sentiment Analysis: Achieve a more nuanced understanding of sentiments expressed in medical reviews, leading to better patient feedback analysis.
- Implementation Guidelines: Develop best practice guidelines for integrating LLMs into healthcare applications, addressing potential limitations and ethical considerations.
By exploring the potential of LLMs in medical review classification, this research aims to contribute to more effective and reliable healthcare data analysis tools.
Past Research
Machine-Guided Tool for Assessing HIV Risk Behavior
Overview
This project focused on developing and evaluating a machine-guided tool designed to assess behaviors associated with the risk of contracting HIV. Conducted across three sites in South Africa, the study aimed to establish the efficacy of machine learning algorithms in identifying individuals at high risk, thereby facilitating targeted counseling and testing interventions.
Key Highlights
- Objective: Determine the utility of a machine-guided tool in computing HIV risk assessment scores based on subjective and objective responses from participants.
- Methodology: Enrolled 200 HIV-negative adults across three South African sites. Participants completed a behavioral questionnaire via a digital application, followed by HIV testing conducted by trained professionals. The tool’s risk assessment scores were compared against actual HIV test results to evaluate accuracy.
- Outcomes: Provided insights into the correlation between specific behavioral patterns and HIV risk, demonstrating the potential of machine learning models in enhancing HIV prevention strategies.
For a detailed overview of the study and its findings, you can access the full article here: Utility of a Machine-Guided Tool for Assessing Risk Behavior Associated With Contracting HIV in Three Sites in South Africa: Protocol for an In-Field Evaluation