1063-India-Evaluating the Performance of Artificial Intelligence (AI)-based Chest X-ray Interpretation Tool vs. Radiological, Microbiological, and Clinical Diagnoses in Detecting Tuberculosis (TB)

Study Overview:
This retrospective observational study aims to assess the sensitivity, specificity, and level of agreement of the qXR Artificial Intelligence (AI) tool in detecting radiological signs of tuberculosis (TB) in chest X-ray (CXR) images from patients. The study will compare the AI tool’s performance against a radiological ground truth, microbiological confirmation, and clinical diagnosis.
A total of 500 chest X-ray images collected from active TB screening sites will be analyzed. Of these, there are images from confirmed TB cases (either microbiologically or clinically), non-TB cases, and EPTB (Extra Pulmonary TB). The goal is to generate evidence on the accuracy of qXR in detecting radiological signs of TB.
Objectives:
•        Compare the performance of qXR in detecting radiological signs of TB against the radiological ground truth.
•        Compare the performance of qXR interpretation against microbiological and clinical ground truth.
•        Evaluate the level of agreement between qXR results and the radiological ground truth.
•        Evaluate the level of agreement between qXR results and microbiological ground truth.
Endpoints:
•        Sensitivity: Sensitivity of qXR in detecting radiological signs of TB compared to radiological ground truth, microbiological, and clinical ground truth.
•        Specificity: Specificity of qXR in detecting radiological signs of TB compared to radiological ground truth, microbiological, and clinical ground truth.

Students will be involved in the following tasks during their in-country placement:
•        Evaluation of Chest X-Rays:
Students will analyze the sensitivity and specificity of the qXR AI tool in detecting TB compared to the radiological ground truth, microbiological results, and clinical diagnosis. This will be done through qualitative analysis, where students will assess the level of agreement between the AI tool’s results and these reference standards. Students will also identify and document discrepancies, providing insights into the potential limitations or strengths of the AI tool in diagnosing TB.
•        Manuscript Preparation:
Based on the findings, students will assist in preparing a manuscript for publication. The student will help interpret the results and will also assist in drafting conclusions and recommendations based on their findings, contributing to the overall publication process.

Type of Setting: Indian Government
Primary Locations:BJGMC Sassoon General Hospitals, Pune.

This placement would be a good fit for someone with a strong interest in leveraging AI and medical imaging for global health research. Ideal candidates should be passionate about advancing diagnostic tools, particularly in the context of tuberculosis, and have an aptitude for data analysis and manuscript preparation. A background in public health, infectious diseases, or radiology would be beneficial, as well as the ability to work independently on research tasks

1.        Language Skills:
•        English (required); local languages- Marathi/Hindi (optional) 
2.        Medical Knowledge about Infectious Diseases (TB):
•        Understanding of TB pathogenesis, transmission, and clinical manifestations, especially pulmonary TB, and familiarity with diagnostic methods.
3.        AI Knowledge in Healthcare:
•        Understanding of AI algorithms, like qXR, and their application in interpreting chest X-rays (CXR) for TB diagnosis.
4.        Data Analysis:
•        Proficiency in quantitative software (e.g., R) for analyzing data.
5.        Manuscript Writing Skills:
•        Understanding of mixed-method research approaches and experience in manuscript writing, including literature reviews and contributing to scientific papers.
6.        Additional Skills:
•        Ability to work in a diverse and multicultural environment, strong verbal and written communication skills for report writing and team collaboration, and strong organizational skills for monitoring study progress, documentation, and task management.

Name: Vidya Mave, lkodali3@jh.edu 

I would describe my communication and mentorship styles as... 

I would describe my communication and mentorship styles as supportive and collaborative, focusing on clear communication, accessibility, and fostering a positive learning environment.

A complementary student would have a working style and mentorship expectations that are..

A complementary student would have a working style and mentorship expectations that are proactive, communicative, and adaptable, with a focus on collaboration and openness to feedback.

Estimated Costs:

Housing: $500/month
Food: $250/month 
Local Transportation: $100/month 
Internet: $20/month data sim card

Total: $870/month

Additional support from PI:

Site will provide a stipend to help offset expenses.