Evaluating and Refining Al Model for TB Identification from Chest X-rays
试验速览
- 阶段
- 不适用
- 状态
- 尚未招募
- 发起方
- 入组人数
- 1,000
- 试验地点
- 1
- 主要终点
- Diagnostic accuracy of ImagiXAI-CXR for Active TB detection from chest X-rays, quantified by sensitivity, specificity, PPV, NPV, and AUC, using confirmatory clinical and radiological findings as reference standards, assessed at the time of retrospective analysis.
研究概览
简要总结
This retrospective observational study at KLE Hospital, Belagavi, aims to evaluate the diagnostic accuracy of ImagiXAI-CXR, an AI-based system for detecting tuberculosis (TB) from chest X-rays. A total of 1,000 retrospective CXR cases from the hospital’s PACS archive will be analyzed, including both TB-positive and TB-negative patients. Reference standards will include confirmatory tests such as High-Resolution CT (HRCT), nucleic acid amplification tests (NAAT), and documented clinical diagnoses. The primary objective is to determine the sensitivity, specificity, positive predictive value, negative predictive value, and area under the ROC curve (AUC) for Active TB detection. All AI misclassifications will be reviewed by a radiologist and clinical team to identify error patterns, and findings will be used for iterative AI model refinement. The study’s outcomes will inform the integration of AI into real-world TB screening workflows to improve diagnostic accuracy and public health impact.
研究设计
- 研究类型
- Observational
入排标准
- 年龄范围
- 18.00 Year(s) 至 90.00 Year(s)（—）
- 性别
- All
入选标准
- •Patients who have undergone a chest X-ray as part of their diagnostic evaluation for suspected TB or other respiratory illness within the study period.
- •Availability of confirmatory test results (e.g., NAAT such as GeneXpert or Truenat, HRCT, or clinical diagnosis).
- •Complete imaging and clinical records available in the PACS and hospital database.
排除标准
- •Patients younger than 18 years of age.
- •Poor-quality chest X-rays (e.g., underexposure, overexposure, excessive motion artifacts, or significant image degradation).
- •Incomplete imaging records or missing confirmatory test results.
- •Cases lacking adequate clinical metadata or follow-up information.
- •Duplicate imaging records of the same episode without added diagnostic value.
结局指标
主要结局
Diagnostic accuracy of ImagiXAI-CXR for Active TB detection from chest X-rays, quantified by sensitivity, specificity, PPV, NPV, and AUC, using confirmatory clinical and radiological findings as reference standards, assessed at the time of retrospective analysis.
时间窗: 16 weeks
次要结局
未报告次要终点
研究者
Dr Pradeep S Goudar
KLES Dr. Prabhakar Kore Hospital & Medical Research Centre