DOI: https://doi.org/10.36347/sjams.2024.v12i12.038

Pages: 1928-1934

Introduction: Artificial Intelligence (AI) is transforming radiology globally by enhancing diagnostic accuracy, reducing human error, and optimizing workflow. In Bangladesh, where radiological services are limited due to inadequate infrastructure and a shortage of trained professionals, AI holds significant potential to improve diagnostic services, especially in underserved regions. Aim of the study: The study aimed to assess the integration of artificial intelligence into radiological practice in Bangladesh to enhance diagnostic capabilities and shape the future of radiology in the country. Methods: A meta-analysis and qualitative synthesis were conducted using data from ten studies that assessed AI-based radiological tools across CT, MRI, ultrasound, and X-ray modalities. Key variables analyzed included sensitivity, specificity, accuracy, AUC, and risk of bias. Implementation readiness in Bangladesh was evaluated based on infrastructure, personnel, policy, and data availability. Results: The meta-analysis reviewed ten studies on imaging modalities such as X-ray, CT, MRI, and ultrasound, with sample sizes from 500 to 23,000. AI models showed strong diagnostic performance, with a pooled sensitivity of 90% and specificity of 88%, and an AUC of 0.92. Notably, MRI detection of multiple sclerosis using 2D-3D CNN achieved 98.8% accuracy, while brain tumor detection reached 94.5%. In Bangladesh, AI tools like qXR showed 90.2% sensitivity for tuberculosis in chest X-rays. Most of the 22 studies focused on image interpretation, with low risk of bias. Subgroup analysis indicated slightly lower sensitivity (89%) and specificity (84%) in low- and middle-income countries compared to high-income countries (91% and 88%). Bangladesh's digital imaging infrastructure was rated moderate, highlighting gaps in trained personnel, policies, and data availability, and indicating a need for improvements to support AI in radiological diagnostics. Conclusion: AI in radiology shows promise in