Malaizyo Gabriel Muzumala Delivers Insightful Talk on Semi-Automated Classification and Detection of Community Acquired Pneumonia at Academic/Industry Colloquium

Introduction to the Seminar

On May 15th, as part of the 2023/24 CSC 5741 course series, the Academic/Industry Colloquium featured a groundbreaking presentation by Malaizyo Gabriel Muzumala, a research student from the Enterprise Medical Imaging in Zambia (EMIZ) project. His talk, titled "Semi-Automated Classification and Detection of Community Acquired Pneumonia," provided valuable insights into the integration of Artificial Intelligence (AI) in medical imaging, with a focus on practical applications in low-resource settings.

Abstract

The use of Artificial Intelligence (AI) in radiology is becoming increasingly mainstream, yet its adoption remains limited in low-resource settings, particularly in the Global South. In his talk, Malaizyo Gabriel Muzumala presented a study exploring the impact of AI on the turnaround time of medical image interpretation and its influence on radiologists' workload.

This study implemented two AI models—a classification model and a detection model—to assist in the semi-automated interpretation of medical images for diagnosing community-acquired pneumonia. A Web-based DICOM Viewer was developed to interface with these AI models. To determine the optimal model configuration, a focus group discussion was held with a radiologist and a radiology resident.

A comparative controlled experiment involving 12 radiology residents at a large University Teaching Hospital assessed the AI models' impact on workload and perceived usefulness. The NASA Task Load Index (TLX) and Technology Acceptance Model (TAM 2) questionnaires were used to measure these factors. The results showed a significant reduction in perceived workload with the AI solution, reflected in a NASA-TLX score of 1.86. Additionally, the AI solution's perceived usefulness was confirmed by positive responses across all eight TAM 2 constructs.

This study highlights the potential for AI to facilitate the semi-automated interpretation of medical images in low-resource settings, demonstrating how such technologies can reduce workload and improve efficiency in medical imaging practices.

Conclusion

The colloquium provided an excellent opportunity for students, professionals, and enthusiasts to explore the intersection of AI and medical imaging. Malaizyo Muzumala’s presentation offered a comprehensive overview of his research, showcasing how AI can revolutionize healthcare in resource-limited environments.

The event was a success, sparking engaging discussions and inspiring further exploration into the practical applications of AI in medical imaging.