0
0
January 25, 2025
A new study has shown that Artificial Intelligence (AI) could be the next big ally for radiologists, offering a breakthrough in the diagnosis of lung diseases like pneumonia, COVID-19, and other respiratory conditions. Researchers from Charles Darwin University (CDU), United International University, and Australian Catholic University (ACU) have successfully developed and trained an AI model capable of analyzing lung ultrasound videos with an impressive accuracy rate of 96.57%.
The study, recently published in the Journal of Frontiers in Computer Science, highlights the innovative potential of AI in medical imaging. The AI model operates by meticulously examining each frame of a lung ultrasound video to extract key features and evaluate the patterns of lung behavior over time. It then uses these insights to classify the ultrasound into specific diagnostic categories, such as normal, pneumonia, COVID-19, and more.
Associate Professor Niusha Shafiabady, an adjunct faculty member at CDU and co-author of the study, explained that the AI’s 96.57% accuracy rate was verified by medical professionals, enhancing its credibility. “The model also uses explainable AI techniques, allowing radiologists to understand why it made certain decisions,” Shafiabady said. “This increases the transparency of the results, making it easier for healthcare providers to trust and apply the technology.”
The model’s use of explainable AI means that it not only offers precise diagnoses but also provides visual cues, such as heatmaps, that show which areas of the lung the AI focused on to make its decision. This allows doctors to quickly identify critical regions and improve the accuracy of their diagnoses. According to the researchers, this approach also has the potential to improve clinical transparency and assist in the training of medical professionals.
Future possibilities for this model include its ability to detect a broader range of lung diseases. With further training and data, it could identify conditions such as tuberculosis, asthma, pulmonary fibrosis, black lung, chronic lung disease, and even lung cancer. In addition, the model could be adapted for other imaging techniques like CT scans and X-rays, opening up new avenues for research in medical AI.
The collaborative study was led by researchers from United International University in Bangladesh, with contributions from CDU’s Dr. Asif Karim, Dr. Sami Azam, Dr. Kheng Cher Yeo, Professor Friso De Boer, and Associate Professor Shafiabady, who also works at ACU.
Disclaimer: The AI model described in this article is still undergoing research and development. While the results are promising, it is not yet available for widespread clinical use. Medical professionals are advised to use their clinical judgment and consult with experts before relying on AI-based diagnostic tools.
