Lung diseases like COVID-19, pneumonia, and other respiratory conditions are normally not desired, especially in low-resource settings. Traditional methods like CT scans and X-rays are injurious to the patient because they expose them to radiation, and the equipment may not be readily available. Lung ultrasound (LUS) is less expensive, transportable, and safer, but its imaging takes specialized skills.
To combat this, scientists have developed a powerful AI-based system that can read lung ultrasound videos with high precision. The system utilizes various deep learning techniques to improve video quality, maintain the order of the ultrasound frames, and extract the most important features for disease diagnosis. The hybrid 3D deep learning model used in this study can identify healthy lungs, pneumonia, and COVID-19 with an astounding accuracy of 96.57%.
Among the most significant innovations is the explainability feature of the system—it drops visual cues on what parts of the ultrasound image contributed to its decision, allowing for higher transparency and credibility with medical professionals. In automating and enhancing LUS interpretation, this AI technology has the potential to assist physicians in faster and more accurate diagnoses, reducing reliance on expensive imaging equipment and improving patient care worldwide.
This innovation has the potential to transform the diagnosis of respiratory disease, and specifically AI-based lung ultrasound, specifically in hospitals and in the field.
Full text: Arefin Ittesafun Abian, Mohaimenul Azam Khan Raiaan, Asif Karim, Sami Azam, Nur Mohammad Fahad, Niusha Shafiabady, Kheng Cher Yeo, Friso De Boer, Automated diagnosis of respiratory diseases from lung ultrasound videos ensuring XAI: an innovative hybrid model approach, Frontiers in Computer Science, 6, https://doi.org/10.3389/fcomp.2024.1438126