New AI-Assisted Ultra-Low-Dose CT Scans Detect Pneumonia
A NEW study has demonstrated that artificial intelligence-enhanced ultra-low-dose CT (ULDCT) scans can accurately detect pneumonia in immunocompromised individuals while significantly reducing radiation exposure.
The prospective study, conducted between September 2020 and December 2022, included 54 immunocompromised adults (median age: 62 years), 63% of whom were male. Each participant underwent both a normal-dose CT scan and a ULDCT scan, with the latter using a reduced radiation dose and an AI algorithm to minimise image noise. Two radiologists, blinded to clinical information, analysed the images from all three methods—normal-dose CT, raw ULDCT, and AI-enhanced ULDCT—to evaluate pneumonia and associated lung abnormalities.
Results showed that the median radiation dose of ULDCT was only 1.95% of that of normal-dose CT (0.12 mSv vs. 6.15 mSv), substantially lowering patient exposure. The AI-enhanced ULDCT images demonstrated improved accuracy over raw ULDCT, correctly identifying pneumonia in all cases with an accuracy rate of 100%, compared to 96–98% for raw ULDCT. Notably, AI-assisted ULDCT was also more precise in detecting features associated with invasive fungal pneumonia.
Fine structural details, such as the tree-in-bud pattern (a sign of airway disease), were better visualised with denoised ULDCT, achieving an accuracy rate of 93% compared to 78–80% for raw ULDCT. Similarly, interlobular septal thickening was more accurately identified (78–83% vs. 61–67%), while intralobular septal thickening detection significantly improved from 0% to 85–87%.
The findings underscore the potential of AI-enhanced ULDCT as a safer alternative for immunocompromised individuals, who often require frequent imaging. By delivering accurate pneumonia detection while dramatically reducing radiation exposure, this method represents a promising advancement in pulmonary imaging.
Researchers suggest that AI-assisted ULDCT could be integrated into routine clinical practice to improve patient safety without compromising diagnostic accuracy. Further studies are recommended to assess its application in broader patient populations.
This breakthrough highlights the growing role of artificial intelligence in medical imaging, offering a balance between diagnostic precision and radiation safety.
Victoria Antoniou, EMJ
Reference
Klug M et al. Denoised ultra-low-dose chest CT to assess pneumonia in individuals who are immunocompromised. Radiology. 2025;DOI:10.1148/ryct.240189.
Each article is made available under the terms of the Creative Commons Attribution-Non Commercial 4.0 License.
canisgallicus 