Clinical value of deep learning image reconstruction on the diagnosis of pulmonary nodule for ultra-low-dose chest CT imaging

Denoised Ultra-Low-Dose Chest CT to Assess Pneumonia in Individuals Who Are Immunocompromised

Purpose To evaluate the accuracy of chest ultra-low-dose CT (ULDCT) as compared with normal-dose CT in the evaluation of pneumonia in individuals who are immunocompromised. Materials and Methods This prospective study included 54 adults who were immunocompromised (median age, 62 years [IQR, 47.75-69.25 years]; 34 [63%] male participants) referred for a chest CT scan between September 2020 and December 2022 to evaluate for pneumonia. Each participant underwent two scans: normal-dose CT (120 kVp and automatic current modulation) and ULDCT (100 kVp and constant current of 10 mA). ULDCT images underwent a postprocessing procedure using an artificial intelligence algorithm to reduce image noise. Two radiologists, blinded to all clinical information, examined the images obtained from the three methods (normal-dose CT, ULDCT, and denoised ULDCT) for the presence of pneumonia and associated findings. The normal-dose CT was used as the reference standard, and sensitivity, specificity, positive and negative predictive values, and accuracy were calculated. Results The median effective radiation dose of ULDCT scans (0.12 mSV) was 1.95% of that of the normal-dose CT (6.15 mSV). Ten of the 54 participants were correctly identified as having no pneumonia, with similar accuracy between denoised ULDCT and ULDCT (100% vs 96%-98%, respectively). Both methods allowed for detection of pneumonia and features associated with invasive fungal pneumonia, but accuracy was slightly better with denoised ULDCT (accuracy, 100% vs 91%-98%). Fine details were better visualized in denoised ULDCT images: tree-in-bud pattern (accuracy, 93% vs 78%-80%), interlobular septal thickening (accuracy, 78%-83% vs 61%-67%), and intralobular septal thickening (accuracy, 85%-87% vs 0%). Conclusion Denoised ULDCT imaging showed better accuracy than ULDCT in identifying lungs with or without pneumonia in individuals who were immunocompromised. Keywords: CT, Pulmonary, Lung, Infection, Technology Assessment Supplemental material is available for this article. © RSNA, 2025.

Detection, measurement, and diagnosis of lung nodules by ultra-low-dose CT in lung cancer screening: a systematic review

Objective

There is a lack of recent meta-analyses and systematic reviews on the use of ultra-low-dose CT (ULDCT) for the detection, measurement, and diagnosis of lung nodules. This review aims to summarize the latest advances of ULDCT in these areas.

Methods

A systematic review of studies in PubMed and Web of Science was conducted, using search terms specific to ULDCT and lung nodules. The included studies were published in the last 5 years (January 2019-August 2024). Two reviewers independently selected articles, extracted data, and assessed the risk of bias and concerns using the Quality Assessment of Diagnostic Accuracy Studies-II (QUADAS-II) tool. The standard-dose, low-dose, or contrast-enhanced CT served as the reference-standard CT to evaluate ULDCT.

Results

The literature search yielded 15 high-quality articles on a total of 1889 patients, of which 10, 3, and 2 dealt with the detection, measurement, and diagnosis of lung nodules. QUADAS-II showed a generally low risk of bias. The mean radiation dose for ULDCT was 0.22 ± 0.10 mSv (7.7%) against 2.84 ± 1.80 mSv for reference-standard CT. Nodule detection rates ranged from 86.1% to 100%. The variability of diameter measurements ranged from 2.1% to 14.4% against contrast-enhanced CT and from 3.1% to 8.29% against standard CT. The diagnosis rate of malignant nodules ranged from 75% to 91%.

Conclusions

ULDCT proves effective in detecting lung nodules while substantially reducing radiation exposure. However, the use of ULDCT for the measurement and diagnosis of lung nodules remains challenging and requires further research.

Advances in knowledge

When ULDCT reduces radiation exposure to 7.7%, it detects lung nodules at a rate of 86.1%-100%, with a measurement variance of 2.1%-14.4% and a diagnostic accuracy for malignancy of 75%-91%, suggesting the potential for safe and effective lung cancer screening.