Ultra-low-dose CT with 1024 matrix and deep learning reconstruction reduces radiation dose by approximately 82% with noninferior image quality: a non-randomized controlled trial

To address cumulative radiation exposure from repeated scans during CT-guided percutaneous lung biopsy, this study evaluated the impact of low-dose and ultra-low-dose CT combined with a 1024×1024 matrix and deep learning reconstruction algorithms on image quality and radiation dose, investigating the clinical feasibility of this protocol in meeting localization requirements for puncture procedures.

In this prospective non-randomized controlled trial, 100 consecutive patients undergoing CT-guided thoracic biopsy were enrolled at the Department of Radiology, First Affiliated Hospital of Dalian Medical University, between April and December 2025. Multiple repeated localizations were performed during the procedure. The standard-dose (CD) protocol utilized 120 kV, 180 mA, a 512×512 reconstruction matrix, and ClearView reconstruction at 60% strength (CD-CV60 group). Low-dose (LD) protocols employed 120 kV with automatic tube current modulation (O-Dose coefficient 1.0), a 1024×1024 matrix, and ClearInfinity reconstruction at 40%, 60%, and 80% strength (LD-CI40, LD-CI60, LD-CI80 groups). Ultra-low-dose (ULD) protocols used 120 kV with automatic tube current modulation (O-Dose coefficient 0.7), a 1024×1024 matrix, and ClearInfinity reconstruction at 40%, 60%, and 80% strength (ULD-CI40, ULD-CI60, ULD-CI80 groups); Except for the tube-current setting and reconstruction parameters, all other scanning parameters remained identical across protocols. CT attenuation values and standard deviation (SD) of the aorta and chest wall fat were measured to calculate signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR). Subjective image quality, including lesion margin delineation, homogeneity of lesion density, and artifact severity, was assessed using a 5-point Likert scale. Volume CT dose index (CTDI_vol), dose-length product (DLP), effective dose (ED), and size-specific dose estimate (SSDE) were recorded.

Interobserver agreement for lesion-margin delineation, lesion-density homogeneity, and artifact severity was good, with linearly weighted κ values of 0.801 (95%CI: 0.753 to 0.844), 0.758 (95%CI: 0.709 to 0.805), and 0.815 (95%CI: 0.768 to 0.859), respectively. Mean CTDIvol for CD, LD, and ULD protocols were 11.13 mGy, 3.98 mGy, and 2.01 mGy, respectively; the corresponding DLP values were 389.55, 139.13, and 70.35 mGy/cm; the ED values were 5.45, 1.95, and 0.98 mSv; and corresponding SSDE values were 14.37 mGy, 5.27 mGy, and 2.60 mGy. Compared to the CD protocol, CTDI_vol was reduced by 64.2% and 81.9% in the LD and ULD protocols, respectively, DLP was reduced by 64.2% and 81.9%, ED by 64.2% and 82.0%, and SSDE by 63.3% and 81.9%, respectively. Regarding objective quality, background SD progressively decreased with increasing ClearInfinity strength in LD and ULD protocols, whereas SNR and CNR generally increased. SNR and CNR in LD-CI60, LD-CI80, ULD-CI60, and ULD-CI80 groups were significantly higher than those in the CD-CV60 group (P<0.05). For subjective quality, most metrics initially improved then declined as ClearInfinity strength increased. Among LD and ULD protocols, images reconstructed with 60% ClearInfinity demonstrated superior subjective scores; notably, lesion density homogeneity in the LD-CI60 group was significantly higher than in CD-CV60 (P<0.05), while other 60% reconstruction groups showed no statistically significant differences compared to CD-CV60. These 60% reconstruction images consistently outperformed their respective 40% and 80% counterparts within the same dose protocol (P<0.05). The mean difference in patient-level composite image-quality scores between the ULD-CI60 and CD-CV60 groups (ULD-CI60-CD-CV60) was −0.002 points (95%CI:−0.100 to 0.097), and the lower limit of the 95%CI was higher than the prespecified noninferiority margin of -0.5 points. Postoperative complications occurred in 33 patients (33.0%), comprising isolated pneumothorax (14 cases, 14.0%), isolated pulmonary hemorrhage (12 cases, 12.0%), and combined pneumothorax with hemorrhage (7 cases, 7.0%); no severe complications such as air embolism occurred.

During CT-guided percutaneous lung biopsy, the ULD-CI60 protocol achieves image quality sufficient for puncture localization and path planning while reducing radiation dose by approximately 81.9%.