A total of 249 patients (126 male, 123 female) with a mean age of 50.0 ± 12.9 years (range, 19–78 years) were included. The mean body mass index (BMI) was 22.9 ± 3.0 kg/m2 (range, 12.5–32.6 kg/m2), with 16 patients with BMI < 18.5 kg/m2 and 91 patients with BMI ≥ 24 kg/m2. Patient characteristics are summarized in Table 2.

According to the reference standard, a total of 637 lung nodules were identified. Among them, 368 (57.8%) were solid, 35 (5.5%) were part-solid, and 234 (36.7%) were pure ground-glass. The nodules were divided according to the type (solid, part-solid, and pure ground-glass) and Lung-RADS category for subgroup analysis.

The mean CTDIvol (0.5 mGy × cm ± 0.0 vs. 5.6 ± 1.2 mGy × cm; p < 0.001) and dose-length product (23.0 mGy ×cm ± 1.4 vs. 258.4 mGy × cm ± 54.4; p < 0.001) were significantly lower with ULDCT than with RDCT. The mean effective dose of RDCT and ULDCT was 3.6 mSv ± 0.8 and 0.3 mSv ± 0.0, respectively, indicating a significant dose reduction of 91.2% (p < .001).

Figure 2 shows box plots of the qualitative scoring results. For ULDCT images, all image quality parameters were ranked superior with EDM when compared with HIR and VMIs (p < 0.05 for all comparisons). Notably, the average scores of ULDCT EDM in each assessment item were higher or equal to the average (score 3 on a 5-point scale). The inter-observer agreement (kappa value) for streak artifact, image noise, clarity of small vessels, visibility of nodules, and the overall image quality were 0.75, 0.70, 0.71, 0.74, and 0.68, respectively.

The box plots show the qualitative image quality scoring results. For visibility of pulmonary nodules, no statistically significant difference is found between RDCT HIR and ULDCT EDM (p > 0.05). And significant differences are observed among other reconstruction groups (p  < 0.001 for all). DLSCT, dual-layer detector spectral CT; RDCT, regular-dose CT; ULDCT, ultra-low dose CT; HIR, hybrid iterative reconstruction; VMI, virtual monoenergetic image; EDM, electron density mapping

In detail, overall image quality for ULDCT EDM was rated as excellent in 15.5% (99 of 637), good in 48.8% (311 of 637), and moderate in 34.2% (218 of 637). In terms of the visibility of all pulmonary nodules, the difference was not significant between RDCT HIR and ULDCT EDM (4.4 ± 0.8 vs. 4.3 ± 0.8; p > 0.05). For pure GGN, ULDCT EDM scored significantly better than RDCT HIR (4.0 ± 0.8 vs. 3.6 ± 0.7; p < 0.001). For part-solid nodules, a similar result was found (ULDCT EDM 4.8 ± 0.4 vs. RDCT HIR 4.7 ± 0.6, p > 0.05). For solid nodules, RDCT HIR achieved the highest score (4.5 ± 0.7), and EDM and VMI40 keV on ULDCT were of diagnostic quality (score > 3). Examples of RDCT and ULDCT images in participants were provided in Figs. 3–5. Detailed qualitative analysis results were provided in the Supplementary Materials (Appendix E2).

DLSCT at different radiation doses and reconstruction algorithms in a 65-year-old female (BMI: 21.2 kg/m2) with a solid nodule in the right lower lobe (circle). RDCT, regular-dose CT; ULDCT, ultra-low dose CT; HIR, hybrid iterative reconstruction; VMI, virtual monoenergetic image; EDM, electron density mapping

DLSCT at different radiation doses and reconstruction algorithms in a 49-year-old female (BMI: 19.5 kg/m2) with a part-solid nodule in the left lower lobe (circle). RDCT, regular-dose CT; ULDCT, ultra-low dose CT; HIR, hybrid iterative reconstruction; VMI, virtual monoenergetic image; EDM, electron density mapping

DLSCT at different radiation doses and reconstruction algorithms in a 67-year-old male (BMI: 25.4 kg/m2) with a pure ground-glass nodule in the right upper lobe (circle). RDCT, regular-dose CT; ULDCT, ultra-low dose CT; HIR, hybrid iterative reconstruction; VMI, virtual monoenergetic image; EDM, electron density mapping

Quantitative assessment results are summarized in Table 3. There were significant differences in CT numbers among different groups (all p < 0.001). ULDCT EDM revealed significantly lower noise in fat, aorta, and muscle than other reconstructions (p < 0.001 for all comparisons). Compared with RDCT HIR, ULDCT EDM showed significantly higher SNR (44.0 ± 77.2 vs. 4.6 ± 6.6; p < 0.001) and CNR (26.7 ± 17.7 vs. 5.0 ± 4.4; p < 0.001) for pulmonary nodules. Furthermore, VMIs at 40 keV and 70 keV of ULDCT yielded lower noise and higher SNR than ULDCT HIR, although the differences were not statistically significant (p > 0.05 for all).

Nodule detection rates were given in Table 4, categorized by nodule type, reconstructions, and Lung-RADS category. Compared with the reference standard, reconstruction techniques of ULDCT images performed good detection of solid nodules, with a similar nodule detection rate (78.3%–82.6%). Among all reconstructions, only 1 of 35 part-solid nodules, sizing 10 mm, was missed on VMI70keV. For pure GGNs, EDM showed a significantly higher nodule detection rate (85.5%, 200/234), compared with HIR (76.9%, 180/234), VMI40keV (76.5%, 179/234), and VMI70keV (71.8%, 168/234). On EDM, 34 pure GGNs not identified, and all of them were less-than 6 mm in diameter.

The results of inter-scan agreement and inter-observer agreement were summarized in Table 5. Inter-scan agreements between ULDCT and SDCT were good (all κw > 0.60, both reader 1 and reader 2). The inter-observer agreements between the two readers were good for ULDCT HIR and ULDCT VMI70keV, and excellent for ULDCT EDM and ULDCT VMI40keV.

The consistency of measurement between RDCT and ULDCT was excellent for diameter ((ICC value) RDCT vs. ULDCT HIR: 0.907, RDCT vs. ULDCT EDM: 0.913, RDCT vs. ULDCT VMI40keV: 0.901, RDCT vs. ULDCT VMI70keV: 0.895) and volume ((ICC value) RDCT vs. ULDCT HIR: 0.977, RDCT vs. ULDCT EDM: 0.989, RDCT vs. ULDCT VMI40keV: 0.981, RDCT vs. ULDCT VMI70keV: 0.975). Consistency in nodule diameter and volume measurements between the two readers was excellent (all ICC values > 0.80). Figure 6 shows the Bland-Altman plots of RDCT HIR vs. ULDCT EDM for diameter and volume measurements. Subgroup analysis based on nodule type and size is provided in the Supplementary Materials (Appendix E3).

Bland-Altman plots show the consistency of diameter measurement (a) and volume measurement (b) between RDCT HIR and ULDCT EDM. RDCT, regular-dose CT; ULDCT, ultra-low dose CT; HIR, hybrid iterative reconstruction; EDM, electron density mapping