The objective of this study was to investigate the effect of ASiR-V algorithm of different strength levels on the detection and characterization of pulmonary nodules in ULDCT at different dose levels. To further isolate the impact of tube voltage (kV), similar doses were generated at tube voltages of 70 kV and 100 kV through tube current modulation. Under the conditions of 0.51-0.53mSv in ULDCT, the combination of 70 kV and 70%ASiR-V generated the highest detection rate of 72.5% for GGNs, while the combination of 100 kV and high ASiR-V strength levels was better in preserving the forms of nodules.

In our study, the average effective doses of ULDCT were 0.105mSv, 0.33mSv, 0.52mSv, and the effective dose of the reference low dose CT was 1.28mSv. We found that the dose of ULDCT was reduced by 91.79%, 74.21%, 59.38%, respectively, but the sensitivity of nodule detection was only reduced by 27.08%, 9.89%, and 5.73%, respectively. Although the sensitivity for nodule detection was decreased, the majority (78%) of the undetected nodules had size less than 5 mm. It is well known that small nodules smaller than 5 mm have a very low risk of developing malignant tumors (less than 1%) [1, 21,22,23].

The overall DR of our ULDCT was 68.68%, with a maximum of 72.5% for GGNs (162 of which were not detected with a diameter of 5 mm). Botelho [24] et al. suggested that the minimum radiation dose to meet the diagnostic requirements for patients with a diameter of 5 mm should be 0.238 mSv when using fixed tube currents. Our study showed that PNs of 5 mm with an attenuation value of 100HU could be detected 100% (20/20) at 0.105mSv (Dose-1), but the detection ability of GGNs was limited at Dose-1 regardless of whether 70 kV or 100 kV was used. However, for nodules larger than 5 mm in Dose-1, CT attenuation values of 100HU and − 630HU could be detected at high detection rates (100HU: 70 kV:59/60, 100 kV:60/60; -630HU: 70 kV:58/60, 100 kV:60/60); and nodules with CT attenuation value of -800HU were mostly undetectable regardless of their sizes. At doses above 0.33mSv (Does-2 and Dose-3), there was essentially no change in the detection of nodules larger than 5 mm with CT attenuation values of 100HU and − 630 HU (100HU: 70 kV:119/120, 100 kV:119/120; -630HU: 70 kV:118/120; 100 kV:120/120). At the same times, the detection of 5 mm nodules with attenuation value of -800HU remained poor, but the detection for sizes larger than 5 mm increased significantly (70 kV:43/120; 100 kV:43/120). Considering that the minimum acceptable sensitivity of the screening test is 80% [25], ULDCT is not recommended for screening GGNs with CT attenuation of -800 HU or lower, and a higher radiation dose is recommended.

With regard to image quality, we found that when the dose was reduced, the image noise increased, the edge of the nodule was irregular, and the measurement error was prone to occur. The lower the radiation dose, the more serious the error and the larger the deformation index, and there was a significant difference in DC and SP between the 70 kV and 100 kV groups (p < 0.05). The effect of different dose levels on GGNs was stronger than SNs. At 70 kV, DC and SP decreased gradually with the increase of reconstruction strength. We found that the low kV and iterative reconstruction algorithm at high strength levels had the greatest effect in reducing DC and SP on the nodules with low CT attenuation values. Other studies have demonstrated that in ULDCT the use of iterative reconstruction algorithms, such as ASiR-V, and deep learning-based reconstruction algorithms could significantly reduce image noise and improve image quality [26,27,28]. The influence of different iteration strength levels on the image is mainly reflected in image quality, resolution, and noise level. Sui et al. [29] showed that there was no effect on the size measurement of nodules at low and ultra-low doses. However, our study found that there were deviations in the size measurement of nodules when combined with different ASiR-V levels. In this study, three representative strength levels of low, medium, high (30%, 50%, 70%) were used. however, we found that some features of the nodules could be distorted when iterative reconstruction algorithms with low strength levels were used, resulting in errors in diagnosis. Therefore, based on our results we recommend using 50% and 70% ASiR-V for image reconstruction to better preserve the nodule characteristics in ULDCT.

This study has several limitations: First, the sample size was small, only three kinds of CT attenuation value nodules were analyzed; Second, AI software from only one commercial company was used to obtain information. After the commercial AI software was obtained, the data training was not re-conducted. As far as we know, the general commercial AI software rarely performed ULDCT training, so the obtained information may not be 100% consistent with the training data, resulting in certain deviations in data analysis (for example, the nodules of -800HU failed to be detected, resulting in a low detection results). It is suggested that multiple software should be used for verification in the future. Third, this study was carried out on a phantom of the lung, which should be extended to real patient image analysis in the future.

In conclusion, the use of ULDCT combined with ASiR-V provides acceptable image quality at greatly reduced radiation exposure to patients, and there are no significant differences in the detection of nodules between 70 kV and 100 kV. At the same time, it is not recommended to choose too low dose conditions for finding GGNs. We recommend that dose levels above 0.33mSv be considered for screening, to ensure nodule detection and characteristic assessment. For patients with small nodules, 100 kV combined with higher ASiR-V strength levels (more than 50%) should be used to follow up the changes of nodules.