The effect of deep learning reconstruction on abdominal CT densitometry and image quality: a systematic review and meta-analysis

Objective To determine the difference in CT values and image quality of abdominal CT images reconstructed by filtered back-projection (FBP), hybrid iterative reconstruction (IR), and deep learning reconstruction (DLR). Methods PubMed and Embase were systematically searched for articles regarding CT densitometry in the abdomen and the image reconstruction techniques FBP, hybrid IR, and DLR. Mean differences in CT values between reconstruction techniques were analyzed. A comparison between signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of FBP, hybrid IR, and DLR was made. A comparison of diagnostic confidence between hybrid IR and DLR was made. Results Sixteen articles were included, six being suitable for meta-analysis. In the liver, the mean difference between hybrid IR and DLR was − 0.633 HU (p = 0.483, SD ± 0.902 HU). In the spleen, the mean difference between hybrid IR and DLR was − 0.099 HU (p = 0.925, SD ± 1.061 HU). In the pancreas, the mean difference between hybrid IR and DLR was − 1.372 HU (p = 0.353, SD ± 1.476 HU). In 14 articles, CNR was described. In all cases, DLR showed a significantly higher CNR. In 9 articles, SNR was described. In all cases but one, DLR showed a significantly higher SNR. In all cases, DLR showed a significantly higher diagnostic confidence. Conclusions There were no significant differences in CT values reconstructed by FBP, hybrid IR, and DLR in abdominal organs. This shows that these reconstruction techniques are consistent in reconstructing CT values. DLR images showed a significantly higher SNR and CNR, compared to FBP and hybrid IR. Key Points CT values of abdominal CT images are similar between deep learning reconstruction (DLR), filtered back-projection (FBP), and hybrid iterative reconstruction (IR). DLR results in improved image quality in terms of SNR and CNR compared to FBP and hybrid IR images. DLR can thus be safely implemented in the clinical setting resulting in improved image quality without affecting CT values.

Multi-Slice Spiral Computed Tomography Image Features under Hybrid Iterative Reconstruction Algorithm in Staging Diagnosis of Bladder Cancer

Objective. This study was aimed to explore the accuracy of multi-slice spiral computed tomography (CT) scan in preoperative staging diagnosis of bladder cancer based on hybrid iterative reconstruction algorithm, so as to provide a more reasonable supporting basis for guiding clinical work in the future. Methods. Retrospectively, 120 patients admitted to hospital from July 2019 to April 2021, who were confirmed to be with urothelial carcinoma of the bladder by pathological examination after surgical treatment, were selected. CT images before processing were set as the control group and those after processing were set as the observation group according to whether they were processed by the hybrid iterative algorithm. Postoperative pathological examination was utilized as the standard for analysis. The accuracy and consistency of the two methods were compared. Results. The accuracy of the results of each stage of the observation group (T1 stage: 91.09%, T2 stage: 89.66%, T3 stage: 88.89%, and T4 stage: 88.89%) and consistency (T1 stage: 0.66, T2 stage: 0.69, T3 stage: 0.71, and T4 stage: 0.82) were higher than those of the control group (accuracy: T1—57.01%, T2—48.28%, T3—44.44%, and T4—44.44%). The consistency was as follows: T1—0.32, T2—0.24, T3—0.37, and T4—0.43, and the comparison was statistically significant ( P < 0.05). Conclusion. The adoption value of the image features based on the hybrid iterative reconstruction algorithm in the diagnosis of bladder cancer staging was higher than that of the conventional multi-slice spiral CT, indicating that the hybrid iterative reconstruction algorithm had a good adoption prospect in clinical examination.

Diagnostic performance and image quality of low-tube voltage and low-contrast medium dose protocol with hybrid iterative reconstruction for hepatic dynamic CT

Objective: To evaluate the diagnostic performance and image quality of the low-tube voltage and low-contrast medium dose protocol for hepatic dynamic CT. Methods: This retrospective study was conducted between January and May 2018. All patients underwent hepatic dynamic CT using one of the two protocols: tube voltage, 80 kVp and contrast dose, 370 mgI/kg with hybrid iterative reconstruction or tube voltage, 120 kVp and contrast dose, 600 mgI/kg with filtered back projection. Two radiologists independently scored lesion conspicuity and image quality. Another radiologist measured the CT numbers of abdominal organs, muscles, and hepatocellular carcinoma (HCC) in each phase. Lesion detectability, HCC diagnostic ability, and image quality of the arterial phase were compared between the two protocols using the non-inferiority test. CT numbers and HCC-to-liver contrast were compared between the protocols using the Mann–Whitney U test. Results: 424 patients (70.5 ± 10.1 years) were evaluated. The 80-kVp protocol showed non-inferiority in lesion detectability and diagnostic ability for HCC (sensitivity, 85.7–89.3%; specificity, 96.3–98.6%) compared with the 120-kVp protocol (sensitivity, 91.0–93.3%; specificity, 93.6–97.3%) (p < 0.001–0.038). The ratio of fair image quality in the 80-kVp protocol also showed non-inferiority compared with that in the 120-kVp protocol in assessments by both readers (p < 0.001). HCC-to-liver contrast showed no significant differences for all phases (p = 0.309–0.705) between the two protocols. Conclusion: The 80-kVp protocol with hybrid iterative reconstruction for hepatic dynamic CT can decrease iodine doses while maintaining diagnostic performance and image quality compared with the 120-kVp protocol. Advances in knowledge: The 80- and 120-kVp protocols showed equivalent hepatic lesion detectability, diagnostic ability for HCC, image quality, and HCC-to-liver contrast. The 80-kVp protocol showed a 38.3% reduction in iodine dose compared with the 120-kVp protocol.

Multislice Spiral Computed Tomography under Hybrid Iterative Reconstruction Algorithm in Staging Diagnosis of Bladder Cancer and Ceramide Glycosylation

This study was to discuss the application of multislice spiral computed tomography (CT) in the staging diagnosis of bladder cancer and the effect of ceramide glycosylation. The hybrid iterative reconstruction algorithm was applied. Immunohistochemistry and western blot were used to detect the normal bladder tissues (30 cases) of GCS in group 1 (100 cases) and group 2. The scanned images of all the research objects were obtained, the images with the iterative reconstruction algorithm were reconstructed, and statistical analysis on the CT value under the algorithm was conducted. The results showed that the image quality, blood vessel sharpness, average image score, signal-to-noise ratio, and radiation dose after the spiral CT and iterative reconstruction algorithm all increased, while the noise value decreased. The optical density value of glucosylceramide synthase in group 2 patients increased by 71%, and the optical density value of group 1 increased by 29%. The optical density expression of glucosylceramide synthase in group 1 patients was significantly higher than that in the control group, and there was a statistical difference between the two ( P < 0.05 ). Among the results of multislice spiral CT for tumor staging, the lesions larger than 5 cm and in the range of 1.1–2 cm in diameter were more sensitive. In 41 patients, there were multiple lesions. A total of 142 cancer lesions were found. The diameter of the tissue ranged from 0.5 to 6.8 cm, with an average diameter of 2.03 ± 0.35 cm. The optical density of glucosylceramide synthase in the group 1 was 5526, and the optical density in group 2 was 2576. The OD expression of GCS in group 1 was greatly higher in contrast to that in group 2, and there was a statistical difference between the two groups ( P < 0.05 ). The multislice spiral CT examination under this algorithm found that the diagnosis and staging accuracy of lesions with a diameter greater than 5 cm and tumor diameters in the range of 1.1 to 2 cm was higher. The image processed by the hybrid iterative reconstruction algorithm had good effect, high definition, and accuracy.

Low-dose whole-body CT using deep learning image reconstruction: image quality and lesion detection

Objectives: To evaluate image quality and lesion detection capabilities of low-dose (LD) portal venous phase whole-body computed tomography (CT) using deep learning image reconstruction (DLIR). Methods: The study cohort of 59 consecutive patients (mean age, 67.2 years) who underwent whole-body LD CT and a prior standard-dose (SD) CT reconstructed with hybrid iterative reconstruction (SD-IR) within one year for surveillance of malignancy were assessed. The LD CT images were reconstructed with hybrid iterative reconstruction of 40% (LD-IR) and DLIR (LD-DLIR). The radiologists independently evaluated image quality (5-point scale) and lesion detection. Attenuation values in Hounsfield units (HU) of the liver, pancreas, spleen, abdominal aorta, and portal vein; the background noise and signal-to-noise ratio (SNR) of the liver, pancreas, and spleen were calculated. Qualitative and quantitative parameters were compared between the SD-IR, LD-IR, and LD-DLIR images. The CT dose-index volumes (CTDIvol) and dose-length product (DLP) were compared between SD and LD scans. Results: The image quality and lesion detection rate of the LD-DLIR was comparable to the SD-IR. The image quality was significantly better in SD-IR than in LD-IR (p < 0.017). The attenuation values of all anatomical structures were comparable between the SD-IR and LD-DLIR (p = 0.28–0.96). However, background noise was significantly lower in the LD-DLIR (p < 0.001) and resulted in improved SNRs (p < 0.001) compared to the SD-IR and LD-IR images. The mean CTDIvol and DLP were significantly lower in the LD (2.9 mGy and 216.2 mGy•cm) than in the SD (13.5 mGy and 1011.6 mGy•cm) (p < 0.0001). Conclusion: LD CT images reconstructed with DLIR enable radiation dose reduction of >75% while maintaining image quality and lesion detection rate and superior SNR in comparison to SD-IR. Advances in knowledge: Deep learning image reconstruction algorithm enables around 80% reduction in radiation dose while maintaining the image quality and lesion detection compared to standard-dose whole-body CT.

Model-based reconstruction algorithm in the detection of acute trauma-related lesions in brain CT examinations

Background and purpose To evaluate the added value of a model-based reconstruction algorithm in the assessment of acute traumatic brain lesions in emergency non-enhanced computed tomography, in comparison with a standard hybrid iterative reconstruction approach. Materials and methods We retrospectively evaluated a total of 350 patients who underwent a 256-row non-enhanced computed tomography scan at the emergency department for brain trauma. Images were reconstructed both with hybrid and model-based iterative algorithm. Two radiologists, blinded to clinical data, recorded the presence, nature, number, and location of acute findings. Subjective image quality was performed using a 4-point scale. Objective image quality was determined by computing the signal-to-noise ratio and contrast-to-noise ratio. The agreement between the two readers was evaluated using k-statistics. Results A subjective image quality analysis using model-based iterative reconstruction gave a higher detection rate of acute trauma-related lesions in comparison to hybrid iterative reconstruction (extradural haematomas 116 vs. 68, subdural haemorrhages 162 vs. 98, subarachnoid haemorrhages 118 vs. 78, parenchymal haemorrhages 94 vs. 64, contusive lesions 36 vs. 28, diffuse axonal injuries 75 vs. 31; all P<0.001). Inter-observer agreement was moderate to excellent in evaluating all injuries (extradural haematomas k=0.79, subdural haemorrhages k=0.82, subarachnoid haemorrhages k=0.91, parenchymal haemorrhages k=0.98, contusive lesions k=0.88, diffuse axonal injuries k=0.70). Quantitatively, the mean standard deviation of the thalamus on model-based iterative reconstruction images was lower in comparison to hybrid iterative one (2.12 ± 0.92 vsa 3.52 ± 1.10; P=0.030) while the contrast-to-noise ratio and signal-to-noise ratio were significantly higher (contrast-to-noise ratio 3.06 ± 0.55 vs. 1.55 ± 0.68, signal-to-noise ratio 14.51 ± 1.78 vs. 8.62 ± 1.88; P<0.0001). Median subjective image quality values for model-based iterative reconstruction were significantly higher ( P=0.003). Conclusion Model-based iterative reconstruction, offering a higher image quality at a thinner slice, allowed the identification of a higher number of acute traumatic lesions than hybrid iterative reconstruction, with a significant reduction of noise.

Deep learning–based reconstruction may improve non-contrast cerebral CT imaging compared to other current reconstruction algorithms

Objectives To evaluate image quality and reconstruction times of a commercial deep learning reconstruction algorithm (DLR) compared to hybrid-iterative reconstruction (Hybrid-IR) and model-based iterative reconstruction (MBIR) algorithms for cerebral non-contrast CT (NCCT). Methods Cerebral NCCT acquisitions of 50 consecutive patients were reconstructed using DLR, Hybrid-IR and MBIR with a clinical CT system. Image quality, in terms of six subjective characteristics (noise, sharpness, grey-white matter differentiation, artefacts, natural appearance and overall image quality), was scored by five observers. As objective metrics of image quality, the noise magnitude and signal-difference-to-noise ratio (SDNR) of the grey and white matter were calculated. Mean values for the image quality characteristics scored by the observers were estimated using a general linear model to account for multiple readers. The estimated means for the reconstruction methods were pairwise compared. Calculated measures were compared using paired t tests. Results For all image quality characteristics, DLR images were scored significantly higher than MBIR images. Compared to Hybrid-IR, perceived noise and grey-white matter differentiation were better with DLR, while no difference was detected for other image quality characteristics. Noise magnitude was lower for DLR compared to Hybrid-IR and MBIR (5.6, 6.4 and 6.2, respectively) and SDNR higher (2.4, 1.9 and 2.0, respectively). Reconstruction times were 27 s, 44 s and 176 s for Hybrid-IR, DLR and MBIR respectively. Conclusions With a slight increase in reconstruction time, DLR results in lower noise and improved tissue differentiation compared to Hybrid-IR. Image quality of MBIR is significantly lower compared to DLR with much longer reconstruction times. Key Points • Deep learning reconstruction of cerebral non-contrast CT results in lower noise and improved tissue differentiation compared to hybrid-iterative reconstruction. • Deep learning reconstruction of cerebral non-contrast CT results in better image quality in all aspects evaluated compared to model-based iterative reconstruction. • Deep learning reconstruction only needs a slight increase in reconstruction time compared to hybrid-iterative reconstruction, while model-based iterative reconstruction requires considerably longer processing time.