ABSTRACT
Objective To evaluate the impact of deep learning reconstruction algorithm on the image quality of head and neck CT angiography (CTA) at 100 kVp. Methods CT scanning was performed at 100 kVp for the 37 patients who underwent head and neck CTA in PUMC Hospital from March to April in 2021.Four sets of images were reconstructed by three-dimensional adaptive iterative dose reduction (AIDR 3D) and advanced intelligent Clear-IQ engine (AiCE) (low,medium,and high intensity algorithms),respectively.The average CT value,standard deviation (SD),signal-to-noise ratio (SNR),and contrast-to-noise ratio (CNR) of the region of interest in the transverse section image were calculated.Furthermore,the four sets of sagittal maximum intensity projection images of the anterior cerebral artery were scored (1 point:poor,5 points:excellent). Results The SNR and CNR showed differences in the images reconstructed by AiCE (low,medium,and high intensity) and AIDR 3D (all P<0.01).The quality scores of the image reconstructed by AiCE (low,medium,and high intensity) and AIDR 3D were 4.78±0.41,4.92±0.27,4.97±0.16,and 3.92±0.27,respectively,which showed statistically significant differences (all P<0.001). Conclusion AiCE outperformed AIDR 3D in reconstructing the images of head and neck CTA at 100 kVp,being capable of improving image quality and applicable in clinical examinations.
Subject(s)
Humans , Computed Tomography Angiography/methods , Radiation Dosage , Deep Learning , Radiographic Image Interpretation, Computer-Assisted/methods , Signal-To-Noise Ratio , AlgorithmsABSTRACT
Objective:To explore the application value of deep learning reconstruction (DLR) in low-dose brain CT imaging in children with craniocerebral trauma.Methods:The CT data of 51 children with craniocerebral trauma complicated with intracerebral hemorrhage who received low dose brain CT were retrospectively collected in Hunan Children′s Hospital between June 2020 and February 2021. All images were reconstructed at 1.25 mm and 5 mm slice thickness utilizing two reconstruction algorithms and divided into six subgroups: ASIR-V with three different blending ratios (0, 50%, 100%), and DLR with three different reconstruction strengths [low (L), media (M) and high (H)]. The objective parameters including CT value, signal to noise ratio (SNR) and contrast to noise ratio (CNR) of dorsal thalamus (grey matter), white matter of frontal lobe and hemorrhagic lesion, as well as basicranial artifact noise (SD) and background SD were measured and calculated. Subjective evaluation was performed with a 5-point scale scoring. Objective parameters and subjective scores were compared among different groups using randomized block analysis of variance and Friedman test, respectively. The objective and subjective differences between 1.25 mm DLR-H and ASIR-V50% images were analyzed using paired samples t-test and correlated sample rank sum test. Results:The average CT dose index volume, dose length product and size-specific dose estimate of head CT were 17.7 (11.9, 21.1) mGy, 248.4 (142.2, 338.1) mGy·cm and (15.7±2.8) mGy. With the same thickness, the difference of CT values between the DLR and ASIR-V groups were stastistically significant ( P<0.05). The subjective scores of DLR groups were significantly better than those of ASIR-V; the higher was the reconstruction grade of ASIR-V and DLR, the higher SNR and CNR values and the lower SD value were obtained for each structure (all P<0.05). DLR images showed better objective parameters than ASIR-V50% images. Background:SD was lowest on DLR-H and ASIR-V100% images, with no significant difference found between these two groups. Using 1.25 mm thickness, DLR-H images showed higher SNR (for both gray matter and white matter) and CNR than ASIR-V100% images ( P<0.05). The subjective score was decreased with the slice thickness reduced. However, the average subjective scores of 1.25 mm DLR images were all over 3 points, while those of 1.25 mm ASIR-V images were less than 3 points, which could not fully meet the needs of diagnosis. Images of 1.25 mm DLR-H had higher background SD and artifact SD than 5 mm ASIR-V50% images ( t=2.96, 2.83, P=0.005, 0.007), while the score and other objective parameters were not statistically different between these two groups ( P>0.05). Conclusion:In children′s low-dose cerebral CT, DLR can improve image quality, with the DLR-H images displaying the highest image quality. It can also increase the SNR and CNR of gray and white matter of images with thin thickness.
ABSTRACT
【Objective】 To explore the value of deep learning reconstruction algorithm (DLIR) in improving image quality of portal vein. 【Methods】 We retrospectively enrolled 32 patients who underwent double-phasic enhanced abdominal CT scanning. Images at the portal venous phase were reconstructed using the 50% adaptive statistical iterative reconstruction (ASIR-V), DLIR at medium (DLIR-M) and high strength (DLIR-H). The CT value and image noise (standard deviation) of the main portal vein, the right portal vein branch, the left portal vein branch, and the paravertebral muscle were measured, and the contrast-noise-ratio (CNR) for vessels were calculated. Moreover, the edge-rising-slope (ERS) of the main portal vein edge was measured to evaluate image spatial resolution. The overall image noise, image contrast, and portal vein branch display were evaluated using a 5-point grading scale and image artifacts using a 4-point grading scare by two experienced radiologists. In addition, we calculated the display rate of small branches of the portal vein in the three reconstruction algorithms. 【Results】 Image noise of the DLIR images in the main portal vein, right branch and left branch was significantly lower than that of ASIR-V 50% images, of which the DLIR-H images had the lowest noise and highest CNR. The ERS of the DLIR images in the main portal vein was significantly higher than that of the ASIR-V 50% images. For qualitative analyses, the DLIR images were significantly better than the ASIR-V 50% ones (P<0.01). In addition, the display rates of small branches of the portal vein in DLIR images were (DLIR-M: 93.75%; DLIR-H: 100%), significantly higher than that of ASIR-V 50% (68.75%). 【Conclusion】 Compared with ASIR-V 50% images, DLIR images can significantly reduce the image noise and improve the spatial resolution of the portal vein and the display rate of small branches of the portal vein.
ABSTRACT
Objective:To investigate the feasibility of chest ultra-low dose CT (ULDCT) using deep learning reconstruction (DLR) for lung cancer screening, and to compare its image quality and nodule detection rate with ULDCT iterative reconstruction (Hybrid IR) and conventional dose CT (RDCT) Hybrid IR.Methods:The patients who underwent chest CT examination for pulmonary nodules in Peking Union Medical College Hospital from October 2020 to March 2021 were prospectively included and underwent chest RDCT (120 kVp, automatic tube current), followed by ULDCT (100 kVp, 20 mA). The RDCT images were reconstructed with Hybrid IR (adaptive iterative dose reduction 3D,AIDR 3D), and ULDCT was reconstructed with AIDR3D and DLR. Radiation dose parameters and nodule numbers were recorded. Image quality was assessed using objective noise, signal-to-noise ratio (SNR) of the main trachea and left upper lobe, subjective image scores of the lung and nodules. Subjective scores were scored by 2 experienced radiologists on a Likert 5-point scale. The difference of radiation dose was compared with paired t-test between ULDCT and RDCT.The differences of quantitative indexes, objective image noise and subjective scores of the three reconstruction methods were compared with one-way analysis of variance or Friedman test. Results:Forty-five patients were enrolled, including 17 males and 28 females, aged from 32 to 74 (55±11) years. The radiation dose of ULDCT was (0.17±0.01) mSv, which was significantly lower than that of RDCT [(1.35±0.41) mSv, t=15.46, P<0.001]. There were significant differences in the image noise and SNR in the trachea and lung parenchyma and in the CT value of the trachea among ULDCT-AICE, ULDCT-AIDR 3D and RDCT-AIDR 3D images ( P<0.05). Image noise in the trachea and lung parenchyma and CT value in the trachea of ULDCT-AICE were significantly lower than those of ULDCT-AIDR 3D ( P<0.05) and comparable to RDCT-AIDR 3D ( P>0.05). There were significant differences in subjective image scores of the lung and nodules among ULDCT-AICE, ULDCT-AIDR 3D and RDCT-AIDR 3D images (χ2=50.57,117.20, P<0.001). Subjective image scores of the lung and nodules for ULDCT-AICE were significantly higher than those of ULDCT-AIDR 3D ( P<0.05), and non-inferior to RDCT-ADIR 3D ( P>0.05). All 72 clinically significant nodules detected on RDCT-ADIR 3D were also noted on ULDCT-AICE and ULDCT-AIDR 3D images. Conclusions:Chest ULDCT using DLR can significantly reduce the radiation dose, and compared with Hybrid IR, it can effectively reduce the image noise and improve SNR, and display the pulmonary nodules well. The image quality and nodule detection are not inferior to RDCT Hybrid IR routinely used in clinical practice.
ABSTRACT
Objective:To explore application value of improving quality of the low dose pancreatic CT images by using deep learning reconstruction (DLR).Methods:From August to December 2020, 68 patients who underwent contrast-enhanced pancreatic CT were prospectively collected in Peking Union Medical College Hospital. All patients were randomly divided into routine dose group (34 patients, with tube voltage of 120 kV) and low dose group (34 patients, with tube voltage of 100 kV). All patients underwent non-contrast, arterial phase, parenchymal phase and delay phase scans. The four-phase images of low dose group were reconstructed by using filtered back projection (FBP), hybrid iterative reconstruction (AIDR) and DLR which were marked with LD-FBP, LD-AIDR and LD-DLR, respectively. The four-phase images of routine dose group were reconstructed by using AIDR algorithm which were marked with RD-AIDR. The CT value, image noise (SD), signal to noise ratio (SNR) and contrast to noise ratio (CNR) of pancreas were measured. The ANOVA test was performed in comparison with objective parameters of different reconstruction algorithms, and LSD test was performed in pairwise comparison. The subjective image scores were obtained and were compared using Kruskal-Wallis test.Results:CT value, SD, SNR and CNR of non-contrast, arterial phase, parenchymal phase and delay phase had significant difference among different reconstruction images of routine dose group and low dose group (all P<0.05). The CT value of LD-FBP, LD-AIDR, and LD-DLR images were significantly higher than those of RD-AIDR images in parenchymal phase and delay phase (all P<0.05). There were statistically significant differences in each pairwise comparison of SD and SNR of four phase images (all P<0.05). There were statistically significant differences of CNR among LD-FBP, LD-DLR and RD-AIDR in four phase images (all P<0.05). The CNR of RD-AIDR was better than that of LD-FBP, and CNR of LD-DLR was better than that of RD-AIDR. DLR algorithm improved the SD, SNR and CNR of four phases of pancreatic images. The improvement of SNR was more significant after contrast enhancement, and the improvement of CNR was more significant in the non-contrast and delay phases. Subjective image scores of different reconstruction images were statistically different in four phase images (all P<0.001). Overall image scores of LD-DLR and RD-AIDR had no significant differences in four phase ( Z value of four phases were 1.00, 2.24, 0.45 and 1.34, respectively; P value of four phases were 0.317, 0.025, 0.655 and 0.180, respectively). Conclusion:The DLR technology can decrease radiation dose of pancreatic CT, improve image quality and satisfy diagnostic requirement. The DLR technology can also reduce image noise, improve the SNR and CNR in low dose contrast-enhanced pancreatic CT.