Application value of combining low tube voltage and deep learning image reconstruction algorithm in head and neck CT angiography with reduced radiation and contrast doses.
Authors
Affiliations (5)
Affiliations (5)
- Department of Radiology, The First Affiliated Hospital of Xi'an Jiaotong University, Xi'an, China.
- Ultrasonic Medicine Department, First Affiliated Hospital of Xi'an Jiaotong University, Xi'an, China.
- GE Healthcare, Computed Tomography Research Center, Beijing 100176, China.
- Department of Neurology, The First Affiliated Hospital of Xi'an Jiaotong University, Xi'an, China.
- Department of Radiology, The First Affiliated Hospital of Xi'an Jiaotong University, Xi'an, China. Electronic address: [email protected].
Abstract
To demonstrate the value of combining low tube voltage (80kVp) and deep learning image reconstruction (DLIR) in head and neck CT angiography (CTA) with reduced radiation and contrast doses, in comparison with routine dose scans with 100kVp and adaptive statistical iterative reconstruction V (ASIR-V). This prospective study included 100 patients scheduled for head and neck CTA due to vascular diseases. Patients were prospectively and randomly assigned in a 1:1 ratio to either the lower-dose group (group A, n = 50) with 80kVp, automatic tube current modulation (ATCM) scanning and contrast medium (CM) dose rate of 1.2gI/s for 10 s, and conventional group (group B, n = 50) with 100kVp, ATCM and CM dose rate of 1.6gI/s for 10 s. Group A used the high-level DLIR (DLIR-H) (subgroup A1), medium-level DLIR (DLIR-M) (subgroup A2), and 50% ASIR-V (A3) for image reconstruction; while Group B used ASIR-V50% for image reconstruction. The CT attenuation and standard deviation (SD) values of the sternocleidomastoid muscle (SCM) at the level of hypopharynx and epiglottis and the cerebral white matter (WM) at the level of splenium of corpus callosum were measured. For both quantitative and qualitative assessments, we evaluated image quality across the four groups. Objectively, we calculated the signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and edge rise slope (ERS) to quantify vessel sharpness. Subjectively, image quality was assessed using subjective image quality scores (SIQS). Finally, using digital subtraction angiography (DSA) as the reference standard, we compared the diagnostic accuracy of the four groups for determining the degree of vessel stenosis. Group A showed significant reductions in contrast dosage, injection rate, and radiation dose by 25.61%, 28.38%, and 36.11%, respectively, compared to group B (all P < 0.001). After contrast dose reduction, the vessel attenuation values in group A were similar to that in group B. However, the four image groups differed significantly in terms of background noise (all P < 0.05) with group A1 having the lowest noise. Group A1 also had significantly higher SNR and CNR values compared to group B in all vessels (all P < 0.05). In addition, Group A1 had the highest SIQS, followed by A2, B, and A3 with good agreement between the two reviewers in all groups (κ values 0.88---1). With DSA as the reference standard, analysis of the 23 segments with > 50% stenosis showed that group A1 with DLIR-H exhibited a strong correlation with DSA in quantifying vascular stenosis (Pearson's r = 0.97) and achieved the lowest MSE (14.83) among the four reconstruction types. Compared to the standard-dose protocol using 100 kVp and ASIR-V50%, a protocol using 80 kVp combined with DLIR-H significantly reduces the radiation dose (by 36%), contrast dose (26%), and injection rate (28%) in head and neck CTA while still significantly improving image quality and providing better diagnostic agreement with DSA. Combination of low tube voltage and DLIR-H may be used to reduce patient doses while improving image quality and diagnostic accuracy in head and neck CT imaging.