How large matrices improve quality in ultra-high-resolution CT

According to a new study recently published in Academic Radiology, in ultra-high-resolution computed tomography (U-HRCT) scans, a large matrix size maintains the spatial resolution and improves the image quality and assessment of lung diseases when compared to a 512-matrix size. 

The researchers noted that commercially U-HRCT scans have "smaller detector elements and x-ray tube focus size" than standard CT.  

"As for matrix size, 512-by-512 has been used in conventional CT, but larger matrix sizes, such as 1024-by-1024 and 2048-by-2048, are available with U-HRCT," said the study's lead author Akinori Hata, MD, from the department of diagnostic and interventional radiology at the Osaka University Graduate School of Medicine in Japan. 

According to study methods, 11 cadaveric lungs and slit phantoms were scanned. Additionally, researchers used for slit phantom scans a 20-mm field of view (FOV) with a 1024 matrix size and a 320-mm FOV with 512, 1024, and 2048 matrix sizes. The 11 cadaveric lungs, on the other hand, were reconstructed using 512, 1024 and 2048 matrix sizes, according to the study.  

After scanning, three objective observers scored the images subjectively from worst to best in terms of six key factors: image quality, noise, streak artifact, vessel, bronchi and image findings.  

 "The median score of the three observers was evaluated by Wilcoxon signed-rank test with Bonferroni correction," Hata said. "Noise was measured quantitatively and evaluated with the Tukey test; a P value of 'less than 0.05' was considered significant."  

Study results included the following:  

  • The maximum spatial resolution was 0.14 mm.
  • Among the 320-mm FOV images, the 2048 matrix had the highest resolution and was significantly better than the 1024 matrix in terms of overall quality, solid nodule, ground-glass opacity, emphysema, intralobular reticulation, honeycombing, and clarity of vessels (P less than 0.05).
  •  The 2048 and 1024 matrices performed better than the 512 matrix (P less than 0.001), except for noise and streak artifact.
  • The visual and quantitative noise decreased significantly in the order of 512, 1024, and 2048 (P less than 0.001).

Hata and his team explained that the study is the first of its kind in evaluating the quality of a 2048-by-2048 matrix obtained by U-HRCT. Additionally, a CT scanner's maximum spatial resolution is dependent on focus size of the x-ray tube and detector element size, among other factors, according to the study.  

"Matrix size influences the spatial resolution of a reconstructed image, but changing the matrix size did not make the spatial resolution of the reconstructed image better than the maximum spatial resolution of the CT scanner," Hata concluded.