- 學位論文
重建演算型式在低輻射劑量肺部電腦斷層之影像評估
Evaluation of Reconstruction Algorithm on Image Quality of Low-Radiaton-Dose Lung CT
摘要
低輻射劑量電腦斷層是肺癌篩檢的重要利器，但在低輻射劑量時，影像品質對許多參數的調整非常敏感，重建演算型式適當的選用，可間接降低輻射劑量，且對影像品質有極大的影響。本研究將臨床肺癌篩檢之影像，以5種重建演算型式重組，由兩位觀測醫師以五分評量表（five-point scale），以及應用於影像評估的Moran I test及量測對比雜訊比（contrast-to-noise ratio）等三種方式，來評估肺部低輻射劑量電腦斷層在使用不同重建演算型式時之影像品質。兩位觀測醫師評估結果，A重建演算型式為4.07± 0.61，B為4.15± 0.58，C為4.37± 0.59，F為4.33± 0.59，L為4.00± 0.73，兩位觀測醫師評估結果，組間同意度檢定有顯著差異（p＜ 0.001）；在Moran I test評估方面，較鮮銳之C、F重建演算型式在Moran I test的線型相似，雖犧牲了部份影像資訊，對肺部影像而言，影像雜訊與空間結構是較為折衷且肺部組織結構仍是清晰且能診斷的，對比雜訊比也呈現類似的結果；然而，更鮮銳的肺部L演算型式，其肺部結構細節雖然更清晰，但雜訊也相對更多，在Moran I test的線型與對比雜訊比在其他演算型式差異較大，在觀測醫師之間形成截然不同的評估結果，這同時也顯示利用較為客觀的評估方式在影像評估上的一致性與重要性。
並列摘要
Low-radiation-dose lung CT is a promising method for lung cancer screening. The image quality is affected by parameters of reconstruction algorithm, so it is important to choose proper reconstruction algorithm under low-radiation-dose. The purpose of this study is to evaluate the effect of five different reconstruction algorithms on image quality of low- radiation-dose lung CT. In this study, each lung cancer screening image (low-radiation-dose lung CT) was divided into five groups (A, B, C, F and L) with usage of five different reconstruction algorithms. These images were subjectively reviewed by two chest radiologists with the five-point scale, and were objectively compared by usage of the Moran I test and contrast-to-noise ratio(CNR). In the subjective comparison, the mean and standard deviation were 4.07± 0.61 in group A, 4.15± 0.58 in group B, 4.37± 0.59 in group C, 4.33± 0.59 in group F, and 4.00± 0.73 in group L. The interobserver agreement have significant different（p＜ 0.001）. In the objective comparison, the result of sharp algorism, C and F group, in Moran I test and CNR was similar. Although loss of some image information in the sharp algorithm, C and F group, it may be a compromise in image noise and spatial resolution because the result of group C and F in the subjective comparison was better than others. The sharper algorism of group L is even more different than others in Moran I test, and is also worse than others in subjective comparison due to more noise. This study showed the importance of objective and subjective methods in comparison of the image quality.