Design and first results of a phantom study on the suitability of iterative reconstruction for lung-cancer screening with low-dose computer tomography

Abstract

In this research computer tomography (CT) iterative reconstruction (IR) algorithms are investigated, specifically the impact of their statistical and model-based strength on image quality in low-dose lung screening CT protocols in comparison to filtered back projection (FBP). It has been probed whether statistical, model-based IR in conjunction with low-dose, and ultra-low-dose protocols are suitable for lungcancer screening. To this end, artificial lung nodules shaped as spheres and spicules made from material with calibrated Hounsfield units (HU) were attached on marked positions in the lung structure of an anthropomorphic phantom. Nodule positions were selected by distinguished radiologists. The phantom with nodules was scanned on a CT Scanner using standard high contrast (SHC), low-dose (LD) and ultra low-dose (ULD) protocol. For reconstruction FBP and the IR algorithm ADMIRE at three different strength levels were used. Volume CT dose index (CTDIvol) and dose-length product were recorded. Radiologists assessed subjective image quality using a six-point Likert scale by reading all image series in terms detectability of lung nodules. As a measurable objective image quality parameter signal-to-noise ratios (SNR) were investigated. The CTDIvol decreases by more than 70% for all protocols and nodules compared to diagnostic reference value for chest CT (p<0.00001). The evaluation of image quality parameters, i.e. SNR, indicates that LD and ULD protocols in conjunction with IR assert high quality lung-nodule detection. The results reveal that IR algorithm with moderate to high strength is an indispensable alternative to FBP in low-dose scanning, thus, potentially suitable for lung-tumour screening.

Validation of iterative CT reconstruction by inter and intra observer performance assessment of artificial lung foci

“…The influence of varying kernels (fixed phantom diameter) with IR could be the reason why more random rankings were obtained, whereas more clinically conditions generated only one random ranking. It was found in [5] that radiologists rate BR69 and BR32 worse than BL57 while evaluating anonymized and randomized single images, so we expected a reliable ranking even with varying kernels. Potentially the ranking test method is more challenging for radiologists to answer than a dichotomous categorical assess-ment or multiple-stage categorical assessment.…”

“…CT protocol and reconstruction. The entire raw data acquisition was performed on a Somatom Force CT, further specifications in [5]. For each setup three CT dose protocols were selected: (i) standard high contrast (SHC; 120kV/51mAs), (ii) low-dose (LD; 120kV/40mAs) and (iii) ultra-low-dose (ULD; 120kV/20mAs).…”

“…Raw data were acquired using FOV 430mm, rotation time 0.5s, delay 2s, beam collimation 192x0.6mm, deactivated CAREDose4D, slice thickness 1mm, increment 1mm. Preparations scans and reconstruction were performed as in [5]. The acquired raw data of each protocol were reconstructed by utilizing three kernels: BL57 is a standard for lung node detection, BR32 is a soft, and BR69 is hard kernel.…”

“…IR reconstruction may outperform traditional analytical methods as image impression alters with increasing algorithm strength. In fact, in [5] has been reported that radiologists have reservations with regard to IR reconstructed image and its possible influence on diagnostics. Furthermore, radiologists evaluated IR images rather different which may be due to professional experience and could mirror their accustoming.…”

Computed tomography lung-cancer screening: does iterative reconstruction matter?

“…The effective dose and convolution kernels effects on the detection of pulmonary nodules in IR were investigated in studies with anthropomorphic lung phantoms by quantifying the perception of expert radiologists' at a Likert scale, contrast-to-noise and signal-to-noise ratio (SNR) (19)(20)(21). Other studies evaluated machine learning (ML), computer-aided detection (CAD) software systems or commercially available deep learning ML based CAD systems by investigating receiver-operating-curve measures in artificial or ex-vivo lung phantoms (22)(23)(24)(25).…”