Evaluation of image quality of a novel computed tomography metal artifact management technique on an anthropomorphic head and neck phantom

Background and purpose

Artefacts caused by dental amalgam implants present a common challenge in computed tomography (CT) and therefore treatment planning dose calculations. The goal was to perform a quantitative image quality analysis of our Artifact Management for Proton Planning (AMPP) algorithm which used gantry tilts for managing metal artefacts on Head and Neck (HN) CT scans and major vendors’ commercial approaches.

Materials and methods

Metal artefact reduction (MAR) algorithms were evaluated using an anthropomorphic phantom with a removable jaw for the acquisition of images with and without (baseline) metal artifacts. AMPP made use of two angled CT scans to generate one artifact-reduced image set. The MAR algorithms from four vendors were applied to the images with artefacts and the analysis was performed with respective baselines. Planar HU difference maps and volumetric HU differences were analyzed.

Results

AMPP algorithm outperformed all vendors’ commercial approaches in the elimination of artefacts in the oropharyngeal region, showing the lowest percent of pixels outside +− 20 HU criteria, 4%; whereas those in the MAR-corrected images ranged from 26% to 67%. In the region of interest within the affected slices, the commercial MAR algorithms showed inconsistent performance, whereas the AMPP algorithm performed consistently well throughout the phantom’s posterior region.

Conclusions

A novel MAR algorithm was evaluated and compared to four commercial algorithms using an anthropomorphic phantom. Unanimously, the analysis showed the AMPP algorithm outperformed vendors’ commercial approaches, showing the potential to be broadly implemented, improve visualizations in patient anatomy and provide accurate HU information.

Computerized tomography before the final treatment cycle of neoadjuvant chemotherapy or induction chemotherapy in muscle-invasive urinary bladder cancer, cannot predict pathoanatomical outcomes and does not reflect prognosis-results of a single centre retrospective prognostic study

Background

Evaluating the routine of using control computer tomography (cCT) for determining the response status of muscle-invasive bladder cancer (MIBC) prior to final cycle of neoadjuvant chemotherapy (NAC) or induction chemotherapy (IC), in terms of predicting histopathological pTNM-staging and pathoanatomical responses/non-responses. Secondly, predicting two and three-year overall survival (OS).

Methods

Seventy-seven patients with localized MIBC (cT2-4aN0M0) and 3 patients with minimal nodal dissemination (cN1-2), undergoing NAC or IC and radical cystectomy (RC), the years 2006–2014 at Norrland university hospital in Umeå, Sweden. Baseline pre-cystectomy CTs and cCTs prior to final chemotherapy-cycle, were reviewed and underwent attempted RECIST-criteria categorization, into five response/non-response related subgroups (n=71). The diagnostic accuracy of cCT in comparison with pTNM was assessed using sensitivity, specificity, positive- and negative likelihood ratios. OS for 2 and 3 years was calculated, both in relation to histopathological pTNM-stages in all patients (n=80) and for the patients with cCT-evaluated categories (n=71). Multivariable analysis for OS, was performed in correlation to pTNM-stages firstly, and to radiological staging secondly.

Results

The sensitivity of cCT to predict non-responders according to pTMN was 64% and specificity 36%. The positive likelihood ratio=1 and the negative likelihood ratio =1. CT-evaluations couldn’t accurately predict pTNM-stages in terms of response/non-response. No statistically significant results were found in correlating cCTs with two and three-year OS.

Conclusions

cCT prior to planned final preoperative chemotherapy-cycle in MIBC patients undergoing NAC or IC, has a poor correlation with pTNM and cannot predict pathoanatomical responses. Prediction of OS based on cCTs is unfeasible.