Search

Rationale and Objective: Uterine leiomyomas with high signal intensity on T2-weighted imaging (T2WI) can be difficult to distinguish from sarcomas. This study assessed the feasibility of using machine learning to differentiate uterine sarcomas from leiomyomas with high signal intensity on T2WI on multiparametric magnetic resonance imaging., Materials and Methods: This retrospective study included 80 patients (50 with benign leiomyoma and 30 with uterine sarcoma) who underwent pelvic 3 T magnetic resonance imaging examination for the evaluation of uterine myometrial smooth muscle masses with high signal intensity on T2WI. We used six machine learning techniques to develop prediction models based on 12 texture parameters on T1WI and T2WI, apparent diffusion coefficient maps, and contrast-enhanced T1WI, as well as tumor size and age. We calculated the areas under the curve (AUCs) using receiver-operating characteristic analysis for each model by 10-fold cross-validation and compared these to those for two board-certified radiologists., Results: The eXtreme Gradient Boosting model gave the highest AUC (0.93), followed by the random forest, support vector machine, multilayer perceptron, k-nearest neighbors, and logistic regression models. Age was the most important factor for differentiation (leiomyoma 44.9 ± 11.1 years; sarcoma 58.9 ± 14.7 years; p < 0.001). The AUC for the eXtreme Gradient Boosting was significantly higher than those for both radiologists (0.93 vs 0.80 and 0.68, p = 0.03 and p < 0.001, respectively)., Conclusion: Machine learning outperformed experienced radiologists in the differentiation of uterine sarcomas from leiomyomas with high signal intensity on T2WI., (Copyright © 2019. Published by Elsevier Inc.)

Rationale and Objectives: To assess the effect of model-based iterative reconstruction (MBIR) on image quality and diagnostic performance of low-radiation-dose computed tomography colonography (CTC) in the preoperative assessment of colorectal cancer., Materials and Methods: This study included 30 patients with colorectal cancer referred for surgical treatment. All patients underwent CTC with a standard dose (SD) protocol in the supine position and a low-dose (LD; radiation dose reduction of approximately 85%) protocol in the prone position. The SD protocol images were post-processed using filtered back projection (FBP), whereas the LD protocol images were post-processed using FBP and MBIR. Objective and subjective image quality parameters were compared among the three different methods. Preoperative evaluations, including site, length, and tumor and node staging were performed, and the findings were compared to the postsurgical findings., Results: The mean image noise of SD-FBP, LD-FBP, and LD-MBIR images was 17.3 ± 3.2, 40.5 ± 10.9, and 11.2 ± 2.0 Hounsfield units, respectively. There were significant differences for all comparison combinations among the three methods (P < .01). For image noise, the mean visual scores were significantly higher for SD-FBP and LD-MBIR than for LD-FBP, and the scores for SD-FBP and LD-MBIR were equivalent (3.9 ± 0.3 [SD-FBP], 2.0 ± 0.5 [LD-FBP], and 3.7 ± 0.3 [LD-MBIR]). Preoperative information was more accurate under SD-FBP and LD-MBIR than under LD-FBP, and the information was comparable between SD-FBP and LD-MBIR., Conclusion: MBIR can yield significantly improved image quality on low-radiation-dose CTC and provide preoperative information equivalent to that of standard-radiation-dose protocol., (Copyright © 2018 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved.)

Rationale and Objectives: We compared the effect of iterative model reconstruction (IMR), filtered back projection (FBP), and hybrid iterative reconstruction (HIR) on coronary artery calcium (CAC) scoring., Materials and Methods: CAC scans of 30 consecutive patients (18 men and 12 women, age 70.1 ± 12.2 years) were reconstructed with FBP, HIR, and IMR, and the image noise was measured on all images. Two radiologists independently measured the CAC scores using semiautomated software, and interobserver agreement was evaluated. Statistical analysis included the Spearman correlation coefficient and Bland-Altman analysis., Results: The mean image noise on FBP, HIR, and IMR images was 48.0 ± 7.9, 29.6 ± 4.8, and 9.3 ± 1.3 Hounsfield units, respectively. The difference among all reconstruction combinations was significant (P < .01). The CAC score on HIR and IMR scans was 4.2% and 8.9% lower, respectively, than the CAC score on FBP images. There was no significant difference in the mean CAC score among the three reconstructions. The interobserver correlation was excellent for all three reconstructions (r 2  = 0.96 FBP, 0.99 HIR, 0.99 IMR); the best Bland-Altman measure of agreement was with IMR, followed by HIR and FBP., Conclusion: For CAC scoring, IMR can reduce the image noise and blooming artifacts, and consequently lowers the measured CAC score. IMR can lessen measurement variability and yield stable, reproducible measurements., (Copyright © 2017 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved.)

Rationale and Objectives: The purpose of this study was to evaluate the noise and image quality of images reconstructed with a knowledge-based iterative model reconstruction (knowledge-based IMR) in ultra-low dose cardiac computed tomography (CT)., Materials and Methods: We performed submillisievert radiation dose coronary CT angiography on 43 patients. We also performed a phantom study to evaluate the influence of object size with the automatic exposure control phantom. We reconstructed clinical and phantom studies with filtered back projection (FBP), hybrid iterative reconstruction (hybrid IR), and knowledge-based IMR. We measured effective dose of patients and compared CT number, image noise, and contrast noise ratio in ascending aorta of each reconstruction technique. We compared the relationship between image noise and body mass index for the clinical study, and object size for phantom study., Results: The mean effective dose was 0.98 ± 0.25 mSv. The image noise of knowledge-based IMR images was significantly lower than those of FBP and hybrid IR images (knowledge-based IMR: 19.4 ± 2.8; FBP: 126.7 ± 35.0; hybrid IR: 48.8 ± 12.8, respectively) (P < .01). The contrast noise ratio of knowledge-based IMR images was significantly higher than those of FBP and hybrid IR images (knowledge-based IMR: 29.1 ± 5.4; FBP: 4.6 ± 1.3; hybrid IR: 13.1 ± 3.5, respectively) (P < .01). There were moderate correlations between image noise and body mass index in FBP (r = 0.57, P < .01) and hybrid IR techniques (r = 0.42, P < .01); however, these correlations were weak in knowledge-based IMR (r = 0.27, P < .01)., Conclusion: Compared to FBP and hybrid IR, the knowledge-based IMR offers significant noise reduction and improvement in image quality in submillisievert radiation dose cardiac CT., (Copyright © 2016 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved.)

Rationale and Objectives: We investigated the effects of small focal spot (SFS) imaging and iterative model reconstruction (IMR) on the image quality of computed tomography angiographs (CTA) in patients with peripheral arterial disease., Materials and Methods: We divided 60 consecutive patients with suspected or confirmed peripheral artery disease into two equal groups. One group underwent large focal spot scanning under our standard CTA protocol with hybrid iterative reconstruction (iDose(4)) (protocol 1), and the other underwent scanning with the SFS protocol and IMR (protocol 2). Quantitative image quality parameters, ie, arterial computed tomography attenuation, image noise, and the contrast-to-noise ratio, were compared and the visual image quality (depiction of each vessel) was scored on a 5-point scale., Results: There was no significant difference in the arterial attenuation among all evaluated slice levels. The mean image noise was significantly lower under protocol 2 and the contrast-to-noise ratio was significantly higher at all slice levels. The visual scores assigned to the two protocols for the depiction of large vessels, such as the abdominal aorta and iliac artery, were comparable. However, the mean visual scores for small vessels in the lower extremities were significantly higher under protocol 2., Conclusion: CTA with SFS and IMR yielded significantly better qualitative and quantitative image quality especially for small vessels., (Copyright © 2016 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved.)