Koichiro Yasaka

Model-based iterative reconstruction technique for ultralow-dose chest CT: comparison of pulmonary nodule detectability with the adaptive statistical iterative reconstruction technique.

PurposeThe purpose of this study was to evaluate whether model-based iterative reconstruction (MBIR) enables dose reduction over adaptive iterative reconstruction (ASIR) while maintaining diagnostic performance. MethodsIn this institutional review board–approved and Health Insurance Portability and Accountability Act–compliant study, 59 patients (mean [SD] age, 64.7 [13.4] years) gave informed consent to undergo reference-, low-, and ultralow-dose chest computed tomography (CT) with 64-row multidetector CT. The reference- and low-dose CT involved the use of automatic tube current modulation with fixed noise indices (31.5 and 70.44 at 0.625 mm, respectively) and were reconstructed with 50% ASIR-filtered back projection blending. The ultralow-dose CT was acquired with a fixed tube current-time product of 5 mA s and reconstructed with MBIR. Two radiologists evaluated 2.5- and 0.625-mm-slice–thick axial images from low-dose ASIR and ultralow-dose MBIR, recorded the pattern of each nodule candidate, and assigned each a confidence score. A reference standard was established by a consensus panel of 2 different radiologists, who identified 84 noncalcified nodules with diameters of 4 mm or greater on reference-dose ASIR (ground-glass opacity, n = 18; partly solid, n = 11; solid, n = 55). Sensitivity in nodule detection was assessed using the McNemar test. Jackknife alternative free-response receiver operating characteristic (JAFROC) analysis was applied to assess the results including confidence scores. ResultsCompared with the low-dose CT, a 78.1% decrease in dose-length product was seen with the ultralow-dose CT. No significant differences were observed between the low-dose ASIR and the ultralow-dose MBIR for overall nodule detection in sensitivity (P = 0.48–0.69) or the JAFROC analysis (P = 0.57). Likewise, no significant differences were seen for ground-glass opacity, partly solid, or solid nodule detection in sensitivity (P = 0.08–0.65) or the JAFROC analysis (P = 0.21–0.90). ConclusionsModel-based iterative reconstruction enables nearly an 80% reduction in radiation dose for chest CT from a low-dose level to an ultralow-dose level, without affecting nodule detectability.

Model-based iterative reconstruction for reduction of radiation dose in abdominopelvic CT: comparison to adaptive statistical iterative reconstruction

PurposeTo evaluate dose reduction and image quality of abdominopelvic computed tomography (CT) reconstructed with model-based iterative reconstruction (MBIR) compared to adaptive statistical iterative reconstruction (ASIR).Materials and methodsIn this prospective study, 85 patients underwent referential-, low-, and ultralow-dose unenhanced abdominopelvic CT. Images were reconstructed with ASIR for low-dose (L-ASIR) and ultralow-dose CT (UL-ASIR), and with MBIR for ultralow-dose CT (UL-MBIR). Image noise was measured in the abdominal aorta and iliopsoas muscle. Subjective image analyses and a lesion detection study (adrenal nodules) were conducted by two blinded radiologists. A reference standard was established by a consensus panel of two different radiologists using referential-dose CT reconstructed with filtered back projection.ResultsCompared to low-dose CT, there was a 63% decrease in dose-length product with ultralow-dose CT. UL-MBIR had significantly lower image noise than L-ASIR and UL-ASIR (all p<0.01). UL-MBIR was significantly better for subjective image noise and streak artifacts than L-ASIR and UL-ASIR (all p<0.01). There were no significant differences between UL-MBIR and L-ASIR in diagnostic acceptability (p>0.65), or diagnostic performance for adrenal nodules (p>0.87).ConclusionMBIR significantly improves image noise and streak artifacts compared to ASIR, and can achieve radiation dose reduction without severely compromising image quality.

Comparison of pure and hybrid iterative reconstruction techniques with conventional filtered back projection: image quality assessment in the cervicothoracic region.

OBJECTIVES To evaluate the impact on image quality of three different image reconstruction techniques in the cervicothoracic region: model-based iterative reconstruction (MBIR), adaptive statistical iterative reconstruction (ASIR), and filtered back projection (FBP). METHODS Forty-four patients underwent unenhanced standard-of-care clinical computed tomography (CT) examinations which included the cervicothoracic region with a 64-row multidetector CT scanner. Images were reconstructed with FBP, 50% ASIR-FBP blending (ASIR50), and MBIR. Two radiologists assessed the cervicothoracic region in a blinded manner for streak artifacts, pixilated blotchy appearances, critical reproduction of visually sharp anatomical structures (thyroid gland, common carotid artery, and esophagus), and overall diagnostic acceptability. Objective image noise was measured in the internal jugular vein. Data were analyzed using the sign test and pair-wise Students t-test. RESULTS MBIR images had significant lower quantitative image noise (8.88 ± 1.32) compared to ASIR images (18.63 ± 4.19, P<0.01) and FBP images (26.52 ± 5.8, P<0.01). Significant improvements in streak artifacts of the cervicothoracic region were observed with the use of MBIR (P<0.001 each for MBIR vs. the other two image data sets for both readers), while no significant difference was observed between ASIR and FBP (P>0.9 for ASIR vs. FBP for both readers). MBIR images were all diagnostically acceptable. Unique features of MBIR images included pixilated blotchy appearances, which did not adversely affect diagnostic acceptability. CONCLUSIONS MBIR significantly improves image noise and streak artifacts of the cervicothoracic region over ASIR and FBP. MBIR is expected to enhance the value of CT examinations for areas where image noise and streak artifacts are problematic.

Deep Learning with Convolutional Neural Network for Differentiation of Liver Masses at Dynamic Contrast-enhanced CT: A Preliminary Study

Purpose To investigate diagnostic performance by using a deep learning method with a convolutional neural network (CNN) for the differentiation of liver masses at dynamic contrast agent-enhanced computed tomography (CT). Materials and Methods This clinical retrospective study used CT image sets of liver masses over three phases (noncontrast-agent enhanced, arterial, and delayed). Masses were diagnosed according to five categories (category A, classic hepatocellular carcinomas [HCCs]; category B, malignant liver tumors other than classic and early HCCs; category C, indeterminate masses or mass-like lesions [including early HCCs and dysplastic nodules] and rare benign liver masses other than hemangiomas and cysts; category D, hemangiomas; and category E, cysts). Supervised training was performed by using 55 536 image sets obtained in 2013 (from 460 patients, 1068 sets were obtained and they were augmented by a factor of 52 [rotated, parallel-shifted, strongly enlarged, and noise-added images were generated from the original images]). The CNN was composed of six convolutional, three maximum pooling, and three fully connected layers. The CNN was tested with 100 liver mass image sets obtained in 2016 (74 men and 26 women; mean age, 66.4 years ± 10.6 [standard deviation]; mean mass size, 26.9 mm ± 25.9; 21, nine, 35, 20, and 15 liver masses for categories A, B, C, D, and E, respectively). Training and testing were performed five times. Accuracy for categorizing liver masses with CNN model and the area under receiver operating characteristic curve for differentiating categories A-B versus categories C-E were calculated. Results Median accuracy of differential diagnosis of liver masses for test data were 0.84. Median area under the receiver operating characteristic curve for differentiating categories A-B from C-E was 0.92. Conclusion Deep learning with CNN showed high diagnostic performance in differentiation of liver masses at dynamic CT.

High-resolution CT with new model-based iterative reconstruction with resolution preference algorithm in evaluations of lung nodules: Comparison with conventional model-based iterative reconstruction and adaptive statistical iterative reconstruction

OBJECTIVE To compare the image quality of high-resolution computed tomography (HRCT) for evaluating lung nodules reconstructed with the new version of model-based iterative reconstruction and spatial resolution preference algorithm (MBIRn) vs. conventional model-based iterative reconstruction (MBIRc) and adaptive statistical iterative reconstruction (ASIR). MATERIALS AND METHODS This retrospective clinical study was approved by our institutional review board and included 70 lung nodules in 58 patients (mean age, 71.2±10.9years; 34 men and 24 women). HRCT of lung nodules were reconstructed using MBIRn, MBIRc and ASIR. Objective image noise was measured by placing the regions of interest on lung parenchyma. Two blinded radiologists performed subjective image analyses. RESULTS Significant improvements in the following points were observed in MBIRn compared with ASIR (p<0.005): objective image noise (24.4±8.0 vs. 37.7±10.4), subjective image noise, streak artifacts, and adequateness for evaluating internal characteristics and borders of nodules. The sharpness of small vessels and bronchi and diagnostic acceptability with MBIRn were significantly better than with MBIRc and ASIR (p<0.008). CONCLUSION HRCT reconstructed with MBIRn provides diagnostically more acceptable images for the detailed analyses of lung nodules compared with MBIRc and ASIR.

Liver Fibrosis: Deep Convolutional Neural Network for Staging by Using Gadoxetic Acid–enhanced Hepatobiliary Phase MR Images

Purpose To investigate the performance of a deep convolutional neural network (DCNN) model in the staging of liver fibrosis using gadoxetic acid-enhanced hepatobiliary phase magnetic resonance (MR) imaging. Materials and Methods This retrospective study included patients for whom input data (hepatobiliary phase MR images, static magnetic field of the imaging unit, and hepatitis B and C virus testing results available, either positive or negative) and reference standard data (liver fibrosis stage evaluated from biopsy or surgical specimens obtained within 6 months of the MR examinations) were available were assigned to the training (534 patients) or the test (100 patients) group. For the training group (54, 53, 81, 113, and 233 patients with fibrosis stages F0, F1, F2, F3, and F4, respectively; mean patient age, 67.4 ± 9.7 years; 388 men and 146 women), MR images with three different section levels were augmented 90-fold (rotated, parallel-shifted, brightness-changed and contrast-changed images were generated; a total of 144 180 images). Supervised training was performed by using the DCNN model to minimize the difference between the output data (fibrosis score obtained through deep learning [FDL score]) and liver fibrosis stage. The performance of the DCNN model was evaluated in the test group (10, 10, 15, 20, and 45 patients with fibrosis stages F0, F1, F2, F3, and F4, respectively; mean patient age, 66.8 years ± 10.7; 71 male patients and 29 female patients) with receiver operating characteristic (ROC) analyses. Results The FDL score was correlated significantly with fibrosis stage (Spearman rank correlation coefficient: 0.63; P < .001). Fibrosis stages F4, F3, and F2 were diagnosed with areas under the ROC curve of 0.84, 0.84, and 0.85, respectively. Conclusion The DCNN model exhibited a high diagnostic performance in the staging of liver fibrosis.

Precision of quantitative computed tomography texture analysis using image filtering

Abstract Quantitative computed tomography (CT) texture analyses for images with and without filtration are gaining attention to capture the heterogeneity of tumors. The aim of this study was to investigate how quantitative texture parameters using image filtering vary among different computed tomography (CT) scanners using a phantom developed for radiomics studies. A phantom, consisting of 10 different cartridges with various textures, was scanned under 6 different scanning protocols using four CT scanners from four different vendors. CT texture analyses were performed for both unfiltered images and filtered images (using a Laplacian of Gaussian spatial band-pass filter) featuring fine, medium, and coarse textures. Forty-five regions of interest were placed for each cartridge (x) in a specific scan image set (y), and the average of the texture values (T(x,y)) was calculated. The interquartile range (IQR) of T(x,y) among the 6 scans was calculated for a specific cartridge (IQR(x)), while the IQR of T(x,y) among the 10 cartridges was calculated for a specific scan (IQR(y)), and the median IQR(y) was then calculated for the 6 scans (as the control IQR, IQRc). The median of their quotient (IQR(x)/IQRc) among the 10 cartridges was defined as the variability index (VI). The VI was relatively small for the mean in unfiltered images (0.011) and for standard deviation (0.020–0.044) and entropy (0.040–0.044) in filtered images. Skewness and kurtosis in filtered images featuring medium and coarse textures were relatively variable across different CT scanners, with VIs of 0.638–0.692 and 0.430–0.437, respectively. Various quantitative CT texture parameters are robust and variable among different scanners, and the behavior of these parameters should be taken into consideration.

Dose-reduced CT with model-based iterative reconstruction in evaluations of hepatic steatosis: How low can we go?

PURPOSE To determine whether dose-reduced CT with model-based iterative image reconstruction (MBIR) is a useful tool with which to diagnose hepatic steatosis. MATERIALS AND METHODS This prospective clinical study approved by our Institutional Review Board included 103 (67 men and 36 women; mean age, 64.3 years) patients who provided written informed consent to undergo unenhanced CT. Images of reference-dose CT (RDCT) with filtered back projection (R-FBP) and low- and ultralow-dose CT (dose-length product; 24 and 9% of that of RDCT) with MBIR (L-MBIR and UL-MBIR) were reconstructed. Mean CT numbers of liver (CT[L]) and spleen (CT[S]), and quotient (CT[L/S]) of CT[L] and CT[S] were calculated from selected regions of interest. Bias and limits of agreement (LOA) of CT[L] and CT[L/S] in L-MBIR and UL-MBIR (vs. R-FBP) were assessed using Bland-Altman analyses. Diagnostic methods for hepatic steatosis of CT[L]<48 Hounsfield units (HU) and CT[L/S]<1.1 were applied to L-MBIR and UL-MBIR using R-FBP as the reference standard. RESULTS Bias was larger for CT[L] in UL-MBIR than in L-MBIR (-3.18HU vs. -1.73HU). The LOA of CT[L/S] was larger for UL-MBIR than for L-MBIR (±0.425 vs. ±0.245) and outliers were identified in CT[L/S] of UL-MBIR. Accuracy (0.92-0.95) and the area under the receiver operating characteristics curve (0.976-0.992) were high for each method, but some were slightly lower in UL-MBIR than L-MBIR. CONCLUSION Dose-reduced CT reconstructed with MBIR is applicable to diagnose hepatic steatosis, however, a low dose of radiation might be preferable.

Metal artefact reduction for patients with metallic dental fillings in helical neck computed tomography: comparison of adaptive iterative dose reduction 3D (AIDR 3D), forward-projected model-based iterative reconstruction solution (FIRST) and AIDR 3D with single-energy metal artefact reduction (SEMAR)

OBJECTIVES To compare the differences in metal artefact degree and the depiction of structures in helical neck CT, in patients with metallic dental fillings, among adaptive iterative dose reduction three dimensional (AIDR 3D), forward-projected model-based iterative reconstruction solution (FIRST) and AIDR 3D with single-energy metal artefact reduction (SEMAR-A). METHODS In this retrospective clinical study, 22 patients (males, 13; females, 9; mean age, 64.6 ± 12.6 years) with metallic dental fillings who underwent contrast-enhanced helical CT involving the oropharyngeal region were included. Neck axial images were reconstructed with AIDR 3D, FIRST and SEMAR-A. Metal artefact degree and depiction of structures (the apex and root of the tongue, parapharyngeal space, superior portion of the internal jugular chain and parotid gland) were evaluated on a four-point scale by two radiologists. Placing regions of interest, standard deviations of the oral cavity and nuchal muscle (at the slice where no metal exists) were measured and metal artefact indices were calculated (the square root of the difference of the squares of them). RESULTS In SEMAR-A, metal artefact was significantly reduced and depictions of all structures were significantly improved compared with those in FIRST and AIDR 3D (p ≤ 0.001, sign test). Metal artefact index for the oral cavity in AIDR 3D/FIRST/SEMAR-A was 572.0/477.7/88.4, and significant differences were seen between each reconstruction algorithm (p < 0.0001, Wilcoxon signed-rank test). CONCLUSIONS SEMAR-A could provide images with lesser metal artefact and better depiction of structures than AIDR 3D and FIRST.

Differentiation of adrenal tumors in patients with hepatocellular carcinoma: adrenal adenoma versus metastasis.

OBJECTIVE To investigate whether computed tomography (CT) attenuation test for differential diagnosis of adrenal nodule is applicable in patients with hepatocellular carcinoma (HCC) which shows similar image characteristics to adrenal adenoma. MATERIALS AND METHODS This retrospective study was approved by our institutional review board, and the requirement for informed consent from study patients was waived. Searching picture archiving and communication system, we identified 3678 patients with HCC who underwent upper abdominal unenhanced CT scans between April 2002 and March 2010, and 114 adrenal nodules (39 adenomas and 75 metastases) were included for analysis. Ten nodules were confirmed pathologically while 104 had imaging diagnosis (enlarged or emerged during the study period). Size, CT number, and the internal characteristics of the lesions were recorded. RESULTS Mean CT numbers of adrenal adenomas were significantly lower than those of metastases (P<0.0001, t-test) on unenhanced CT. Thresholds of 17 and 33 Hounsfield units (HU) provided the following sensitivity, specificity, and accuracy: 46.2%, 100%, and 81.6% at 17HU, and 94.9%, 89.3%, and 91.2% at 33HU, respectively. The area under receiver operating characteristic curve for the CT number test was 0.96. Metastases were significantly larger than adrenal adenoma (P=0.009, t-test). However, the accuracy of testing using mass size was 64.0% at most. All adenomas and metastases were depicted as homogeneous masses with the exception of two metastases that presented as heterogeneous masses (necrotic or lipomatous). CONCLUSION Adrenal adenomas can be differentiated from HCC metastases using CT number on unenhanced CT.