Femoral osteoporosis prediction model using autosegmentation and machine learning analysis with PyRadiomics on abdomen-pelvic computed tomography (CT).

Background: With the advancement of artificial intelligence technology and radiomics analysis, opportunistic prediction of osteoporosis with computed tomography (CT) is a new paradigm in osteoporosis screening. This study aimed to assess the diagnostic performance of osteoporosis prediction by the combination of autosegmentation of the proximal femur and machine learning analysis with a reference standard of dual-energy X-ray absorptiometry (DXA). Methods: Abdomen-pelvic CT scans were retrospectively analyzed from 1,122 patients who received both DXA and abdomen-pelvic computed tomography (APCT) scan from January 2018 to December 2020. The study cohort consisted of a training cohort and a temporal validation cohort. The left proximal femur was automatically segmented, and a prediction model was built by machine-learning analysis using a random forest (RF) analysis and 854 PyRadiomics features. The technical success rate of autosegmentation, diagnostic test, area under the receiver operator characteristics curve (AUC), and precision recall curve (AUC-PR) analysis were used to analyze the training and validation cohorts. Results: The osteoporosis prevalence of the training and validation cohorts was 24.5%, and 10.3%, respectively. The technical success rate of autosegmentation of the proximal femur was 99.7%. In the diagnostic test, the training and validation cohorts showed 78.4% vs. 63.3% sensitivity, 89.4% vs. 98.1% specificity. The prediction performance to identify osteoporosis within the groups used for training and validation cohort was high and the AUC and AUC-PR to forecast the occurrence of osteoporosis within the training and validation cohorts were 90.8% [95% confidence interval (CI), 88.4-93.2%] vs. 78.0% (95% CI, 76.0-79.9%) and 94.6% (95% CI, 89.3-99.8%) vs. 88.8% (95% CI, 86.2-91.5%), respectively. Conclusions: The osteoporosis prediction model using autosegmentation of proximal femur and machine-learning analysis with PyRadiomics features on APCT showed excellent diagnostic feasibility and technical success.

Development and Validation of a Deep-Learning-Based Algorithm for Detecting and Classifying Metallic Implants in Abdominal and Spinal CT Topograms.

PURPOSE: To develop and validate a deep-learning-based algorithm (DLA) that is designed to segment and classify metallic objects in topograms of abdominal and spinal CT. METHODS: DLA training for implant segmentation and classification was based on a U-net-like architecture with 263 annotated hip implant topograms and 2127 annotated spine implant topograms. The trained DLA was validated with internal and external datasets. Two radiologists independently reviewed the external dataset consisting of 2178 abdomen anteroposterior (AP) topograms and 515 spine AP and lateral topograms, all collected in a consecutive manner. Sensitivity and specificity were calculated per pixel row and per patient. Pairwise intersection over union (IoU) was also calculated between the DLA and the two radiologists. RESULTS: The performance parameters of the DLA were consistently >95% in internal validation per pixel row and per patient. DLA can save 27.4% of reconstruction time on average in patients with metallic implants compared to the existing iMAR. The sensitivity and specificity of the DLA during external validation were greater than 90% for the detection of spine implants on three different topograms and for the detection of hip implants on abdominal AP and spinal AP topograms. The IoU was greater than 0.9 between the DLA and the radiologists. However, the DLA training could not be performed for hip implants on spine lateral topograms. CONCLUSIONS: A prototype DLA to detect metallic implants of the spine and hip on abdominal and spinal CT topograms improves the scan workflow with good performance for both spine and hip implants.

Rim enhancement of pancreatic ductal adenocarcinoma: investigating the relationship with DCE-MRI-based radiomics and next-generation sequencing.

Purpose: To identify the clinical and genetic variables associated with rim enhancement of pancreatic ductal adenocarcinoma (PDAC) and to develop a dynamic contrast-enhanced (DCE) MRI-based radiomics model for predicting the genetic status from next-generation sequencing (NGS). Materials and methods: Patients with PDAC, who underwent pretreatment pancreatic DCE-MRI between November 2019 and July 2021, were eligible in this prospective study. Two radiologists evaluated presence of rim enhancement in PDAC, a known radiological prognostic indicator, on DCE MRI. NGS was conducted for the tissue from the lesion. The Mann-Whitney U and Chi-square tests were employed to identify clinical and genetic variables associated with rim enhancement in PDAC. For continuous variables predicting rim enhancement, the cutoff value was set based on the Youden's index from the receiver operating characteristic (ROC) curve. Radiomics features were extracted from a volume-of-interest of PDAC on four DCE maps (Ktrans, Kep, Ve, and iAUC). A random forest (RF) model was constructed using 10 selected radiomics features from a pool of 392 original features. This model aimed to predict the status of significant NGS variables associated with rim enhancement. The performance of the model was validated using test set. Results: A total of 55 patients (32 men; median age 71 years) were randomly assigned to the training (n = 41) and test (n = 14) sets. In the training set, KRAS, TP53, CDKN2A, and SMAD4 mutation rates were 92.3%, 61.8%, 14.5%, and 9.1%, respectively. Tumor size and KRAS variant allele frequency (VAF) differed between rim-enhancing (n = 12) and nonrim-enhancing (n = 29) PDACs with a cutoff of 17.22%. The RF model's average AUC from 10-fold cross-validation for predicting KRAS VAF status was 0.698. In the test set comprising 6 tumors with low KRAS VAF and 8 with high KRAS VAF, the RF model's AUC reached 1.000, achieving a sensitivity of 75.0%, specificity of 100% and accuracy of 87.5%. Conclusion: Rim enhancement of PDAC is associated with KRAS VAF derived from NGS-based genetic information. For predicting the KRAS VAF status in PDAC, a radiomics model based on DCE maps showed promising results.

Low-dose versus conventional CT urography using dual-source CT with different time-current product values and the same tube voltage: image quality and diagnostic performance in various diagnoses.

OBJECTIVES: To compare the image quality and diagnostic performance of low-dose CT urography to that of concurrently acquired conventional CT using dual-source CT. METHODS: This retrospective study included 357 consecutive CT urograms performed by third-generation dual-source CT in a single institution between April 2020 and August 2021. Two-phase CT images (unenhanced phase, excretory phase with split bolus) were obtained with two different tube current-time products (280 mAs for the conventional-dose protocol and 70 mAs for the low-dose protocol) and the same tube voltage (90 kVp) for the two X-ray tubes. Iterative reconstruction was applied for both protocols. Two radiologists independently performed quantitative and qualitative image quality analysis and made diagnoses. The correlation between the noise level or the effective radiation dose and the patients' body weight was evaluated. RESULTS: Significantly higher noise levels resulting in a significantly lower liver signal-to-noise ratio and contrast-to-noise ratio were noted in low-dose images compared to conventional images (P < .001). Qualitative analysis by both radiologists showed significantly lower image quality in low-dose CT than in conventional CT images (P < .001). Patient's body weight was positively correlated with noise and effective radiation dose (P < .001). Diagnostic performance for various diseases, including urolithiasis, inflammation, and mass, was not different between the two protocols. CONCLUSIONS: Despite inferior image quality, low-dose CT urography with 70 mAs and 90 kVp and iterative reconstruction demonstrated diagnostic performance equivalent to that of conventional CT for identifying various diseases of the urinary tract. ADVANCES IN KNOWLEDGE: Low-dose CT (25% radiation dose) with low tube current demonstrated diagnostic performance comparable to that of conventional CT for a variety of urinary tract diseases.

Optimal dual-energy computed tomography scan parameters to detect small-sized urinary stones and their composition.

To investigate the optimal scanning parameters of dual-energy computed tomography (DECT), which can accurately determine sensitivity (the detectability of urinary stones) and accuracy (the composition matching of urinary stones), and to apply them to clinical trials. Fifteen urinary stones were chemically analyzed, and their chemical compositions were considered a reference standard with which we compared the uric acid (UA) and non-UA compositions determined using DECT. The urinary stones were placed inside a bolus and scanned with a dual-source CT scanner under various selected dual-energy conditions (A to X) using various solid water phantom thicknesses. These datasets were analyzed using the Siemens syngo.via software tool (integrated into the CT system) for matching the sensitivity and accuracy assessments. This study showed that 80% of the highest sensitivity (detection of urinary stones) and 92% of the highest accuracy (composition matching of urinary stones) were achieved under condition A (a collimation beam width setting of 2 × 32 × 0.6 mm, an automatic exposure control setting of 80/sn140 peak kilovoltage, and a slice thickness of 0.5/0.5 mm) (P < 0.05). Application of the DECT energy parameters presented in the study will help identify the sensitivity and accuracy of UA and non-UA stone analysis, even in patients with small-sized urinary stones and in conditions difficult for analysis.

Ultra-low-dose hepatic multiphase CT using deep learning-based image reconstruction algorithm focused on arterial phase in chronic liver disease: A non-inferiority study.

PURPOSE: This study determined whether image quality and detectability of ultralow-dose hepatic multiphase CT (ULDCT, 33.3% dose) using a vendor-agnostic deep learning model(DLM) are noninferior to those of standard-dose CT (SDCT, 100% dose) using model-based iterative reconstruction(MBIR) in patients with chronic liver disease focusing on arterial phase. METHODS: Sixty-seven patients underwent hepatic multiphase CT using a dual-source scanner to obtain two different radiation dose CT scans (100%, SDCT and 33.3%, ULDCT). ULDCT using DLM and SDCT using MBIR were compared. A margin of -0.5 for the difference between the two protocols was pre-defined as noninferiority of the overall image quality of the arterial phase image. Quantitative image analysis (signal to noise ratio[SNR] and contrast to noise ratio[CNR]) was also conducted. The detectability of hepatic arterial focal lesions was compared using the Jackknife free-response receiver operating characteristic analysis. Non-inferiority was satisfied if the margin of the lower limit of 95%CI of the difference in figure-of-merit was less than -0.1. RESULTS: Mean overall arterial phase image quality scores with ULDCT using DLM and SDCT using MBIR were 4.35 ± 0.57 and 4.08 ± 0.58, showing noninferiority (difference: -0.269; 95 %CI, -0.374 to -0.164). ULDCT using DLM showed a significantly superior contrast-to-noise ratio of arterial enhancing lesion (p < 0.05). Figure-of-merit for detectability of arterial hepatic focal lesion was 0.986 for ULDCT using DLM and 0.963 for SDCT using MBIR, showing noninferiority (difference: -0.023, 95 %CI: -0.016 to 0.063). CONCLUSION: ULDCT using DLM with 66.7% dose reduction showed non-inferior overall image quality and detectability of arterial focal hepatic lesion compared to SDCT using MBIR.

Prospective evaluation of low-dose multiphase hepatic computed tomography for detecting and characterizing hepatocellular carcinoma in patients with chronic liver disease.

BACKGROUND: Knowing the lowest acceptable radiation dose of multiphase hepatic CT may allow us to reduce the radiation dose for detecting HCC. PURPOSE: To prospectively assess the image quality and diagnostic performance of low-dose and ultra-low-dose multiphase hepatic computed tomography using a dual-source CT scanner. METHODS: Three reconstructed different dose scan images (standard-dose, low-dose, and ultra-low-dose) of hepatic multiphase CT were obtained from 67 patients with a dual-source CT scanner. The image quality and the diagnostic performance of the three radiation dose CT scans of the hepatic focal lesion (≥ 0.5 cm) were analyzed by two independent readers using the Liver Imaging Reporting and Data System. RESULTS: Qualitative image quality and signal-to-noise ratio were significantly different among the radiation doses (p < 0.001). In total, 154 lesions comprising 32 hepatocellular carcinomas (HCC) and 122 non-HCC were included. The sensitivities of SDCT, LDCT, and ULDCT were 90.6%(29/32), 81.3%(26/32), and 56.2%(18/32), respectively. The accuracies of SDCT, LDCT, and ULDCT were 98.1%(151/154), 96.1%(148/154), and 89.6%(138/154), respectively. On per-lesion analysis, SDCT and LDCT did not show significantly different sensitivity and accuracy in diagnosing HCC (p = 0.250 and 0.250). CONCLUSIONS: The diagnostic performance of dynamic hepatic LDCT with 33% reduced radiation dose in comparison to SDCT would be acceptable even though its image quality was qualitatively and quantitatively inferior. However, few HCCs could be overlooked. Therefore, with caution, radiation dose reduction by one-third could be implemented for follow-up CT scans for patients suspected of having HCC with caution and further studies are needed in the future.

Comparison of diagnostic performance between dynamic versus static adenosine-stress myocardial CT perfusion to detect hemodynamically significant coronary artery stenosis: A prospective multicenter study.

Myocardial computed tomography perfusion (CTP) imaging is a noninvasive method for detecting myocardial ischemia. This study aimed to compare the diagnostic performance of dynamic and static adenosine-stress CTPs for detecting hemodynamically significant coronary stenosis. We prospectively enrolled 42 patients (mean age, 59.7 ± 8.8 years; 31 males) with ≥40% coronary artery stenosis. All patients underwent dynamic CTP for adenosine stress. The static CTP was simulated by choosing the seventh dynamic dataset after the initiation of the contrast injection. Diagnostic performance was compared with invasive fractional flow reserve (FFR) <0.8 as the reference. Of the 125 coronary vessels in 42 patients, 20 (16.0%) in 16 (38.1%) patients were categorized as hemodynamically significant. Dynamic and static CTP yielded similar diagnostic accuracy (90.4% vs 88.8% using visual analysis, P = .558; 77.6% vs 80.8% using quantitative analysis, P = .534; 78.4% vs 82.4% using combined visual and quantitative analyses, P = .426). The diagnostic accuracy of combined coronary computed tomography angiography (CCTA) and dynamic CTP (89.6% using visual analysis, P = .011; 88.8% using quantitative analysis, P = .018; 89.6% using combined visual and quantitative analyses, P = .011) and that of combined CCTA and static CTP (88.8% using visual analysis, P = .018; 90.4% using quantitative analysis, P = .006; 91.2% using combined visual and quantitative analyses, P = .003) were significantly higher than that of CCTA alone (77.6%). Dynamic CTP and static CTP showed similar diagnostic performance in the detection of hemodynamically significant stenosis.

Dual-energy CT of the liver: True noncontrast vs. virtual noncontrast images derived from multiple phases for the diagnosis of fatty liver.

PURPOSE: To evaluate the difference in liver density and to compare the performance to diagnose fatty liver between true noncontrast (TNC) images and virtual noncontrast (VNC) images generated from dual-energy CT (DECT). MATERIALS AND METHODS: Patients who underwent liver dynamic DECT and MRI were included (n = 49). Two observers measured the liver and spleen densities on TNC images and three VNC images from the arterial, portal and delayed phases of DECT (VNCa, VNCp and VNCd, respectively). The liver-minus-spleen density (density L-S) and liver-to-spleen ratio (density L/S) were calculated. The CT parameters were compared between normal liver patients and fatty liver patients by using the independent t-test. Differences and agreements between measurements on TNC images and VNC images were evaluated by using the paired t-test and Bland-Altman analysis. Receiver operating characteristic (ROC) curve analysis was used to evaluate the diagnostic performance of CT parameters for diagnosing fatty liver. RESULTS: All CT parameters measured on TNC and VNC images were significantly higher in normal liver patients than in fatty liver patients. Although the mean liver densities on VNC images were significantly lower than those on TNC images, all CT parameters showed good agreement between TNC images and VNC images. The diagnostic performances of CT parameters on VNC images were not significantly different from those on TNC images. CONCLUSION: Although the liver and spleen density on VNC images was significantly lower than that on TNC images, the diagnostic performances of CT parameters on three VNC images from multiple phases were similar to those on TNC images for diagnosing fatty liver.

Development and Validation of a Radiomics Model for Differentiating Bone Islands and Osteoblastic Bone Metastases at Abdominal CT.

Background It is important to diagnose sclerotic bone lesions in order to determine treatment strategy. Purpose To evaluate the diagnostic performance of a CT radiomics-based machine learning model for differentiating bone islands and osteoblastic bone metastases. Materials and Methods In this retrospective study, patients who underwent contrast-enhanced abdominal CT and were diagnosed with a bone island or osteoblastic metastasis between 2015 to 2019 at either of two different institutions were included: institution 1 for the training set and institution 2 for the external test set. Radiomics features were extracted. The random forest (RF) model was built using 10 selected features, and subsequent 10-fold cross-validation was performed. In the test phase, the RF model was tested with an external test set. Three radiologists reviewed the CT images for the test set. The sensitivity, specificity, accuracy, and area under the receiver operating characteristic curve (AUC) were calculated for the models and each of the three radiologists. The AUCs of the radiomics model and radiologists were compared. Results A total of 177 patients (89 with a bone island and 88 with metastasis; mean age, 66 years ± 12 [standard deviation]; 111 men) were in the training set, and 64 (23 with a bone island and 41 with metastasis; mean age, 69 years ± 14; 59 men) were in the test set. Radiomics features (n = 1218) were extracted. The average AUC of the RF model from 10-fold cross-validation was 0.89 (sensitivity, 85% [75 of 88 patients]; specificity, 82% [73 of 89 patients]; and accuracy, 84% [148 of 177 patients]). In the test set, the AUC of the trained RF model was 0.96 (sensitivity, 80% [33 of 41 patients]; specificity, 96% [22 of 23 patients]; and accuracy, 86% [55 of 64 patients]). The AUCs for the three readers were 0.95 (95% CI: 0.90, 1.00), 0.96 (95% CI: 0.90, 1.00), and 0.88 (95% CI: 0.80, 0.96). The AUC of radiomics model was higher than that of only reader 3 (0.96 vs 0.88, respectively; P = .03). Conclusion A CT radiomics-based random forest model was proven useful for differentiating bone islands from osteoblastic metastases and showed better diagnostic performance compared with an inexperienced radiologist. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Vannier in this issue.

Comparative performance analysis for abdominal phantom ROI detectability according to CT reconstruction algorithm: ADMIRE.

PURPOSE: We compared and analyzed the detectability performance pertaining to an abdominal phantom including a region of interest (ROI) according to a computed tomography (CT) reconstruction algorithm. METHODS: Three types of reconstruction algorithms (FBP, SAFIRE, and ADMIRE) were used to evaluate the detectability performance using the abdominal phantom (phantom size: 25 × 18 × 28 cm3 ). The vendor default settings for routine multi-detector computed tomography abdominal scans were used. As the quantitative evaluation method, the contrast-to-noise ratio (CNR), difference in coefficient of variation (COV) with the normalization based on the FBP data, and the noise power spectrum (NPS) were measured. RESULTS: The characteristic of the ADMIRE-3 reconstructed image was higher than those of the FBP and SAFIRE-3 reconstructed images. The CNR values of the SAFIRE and ADMIRE images were much higher than the corresponding values of the FBP images. The difference in COV values for the ADMIRE images was ~1.2 times lower than the corresponding values of the SAFIRE images. CONCLUSION: The comparative analysis of the abdominal phantom low-contrast resolution differences for each CT exposure parameters showed that ADMIRE demonstrated better results than SAFIRE and FBP in terms of contrast, CNR, COV difference, and 1D NPS. This indicates that ADMIRE can provide a clearer observation even with the same number of contrast objects as compared to SAFIRE and FBP owing to its better contrast resolution in the central part of the contrast hole at low kV.