Predicting Osteoporosis and Osteopenia by Fusing Deep Transfer Learning Features and Classical Radiomics Features Based on Single-Source Dual-energy CT Imaging.

RATIONALE AND OBJECTIVES: To develop and validate a predictive model for osteoporosis and osteopenia prediction by fusing deep transfer learning (DTL) features and classical radiomics features based on single-source dual-energy computed tomography (CT) virtual monochromatic imaging. METHODS: A total of 606 lumbar vertebrae with dual-energy CT imaging and quantitative CT (QCT) evaluation were included in the retrospective study and randomly divided into the training (n = 424) and validation (n = 182) cohorts. Radiomics features and DTL features were extracted from 70-keV monochromatic CT images, followed by feature selection and model construction, radiomics and DTL features models were established. Then, we integrated the selected two types of features into a features fusion model. We developed a two-level classifier for the hierarchical pairwise classification of each vertebra. All the vertebrae were first classified into osteoporosis and non-osteoporosis groups, then non-osteoporosis group was classified into osteopenia and normal groups. QCT was used as reference. The predictive performance and clinical usefulness of three models were evaluated and compared. RESULTS: The area under the curve (AUC) of the features fusion, radiomics and DTL models for the classification between osteoporosis and non-osteoporosis were 0.981, 0.999, 0.997 in the training cohort and 0.979, 0.943, 0.848 in the validation cohort. Furthermore, the AUCs of the previously mentioned models for the differentiation between osteopenia and normal were 0.994, 0.971, 0.996 in the training cohort and 0.990, 0.968, 0.908 in the validation cohort. The overall accuracy of the previously mentioned models for two-level classifications was 0.979, 0.955, 0.908 in the training cohort and 0.918, 0.885, 0.841 in the validation cohort. Decision curve analysis showed that all models had high clinical value. CONCLUSION: The feature fusion model can be used for osteoporosis and osteopenia prediction with improved predictive ability over a radiomics model or a DTL model alone.

Prediction of osteoporosis using radiomics analysis derived from single source dual energy CT.

BACKGROUND: With the aging population of society, the incidence rate of osteoporosis is increasing year by year. Early diagnosis of osteoporosis plays a significant role in the progress of disease prevention. As newly developed technology, computed tomography (CT) radiomics could discover radiomic features difficult to recognize visually, providing convenient, comprehensive and accurate osteoporosis diagnosis. This study aimed to develop and validate a clinical-radiomics model based on the monochromatic imaging of single source dual-energy CT for osteoporosis prediction. METHODS: One hundred sixty-four participants who underwent both single source dual-energy CT and quantitative computed tomography (QCT) lumbar-spine examination were enrolled in a study cohort including training datasets (n = 114 [30 osteoporosis and 84 non-osteoporosis]) and validation datasets (n = 50 [12 osteoporosis and 38 non-osteoporosis]). One hundred seven radiomics features were extracted from 70-keV monochromatic CT images. With QCT as the reference standard, a radiomics signature was built by using least absolute shrinkage and selection operator (LASSO) regression on the basis of reproducible features. A clinical-radiomics model was constructed by incorporating the radiomics signature and a significant clinical predictor (age) using multivariate logistic regression analysis. Model performance was assessed by its calibration, discrimination and clinical usefulness. RESULTS: The radiomics signature comprised 14 selected features and showed good calibration and discrimination in both training and validation cohorts. The clinical-radiomics model, which incorporated the radiomics signature and a significant clinical predictor (age), also showed good discrimination, with an area under the receiver operating characteristic curve (AUC) of 0.938 (95% confidence interval, 0.903-0.952) in the training cohort and an AUC of 0.988 (95% confidence interval, 0.967-0.998) in the validation cohort, and good calibration. The clinical-radiomics model stratified participants into groups with osteoporosis and non-osteoporosis with an accuracy of 94.0% in the validation cohort. Decision curve analysis (DCA) demonstrated that the radiomics signature and the clinical-radiomics model were clinically useful. CONCLUSIONS: The clinical-radiomics model incorporating the radiomics signature and a clinical parameter had a good ability to predict osteoporosis based on dual-energy CT monoenergetic imaging.

Associations between paraspinal muscles fatty infiltration and lumbar vertebral bone mineral density - An investigation by fast kVp switching dual-energy CT and QCT.

Purpose: To investigate the relationship between paraspinal muscles fat content and lumbar bone mineral density (BMD). Methods: A total of 119 participants were enrolled in our study (60 males, age: 50.88 ± 17.79 years, BMI: 22.80 ± 3.80 kg·m-2; 59 females, age: 49.41 ± 17.69 years, BMI: 22.22 ± 3.12 kg·m-2). Fat content of paraspinal muscles (erector spinae (ES), multifidus (MS), and psoas (PS)) were measured at (ES L1/2-L4/5; MS L2/3-L5/S1; PS L2/3-L5/S1) levels using dual-energy computed tomography (DECT). Quantitative computed tomography (QCT) was used to assess BMD of L1 and L2. Linear regression analysis was used to assess the relationship between BMD of the lumbar spine and paraspinal muscles fat content with age, sex, and BMI. The variance inflation factor (VIF) was used to detect the degree of multicollinearity among the variables. P < .05 was considered to indicate a statistically significant difference. Results: The paraspinal muscles fat content had a fairly significant inverse association with lumbar BMD after controlling for age, sex, and BMI (adjusted R 2 = 0.584-0.630, all P < .05). Conclusion: Paraspinal muscles fat content was negatively associated with BMD.Paraspinal muscles fatty infiltration may be considered as a potential marker to identify BMD loss.

Fast kilovoltage (KV)-switching dual-energy CT virtual noncalcium technique using different reconstruction kernels for identifying acute and chronic vertebral compression fractures.

The vertebral compression fractures (VCFs) represent an incidental finding on thoracic and abdominal dual-energy CT examinations (which use STND reconstruction kernel), which are associated with increased mortality. While the BONE reconstruction kernel shows a superior diagnostic accuracy to find fractures. This study showed STND and BONE reconstruction kernel both had excellent diagnostic performance to detect abnormal edema in acute VCFs. PURPOSE: To investigate whether different reconstruction kernels (STND V.S. BONE) affect the diagnostic performance of dual-energy CT virtual noncalcium technique (VNCa) for identifying acute and chronic vertebral compression fractures (VCFs). METHODS: This retrospective study included 31 consecutive patients with 79 VCFs who underwent both a dual-energy CT and a 3-T MR examination of the spine between August 2018 and March 2019. MR images served as the reference standard. Two independent and blinded radiologists evaluated all vertebral bodies for the presence of abnormal edema on color-coded overlay VNCa images. Two additional radiologists performed a quantitative analysis on VNCa images by calculating water content of vertebral bodies. Receiver operating characteristic curve (ROC) analysis was conducted. Area under the curve (AUC) was calculated. RESULTS: MR imaging depicted 44 edematous and 35 nonedematous VCFs. In visual analysis, the AUCSTND and AUCBONE were 0.932 and 0.943. In quantitative analysis, water content results were significantly different between vertebrae with and without bone marrow edema on MR (P < 0.001). And the AUCSTND and AUCBONE were 0.851 and 0.850 respectively. CONCLUSION: Visual and quantitative analysis of dual-energy CT VNCa technique had excellent diagnostic performance for identifying acute and chronic compression fractures; different reconstruction kernels did not matter.

In vivo quantification of bone mineral density of lumbar vertebrae using fast kVp switching dual-energy CT: correlation with quantitative computed tomography.

BACKGROUND: Osteoporosis is a common, progressive disease related to low bone mineral density (BMD). If it can be diagnosed at an early stage, osteoporosis is treatable. Quantitative computed tomography (QCT) is one of the current reference standards of BMD measurement, but dual-energy computed tomography (DECT) is considered to be a potential alternative. This study aimed to evaluate the feasibility and accuracy of phantomless in vivo DECT-based BMD quantification in comparison with QCT. METHODS: A total of 128 consecutive participants who underwent DECT lumbar examinations between July 2018 and February 2019 were retrospectively analyzed. The density of calcium (water), hydroxyapatite (water), calcium (fat), and hydroxyapatite (fat) [DCa(Wa), DHAP(Wa), DCa(Fat) and DHAP(Fat), respectively] were measured along with BMD in the trabecular bone of lumbar level 1-2 by DECT and QCT. Linear regression analysis was performed to assess the relationship between DECT- and QCT-derived BMD at both the participant level and the vertebral level. Linear regression models were quantitatively evaluated with adjusted R-square, normalized mean squared error (NMSE) and relative error (RE). Bland-Altman analysis was conducted to assess agreement between measurements. P<0.05 was considered statistically significant. RESULTS: Strong correlations were observed between DECT- and QCT-derived BMD at both the participant level and the vertebral level (adjusted R2 =0.983-0.987; NMSE = 1.6-2.1%; RElinear =0.6-0.9%). Bland-Altman plots indicated high agreement between both measurements. DCa(Fat) and DHAP(Fat) showed relatively similar and optimal predictive capability for QCT-derived BMD (both: adjusted R2 =0.987, NMSE =1.6%, RElinear =0.6%). CONCLUSIONS: Fast kVp switching DECT enabled accurate phantomless in vivo BMD quantification of the lumbar spine. DCa(Fat) and DHAP(Fat) had relatively similar and optimal predictive capability.

Fast kilovoltage (KV)-switching dual-energy computed tomography hydroxyapatite (HAP)-water decomposition technique for identifying bone marrow edema in vertebral compression fractures.

BACKGROUND: Hydroxyapatite (HAP) is the main component of bone mineral. The utility of using HAP-water decomposition technique with fast kilovoltage (KV)-switching dual-energy computed tomography (DECT) to detect abnormal edema in vertebral compression fractures (VCFs) has not been widely reported. METHODS: A total of 31 consecutive patients with 80 VCFs who underwent DECT and magnetic resonance imaging (MRI) of the spine were retrospectively enrolled in our study between October 2018 and January 2019. VCFs in MR examinations served as the standard of reference. Two radiologists blindly and independently evaluated color-coded overlay virtual nonhydroxyapatite (VNHAP) images for the presence of abnormal edema. The inter-reader agreement, specificity, sensitivity, accuracy, and predictive values of VNHAP images for edema detection were calculated. The diagnostic accuracy of two readers was compared using McNemar's test. Two additional radiologists performed a quantitative analysis on VNHAP images, receiver operating characteristic (ROC) curve analysis was conducted, and the threshold was calculated. RESULTS: MRI depicted 45 edematous and 35 nonedematous VCFs. For visual analysis, the VNHAP technique showed a sensitivity of 93.3%, a specificity of 97.1%, a positive predictive value (PPV) of 97.7%, a negative predictive value (NPV) of 91.9%, and an accuracy of 95.0%. The inter-reader agreement was almost perfect (k=0.90). The diagnostic accuracy of the two readers showed no significant differences in the assessment of VNHAP images (P=1.00). Significant differences in CT numbers between vertebrae with and without bone marrow edema were found by quantitative analysis (P<0.01). The area under the curve (AUC) of the VNHAP images was estimated to be 0.917. The threshold of 1,003.2 mg/cm3 yielded a sensitivity of 88.9% and a specificity of 82.9% for the differentiation of fresh and old VCFs. CONCLUSIONS: Fast KV-switching DECT HAP-water decomposition technique had excellent diagnostic performance for identifying acute and chronic VCFs in visual and quantitative analyses.