Vertebral Body Compression Fractures and Bone Density: Automated Detection and Classification on CT Images

Population-based opportunistic osteoporosis screening: Validation of a fully automated CT tool for assessing longitudinal BMD changes

METHODS:

The automated BMD tool was retrospectively applied to non-contrast abdominal CT scans in 1603 consecutive asymptomatic adults (mean age, 55.9 years; 770 M/833 F) undergoing longitudinal screening (mean interval, 5.7 years; range, 1.0-12.3 years). The spinal column was automatically segmented, with standardized L1 and L2 anterior trabecular ROI placement. Automated and manual L1 HU values were compared, as were automated supine-prone measures. L1-L2 CT attenuation values were converted to BMD values through a linear regression model. BMD values and changes were assessed according to age and gender.

RESULTS:

Success rate of the automated BMD tool was 99.8 % (four failed cases). Both automated supine vs prone and manual vs automated L1 attenuation measurements showed good agreement. Overall mean annual rate of bone loss was greater in females than males (-2.0% vs -1.0%), but the age-specific rate declined faster in females from age 50 (-2.1%) to age 65 (-0.3%) compared with males (-0.9% to -0.5%). Mean BMD was higher in females than males at age 50 (143.6 vs 135.1  mg cm^-3), but post-menopausal bone loss in females reversed this relationship beyond age 60. By age 70, mean BMD in females and males was 100.8 and 107.7  mg cm^-3 , respectively.

CONCLUSION:

This robust, fully automated CT BMD tool allows for both individualized and population-based assessment. Mean BMD was lower in men than women aged 50-60, but accelerated post-menopausal bone loss in women resulted in lower values beyond age 60.

ADVANCES IN KNOWLEDGE:

This fully automated tool can be applied to routine abdominal CT scans for prospective or retrospective opportunistic BMD assessment, including change over time. Mean BMD was lower in men compared with women aged 50-60 years, but accelerated bone loss in women during this early post-menopausal period resulted in lower BMD values for women beyond age 60.

Predictive Factors for and Detection of Lateral Hinge Fractures Following Open Wedge High Tibial Osteotomy: Plain Radiography Versus Computed Tomography

PURPOSE

To compare the accuracy of plain radiography and computed tomography (CT) in detecting lateral hinge fractures and to identify predictive factors of the lateral hinge fractures after open wedge high tibial osteotomy (HTO).

METHODS

This study included patients who underwent isolated primary open wedge HTO for medial compartment degenerative arthritis from 2015 to 2016 and were evaluated by postoperative CT scans and followed for a minimum of 1 year. This study involved 48 patients (48 knees) undergoing open wedge HTO for primary medial osteoarthritis. Lateral hinge fractures were evaluated on immediate plain radiographs and CT scans according to the Takeuchi classification. Predictive factors of plain radiograph-based hinge fracture and CT-based hinge fracture were investigated.

RESULTS

The detection rate of lateral hinge fracture was significantly higher on CT scans than on plain radiographs (50% vs 14.6%; P < .001). Seven knees (14.6%) showed lateral hinge fractures on postoperative plain radiographs, including 6 with type I and 1 with type III fractures; 24 (50.0%) showed lateral hinge fractures on postoperative CT scans, including 20 with type I, 2 with type II, and 2 with type III fractures. CT scans detected 14 type I, 2 type II, and 1 type III hinge fractures not detected by plain radiographs. The medial opening gap width was the only predictor of lateral hinge fracture on CT scans (adjusted odds ratio, 1.565; P =.034).

CONCLUSIONS

The incidence of lateral hinge fracture of the knee after open wedge HTO was higher on CT scans than on plain radiographs. The medial opening gap width was the only predictor of lateral hinge fractures, suggesting a need for care when spreading the gap to avoid lateral hinge fractures.

LEVEL OF EVIDENCE

Level III, cross-sectional study.

Multiparametric ultrasomics of significant liver fibrosis: A machine learning-based analysis

Objective

To assess significant liver fibrosis by multiparametric ultrasomics data using machine learning.

Materials and Methods

This prospective study consisted of 144 patients with chronic hepatitis B. Ultrasomics—high-throughput quantitative data from ultrasound imaging of liver fibrosis—were generated using conventional radiomics, original radiofrequency (ORF) and contrast-enhanced micro-flow (CEMF) features. Three categories of features were explored using pairwise correlation and hierarchical clustering. Features were selected using diagnostic tests for fibrosis, activity and steatosis stage, with the histopathological results as the reference. The fibrosis staging performance of ultrasomics models with combinations of the selected features was evaluated with machine-learning algorithms by calculating the area under the receiver-operator characteristic curve (AUC).

Results

ORF and CEMF features had better predictive power than conventional radiomics for liver fibrosis stage (both p < 0.01). CEMF features exhibited the highest diagnostic value for activity stage (both p < 0.05), and ORF had the best diagnostic value for steatosis stage (both p < 0.01). The machine-learning classifiers of adaptive boosting, random forest and support vector machine were found to be optimal algorithms with better (all mean AUCs = 0.85) and more stable performance (coefficient of variation = 0.01–0.02) for fibrosis staging than decision tree, logistic regression and neural network (mean AUC = 0.61–0.72, CV = 0.07–0.08). The multiparametric ultrasomics model achieved much better performance (mean AUC values of 0.78–0.85) than the features from a single modality in discriminating significant fibrosis (≥ F2).

Conclusion

Machine-learning-based analysis of multiparametric ultrasomics can help improve the discrimination of significant fibrosis compared with mono or dual modalities.

Key Points

• Multiparametric ultrasomics has achieved much better performance in the discrimination of significant fibrosis (≥ F2) than the single modality of conventional radiomics, original radiofrequency and contrast-enhanced micro-flow.

• Adaptive boosting, random forest and support vector machine are the optimal algorithms for machine learning.

Electronic supplementary material

The online version of this article (10.1007/s00330-018-5680-z) contains supplementary material, which is available to authorized users.

Deep neural networks for automatic detection of osteoporotic vertebral fractures on CT scans

Osteoporotic vertebral fractures (OVFs) are prevalent in older adults and are associated with substantial personal suffering and socio-economic burden. Early diagnosis and treatment of OVFs are critical to prevent further fractures and morbidity. However, OVFs are often under-diagnosed and under-reported in computed tomography (CT) exams as they can be asymptomatic at an early stage. In this paper, we present and evaluate an automatic system that can detect incidental OVFs in chest, abdomen, and pelvis CT examinations at the level of practicing radiologists. Our OVF detection system leverages a deep convolutional neural network (CNN) to extract radiological features from each slice in a CT scan. These extracted features are processed through a feature aggregation module to make the final diagnosis for the full CT scan. In this work, we explored different methods for this feature aggregation, including the use of a long short-term memory (LSTM) network. We trained and evaluated our system on 1432 CT scans, comprised of 10,546 two-dimensional (2D) images in sagittal view. Our system achieved an accuracy of 89.2% and an F1 score of 90.8% based on our evaluation on a held-out test set of 129 CT scans, which were established as reference standards through standard semiquantitative and quantitative methods. The results of our system matched the performance of practicing radiologists on this test set in real-world clinical circumstances. We expect the proposed system will assist and improve OVF diagnosis in clinical settings by pre-screening routine CT examinations and flagging suspicious cases prior to review by radiologists.