Performance of statistical models of shape and appearance for semiautomatic segmentations of spinal vertebrae T4-L4 on digitized vertebral fracture assessment images

Statistical Shape and Appearance Models: Development Towards Improved Osteoporosis Care

PURPOSE OF REVIEW

Statistical models of shape and appearance have increased their popularity since the 1990s and are today highly prevalent in the field of medical image analysis. In this article, we review the recent literature about how statistical models have been applied in the context of osteoporosis and fracture risk estimation.

RECENT FINDINGS

Recent developments have increased their ability to accurately segment bones, as well as to perform 3D reconstruction and classify bone anatomies, all features of high interest in the field of osteoporosis and fragility fractures diagnosis, prevention, and treatment. An increasing number of studies used statistical models to estimate fracture risk in retrospective case-control cohorts, which is a promising step towards future clinical application. All the reviewed application areas made considerable steps forward in the past 5-6 years. Heterogeneities in validation hinder a thorough comparison between the different methods and represent one of the future challenges to be addressed to reach clinical implementation.

A deep-learning model for identifying fresh vertebral compression fractures on digital radiography

OBJECTIVES

To develop a deep-learning (DL) model for identifying fresh VCFs from digital radiography (DR), with magnetic resonance imaging (MRI) as the reference standard.

METHODS

Patients with lumbar VCFs were retrospectively enrolled from January 2011 to May 2020. All patients underwent DR and MRI scanning. VCFs were categorized as fresh or old according to MRI results, and the VCF grade and type were assessed. The raw DR data were sent to InferScholar Center for annotation. A DL-based prediction model was built, and its diagnostic performance was evaluated. The DeLong test was applied to assess differences in ROC curves between different models.

RESULTS

A total of 1877 VCFs in 1099 patients were included in our study and randomly divided into development (n = 824 patients) and test (n = 275 patients) datasets. The ensemble model identified fresh and old VCFs, reaching an AUC of 0.80 (95% confidence interval [CI], 0.77-0.83), an accuracy of 74% (95% CI, 72-77%), a sensitivity of 80% (95% CI, 77-83%), and a specificity of 68% (95% CI, 63-72%). Lateral (AUC, 0.83) views exhibited better performance than anteroposterior views (AUC, 0.77), and the best performance among respective subgroupings was obtained for grade 3 (AUC, 0.89) and crush-type (AUC, 0.87) subgroups.

CONCLUSION

The proposed DL model achieved adequate performance in identifying fresh VCFs from DR.

KEY POINTS

• The ensemble deep-learning model identified fresh VCFs from DR, reaching an AUC of 0.80, an accuracy of 74%, a sensitivity of 80%, and a specificity of 68% with the reference standard of MRI. • The lateral views (AUC, 0.83) exhibited better performance than anteroposterior views (AUC, 0.77). • The grade 3 (AUC, 0.89) and crush-type (AUC, 0.87) subgroups showed the best performance among their respective subgroupings.

Associations between bone attenuation and prevalent vertebral fractures on chest CT scans differ with vertebral fracture locations

Vertebral fracture (VF) locations are bimodally distributed in the spine. The association between VF and bone attenuation (BA) measured on chest CT scans varied according to the location of VFs, indicating that other factors than only BA play a role in the bimodal distribution of VFs.

INTRODUCTION

Vertebral fractures (VFs) are associated with low bone mineral density but are not equally distributed throughout the spine and occur most commonly at T7-T8 and T11-T12 ("cVFs") and less commonly at T4-T6 and T9-T10 ("lcVF"). We aimed to determine whether associations between bone attenuation (BA) and VFs vary between subjects with cVFs only, with lcVFs only and with both cVFs and lcVFs.

METHODS

Chest CT images of T4-T12 in 1237 smokers with and without COPD were analysed for prevalent VFs according to the method described by Genant (11,133 vertebrae). BA (expressed in Hounsfield units) was measured in all non-fractured vertebrae (available for 10,489 vertebrae). Linear regression was used to compare mean BA, and logistic regression was used to estimate the association of BA with prevalent VFs (adjusted for age and sex).

RESULTS

On vertebral level, the proportion of cVFs was significantly higher than of lcVF (5.6% vs 2.0%). Compared to subjects without VFs, BA was 15% lower in subjects with cVFs (p < 0.0001), 25% lower in subjects with lcVFs (p < 0.0001) and lowest in subjects with cVFs and lcVFs (- 32%, p < 0.0001). The highest ORs for presence of VFs per - 1SD BA per vertebra were found in subjects with both cVFs and lcVFs (3.8 to 4.6).

CONCLUSIONS

The association between VFs and BA differed according to VF location. ORs increased from subjects with cVFs to subjects with lcVFs and were highest in subjects with cVFs and lcVFs, indicating that other factors than only BA play a role in the bimodal VF distribution.

TRIAL REGISTRATION

Clinicaltrials.gov identifier: NCT00292552.

Current Applications and Selected Technical Details of Dual-Energy X-Ray Absorptiometry

The application of dual-energy X-ray absorptiometry (DXA) examinations in the assessment of bone mineral density (BMD) in the lumbar spine, hip, and forearm is the basic diagnostic method for recognition of osteoporosis. The constant development of DXA technique is due to the aging of societies and the increasing importance of osteoporosis as a public health problem. In order to assess the degree of bone demineralization in patients with hyperparathyroidism, forearm DXA examination is recommended. The vertebral fracture assessment (VFA) of the thoracic and lumbar spine, performed by a highly-skilled technician, is an interesting alternative to the X-ray examination. The DXA total body examination can be useful in the evaluation of fat redistribution among patients after bariatric surgery, in patients infected with HIV and receiving antiretroviral therapy, and in patients with metabolic diseases and suspected to have sarcopenia. The assessment of visceral adipose tissue (VAT) and detection of abdominal aortic calcifications may be useful in the prediction of cardiovascular events. The positive effect of anti-resorptive therapy may affect some parameters of DXA hip structure analysis (HSA). Long-term anti-resorptive therapy, especially with the use of bisphosphonates, may result in changes in the DXA image, which may herald atypical femur fractures (AFF). Reduction of the periprosthetic BMD in the DXA measurements can be used to estimate the likelihood of loosening the prosthesis and periprosthetic fractures. The present review aims to present current applications and selected technical details of DXA.

Axial-SpineGAN: simultaneous segmentation and diagnosis of multiple spinal structures on axial magnetic resonance imaging images

Providing a simultaneous segmentation and diagnosis of the spinal structures on axial magnetic resonance imaging (MRI) images has significant value for subsequent pathological analyses and clinical treatments. However, this task remains challenging, owing to the significant structural diversity, subtle differences between normal and abnormal structures, implicit borders, and insufficient training data. In this study, we propose an innovative network framework called 'Axial-SpineGAN' comprising a generator, discriminator, and diagnostor, aiming to address the above challenges, and to achieve simultaneous segmentation and disease diagnosis for discs, neural foramens, thecal sacs, and posterior arches on axial MRI images. The generator employs an enhancing feature fusion module to generate discriminative features, i.e. to address the challenges regarding the significant structural diversity and subtle differences between normal and abnormal structures. An enhancing border alignment module is employed to obtain an accurate pixel classification of the implicit borders. The discriminator employs an adversarial learning module to effectively strengthen the higher-order spatial consistency, and to avoid overfitting owing to insufficient training data. The diagnostor employs an automated diagnosis module to provide automated recognition of spinal diseases. Extensive experiments demonstrate that these modules have positive effects on improving the segmentation and diagnosis accuracies. Additionally, the results indicate that Axial-SpineGAN has the highest Dice similarity coefficient (94.9% ± 1.8%) in terms of the segmentation accuracy and highest accuracy rate (93.9% ± 2.6%) in terms of the diagnosis accuracy, thereby outperforming existing state-of-the-art methods. Therefore, our proposed Axial-SpineGAN is effective and potential as a clinical tool for providing an automated segmentation and disease diagnosis for multiple spinal structures on MRI images.

Association between vertebral fractures and coronary artery calcification in current and former smokers in the ECLIPSE cohort

In smokers and former smokers from the ECLIPSE cohort, there is an association between prevalent vertebral fractures (VFs) and coronary artery calcification (CAC). Chest CT scans provide the opportunity to evaluate VFs and CAC, which are potentially important comorbidities, each of which is amenable to effective interventions.

INTRODUCTION

Prevalence of VFs among smokers and patients with chronic obstructive pulmonary disease (COPD) is high, and an association between CAC and osteoporosis has been described. We investigated the associations between VFs and CAC (expressed in Agatston score) in (former) smokers.

METHODS

Current and former smokers from the ECLIPSE study (designed to determine underlying COPD progression mechanisms) were studied. Baseline Agatston score (zero (0), medium (1-400), or high (> 400)), baseline bone attenuation (BA), and prevalent and incident VFs (vertebrae T₁-L₁) were assessed on CT.

RESULTS

A total of 586 subjects were included (mean age 59.8 ± 8.3; 62.3% men; 70.1% with COPD; 21.0% with prevalent VFs; 196 with zero, 266 with medium, and 124 with high Agatston score). Of these, 23.4% suffered incident VFs within 3 years. In multivariate models, prevalent VFs were associated with medium (1.83 [95% CI 1.01-3.30]) and with high (OR = 3.06 [1.45-6.47]) Agatston score. After adjustment for BA, prevalent VFs were still associated with high (OR = 2.47 [1.13-5.40]), but not significantly with medium Agatston score (OR = 1.57 [0.85-2.88]). Similarly, after adjustment for BA, high (OR = 2.06 [1.02-4.13]) but not medium Agatston score (OR = 1.61 [0.88-2.94]) was associated with prevalent VFs. Agatston score at baseline was not associated with short-term VF incidence.

CONCLUSION

In (former) smokers, there was an association between prevalent VFs and Agatston score. Chest CT scans provide the opportunity to also evaluate for VFs and CAC, which are potentially important comorbidities, each of which is amenable to effective interventions.

Vertebral bone attenuation in Hounsfield Units and prevalent vertebral fractures are associated with the short-term risk of vertebral fractures in current and ex-smokers with and without COPD: a 3-year chest CT follow-up study

CT scans performed to evaluate chronic obstructive pulmonary disease (COPD) also enable evaluation of bone attenuation (BA; a measure of bone density) and vertebral fractures (VFs). In 1239 current/former smokers with (n = 999) and without (n = 240) COPD, the combination of BA and prevalent VFs was associated with the incident VF risk.

INTRODUCTION

Chest CT scans are increasingly used to evaluate pulmonary diseases, including COPD. COPD patients have increased risk of osteoporosis and VFs. BA on CT scans is correlated with bone mineral density and prevalent VFs. The aim of this study was to evaluate the association between BA and prevalent VFs on chest CT scans, and the risk of incident VFs in current and former smokers with and without COPD.

METHODS

In participants of the ECLIPSE study with baseline and 1-year and 3-year follow-up CT scans, we evaluated BA in vertebrae T₄-T₁₂ and prevalent and incident VFs.

RESULTS

A total of 1239 subjects were included (mean age 61.3 ± 8.0, 61.1% men, 999 (80.6%) COPD patients). The mean BA was 155.6 ± 47.5 Hounsfield Units (HU); 253 (20.5%) had a prevalent VF and 296 (23.9%) sustained an incident VF within 3 years. BA and prevalent VFs were associated with incident VFs within 1 (per - 1SD HR = 1.38 [1.08-1.76] and HR = 3.97 [2.65-5.93] resp.) and 3 years (per - 1SD HR = 1.25 [1.08-1.45] and HR = 3.10 [2.41-3.99] resp.), while age, sex, body mass index (BMI), smoking status and history, or presence of COPD was not. In subjects without prevalent VFs and BA, and for 1-year incidence, BMI values were associated with incident fractures (1 year, BA per - 1SD HR = 1.52 [1.05-2.19], BMI per SD HR = 1.54 [1.13-2.11]; 3 years, per - 1SD HR = 1.37 [1.12-1.68]).

CONCLUSIONS

On CT scans performed for pulmonary evaluation in (former) smokers with and without COPD, the combination of BA and prevalent VFs was strongly associated with the short-term risk of incident VFs.

Thoracic Kyphosis on Chest CT Scans Is Associated With Incident Vertebral Fractures in Smokers

Greater kyphosis angles lead to increased loading on vertebral bodies in computational models. However, results about the relationship between severity of kyphosis and incident vertebral fracture (VF) risk have been conflicting. Therefore, the aim of this study was to evaluate associations between 1) prevalent VFs and severity of kyphosis, and 2) severity of kyphosis and incident VF risk in smokers with or without chronic obstructive pulmonary disease (COPD). Former and current smokers with or without COPD were included. CT scans were made at baseline, 1-year, and 3-year follow-up. VFs were evaluated on superposed sagittal CT reconstructions. Kyphosis was measured as the angle between the lines above T₄ and below T₉ or T₁₂ . We included 1239 subjects (mean age 61.3 ± 8.0 years, 61.1% male, 80.6% with COPD), of whom 253 (20.4%) had a prevalent VF and 294 (23.7%) an incident VF within 3 years. Presence, number, and severity of prevalent VFs were associated with a greater kyphosis angle. The mean increase in kyphosis angle within 3 years was small but significantly greater in subjects with incident VFs compared with those without (2.2 ± 4.1 versus 1.2 ± 3.9 degrees, respectively, for T₄ to T₁₂ angle, p < 0.001). After adjustment for bone attenuation (BA) and prevalent VFs, baseline kyphosis angle was associated with incident VFs within 1 and 3 years (angle T₄ to T₁₂ per +1 SD, hazard ratio [HR] = 1.34 [1.12-1.61] and HR 1.29 [1.15-1.45], respectively). Our data showed that a greater kyphosis angle at baseline was independently associated with increased risk of incident VFs within 1 and 3 years, supporting the theory that greater kyphosis angle contributes to higher biomechanical loads in the spine. © 2019 American Society for Bone and Mineral Research.

Automated quantitative morphometry of vertebral heights on spinal radiographs: comparison of a clinical workflow tool with standard 6-point morphometry

A workflow tool for measurements of vertebral heights on lateral spine radiographs based on automated placements of 6 points per vertebra was evaluated. The tool helps to standardize point placement among operators. Its success rate is very good in normal vertebrae but lower in vertebrae with more severe fractures. Manual corrections were required in 192 of 1257 analyzed vertebrae.

INTRODUCTION

To evaluate a new workflow tool (SA) for the automated measurements of vertebral heights on lateral spine radiographs.

METHODOLOGY

Lateral radiographs from 200 postmenopausal women were evaluated at two visits. Genant's semi-quantitative fracture assessment (SQ) and manual quantitative morphometry (QM) results were available from prior analyses. Vertebral heights from point placements using SA were compared with manual 6-point placement QM. Differences were quantified as RMS coefficient of variations (rmsCV) and standard deviations (rmsSD).

RESULTS AND CONCLUSIONS

SA required manual corrections in 192 of 1257 vertebrae. SA heights were larger than QM ones by 2.2-3.6%. Correlations (r² > 0.92) between SA and QM were very high. Differences between QM and SA were higher for fractured (SQ = 2; rmsCV% 14.5%) than for unfractured vertebrae (rmsCV% 4.2-4.7%). rmsCV% for QM varied between 3 and 6% and for SA between 2.5 and 7.5%. For SA, highest rmsCV% was obtained for T4 and L4. Manual correction mostly affected the end vertebrae T4 and L4. SA helps to standardize point placement among operators. The algorithm success rate is very good in normal vertebrae but lower in vertebrae with more severe fractures, which are of greater clinical interest but are more readily recognized without morphometric measurements.

High Imminent Vertebral Fracture Risk in Subjects With COPD With a Prevalent or Incident Vertebral Fracture

Subjects with chronic obstructive pulmonary disease (COPD) have an increased risk of vertebral fractures (VFs); however, VF incidence is largely unknown. Therefore, the aim of our study was to determine the incidence of new and/or worsening VF in subjects with COPD. Smokers and subjects with COPD (GOLD II-IV) from the ECLIPSE study with complete set of chest CT scans (baseline and 1- and 3-year follow-up) to evaluate vertebrae T₁ down to L₁ were included. If a VF was diagnosed on the last scan, detailed VF assessment of the previous scans was performed. VFs were scored according to the method of Genant as mild, moderate, or severe. Main outcome measure was the cumulative incidence of new and/or worsening VF at subject level, within 1 and 3 years. Of 1239 subjects (mean age 61 years, 757 males [61%], 999 subjects with COPD), 253 (20.5%) had ≥1 prevalent VF. The cumulative incidence of VFs was 10.1% within 1 year and 24.0% within 3 years. After adjustment for age, sex, body mass index (BMI), pack-years, and smoking status, prevalence and incidence were similar between smokers and COPD GOLD stages. Within 1 year, 29.2% of the subjects with a prevalent VF had an incident VF, compared with 5.1% in absence of prevalent VF (hazard ratio [HR] = 5.1; 95% confidence interval [CI] 3.6-7.4) and 58.5% versus 15.0% within 3 years (HR = 3.6; 95% CI 2.9-4.6). The incidence of VF was higher with increasing number and severity of prevalent VFs. Among subjects having an incident VF within the first year, 57.3% had a subsequent VF within the next 2 years. In this study, more than half of the smokers and subjects with COPD with a prevalent VF or an incident VF within the first year sustained a subsequent VF within 3 years. The 3-year risk was even higher in the presence of multiple or severe prevalent VFs. © 2018 The Authors. Journal of Bone and Mineral Research Published by Wiley Periodicals Inc.

Diagnosis of vertebral deformities on chest CT and DXA compared to routine lateral thoracic spine X-ray

X-ray, CT and DXA enable diagnosis of vertebral deformities. For this study, level of agreement of vertebral deformity diagnosis was analysed. We showed that especially on subject level, these imaging techniques could be used for opportunistic screening of vertebral deformities in COPD patients.

INTRODUCTION

X-ray and CT are frequently used for pulmonary evaluation in patients with chronic obstructive pulmonary disease (COPD) and also enable to diagnose vertebral deformities together with dual-energy X-ray absorptiometry (DXA) imaging. The aim of this research was to study the level of agreement of these imaging modalities for diagnosis of vertebral deformities from T4 to L1.

METHODS

Eighty-seven subjects (mean age of 65; 50 males; 57 COPD patients) who had X-ray, chest CT (CCT) and DXA were included. Evaluable vertebrae were scored twice using SpineAnalyzer™ software. ICCs and kappas were calculated to examine intra-observer variability. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and area under the receiver operating characteristic curve (AUROC) were calculated to compare vertebral deformities diagnosed on the different imaging modalities.

RESULTS

ICCs for height measurements were excellent (> 0.94). Kappas were good to excellent (0.64-0.77). At vertebral level, the AUROC was 0.85 for CCT vs. X-ray, 0.74 for DXA vs. X-ray and 0.77 for DXA vs. CCT. Sensitivity (51%-73%) and PPV (57%-70%) were fair to good; specificity and NPV were excellent (≥ 96%). At subject level, the AUROC values were comparable.

CONCLUSIONS

Reproducibility of height measurements of vertebrae is excellent with all three imaging modalities. On subject level, diagnostic performance of CT (PPV 79-82%; NPV 90-93%), and to a slightly lesser extend of DXA (PPV 73-77%; NPV 80-89%), indicates that these imaging techniques could be used for opportunistic screening of vertebral deformities in COPD patients.

Effect of implementation of guidelines on assessment and diagnosis of vertebral fractures in patients older than 50 years with a recent non-vertebral fracture

We evaluated the impact of a new Dutch guideline on systematic implementation of densitometric Vertebral Fracture Assessment (VFA) in patients with a recent non-vertebral fracture. Systematic implementation resulted in a significant increase of VFA, diagnosis of vertebral fractures (VFs), and percentage of patients eligible for treatment.

INTRODUCTION

VFs are underdiagnosed although they are important predictors of fracture risk, independent of age and bone mineral density (BMD). The Dutch guideline on osteoporosis and fracture prevention recommends VFA in all patients aged >50 years with a recent non-VF. Our aim was to evaluate the effect of systematic implementation of densitometric VFA in patients with a recent non-VF at the fracture liaison service (FLS).

METHODS

VFA was performed on lateral images of the spine using dual-energy X-ray absorptiometry (DXA) and graded according to Genant using Spine Analyzer software.

RESULTS

We evaluated 582 patients before and 484 after implementation (mean age 67 and 66 years; 71 and 74% women, respectively). Performing VFA increased from 4.6 to 97.1% (p < 0.001) and the diagnosis of VFs from 2.2 to 26.2% for grade ≥ 1 (p < 0.001) and from 0.9 to 14.7% for grade ≥ 2 (p < 0.001). Prevalence of VFs increased with age (5.2% in 50-59-year olds to 27.8% in 80+-year olds, p < 0.001), but was similar for both genders, non-VF locations, and BMD. Including patients with osteopenia and a VF increased the percentage of patients eligible for treatment by a quarter, from 31.0% in the pre-guideline to 38.4% in the post-guideline cohort.

CONCLUSIONS

Systematic guideline implementation resulted in a significant increase of VFA, diagnosis of VFs, and percentage of patients eligible for treatment. VFA contributes to documenting the high prevalence of VFs in patients visiting the FLS with a non-VF in both genders, at any age, non-VF location, and BMD.