Automated Segmentation of Trunk Musculature with a Deep CNN Trained from Sparse Annotations in Radiation Therapy Patients with Metastatic Spine Disease.

PURPOSE: Given the high prevalence of vertebral fractures post-radiotherapy in patients with metastatic spine disease, accurate and rapid muscle segmentation could support efforts to quantify muscular changes due to disease or treatment and enable biomechanical modeling for assessments of vertebral loading to improve personalized evaluation of vertebral fracture risk. This study presents a deep-learning approach for segmenting the complete volume of the trunk muscles from clinical CT images trained using sparsely annotated data.

MATERIALS AND METHODS: we extracted 2,009 axial CT images at the midpoint of each vertebral level (T4 to L4) from clinical CT of 148 cancer patients. The key extensor and flexor muscles (up to 8 muscles per side) were manually contoured and labeled per image in the thoracic and lumbar regions. We first trained a 2D nnU-Net deep-learning model on these labels to segment key extensor and flexor muscles. Using these sparse annotations per spine, we trained the model to segment each muscle's entire 3D volume.

RESULTS: The proposed method achieved comparable performance to manual segmentations, as assessed by expert radiologists, with a mean Dice score above 0.769. Significantly, the model drastically reduced segmentation time, from 4.3-6.5 hours for manual segmentation of 14 single axial CT images to approximately 1 minute for segmenting the complete thoracic-abdominal 3D volume.

CONCLUSION: The approach demonstrates high potential for automating 3D muscle segmentation, significantly reducing the manual intervention required for generating musculoskeletal models, and could be instrumental in enhancing clinical decision-making and patient care in radiation oncology.

Pathologic vertebral fractures (PVF) are common and serious complications in patients with metastatic lesions affecting the spine. Accurate assessment of cancer patients' PVF risk is an unmet clinical need. Load-to-strength ratios (LSRs) evaluated in vivo by estimating vertebral loading from biomechanical modeling and strength from computed tomography imaging (CT) have been associated with osteoporotic vertebral fractures in older adults. Here, for the first time, we investigate LSRs of thoracic and lumbar vertebrae of 135 spine metastases patients compared to LSRs of 246 healthy adults, comparable by age and sex, from the Framingham Heart Study under four loading tasks. Findings include: (1) Osteolytic vertebrae have higher LSRs than osteosclerotic and mixed vertebrae; (2). In patients' vertebrae without CT observed metastases, LSRs were greater than healthy controls. (3) LSRs depend on the spinal region (Thoracic, Thoracolumbar, Lumbar). These findings suggest that LSRs may contribute to identifying patients at risk of incident PVF in metastatic spine disease patients. The lesion-mediated difference suggests that risk thresholds should be established based on spinal region, simulated task, and metastatic lesion type.

Automatic analysis of pathologic vertebrae from computed tomography (CT) scans could significantly improve the diagnostic management of patients with metastatic spine disease. We provide the first publicly available annotated imaging dataset of cancerous CT spines to help develop artificial intelligence frameworks for automatic vertebrae segmentation and classification. This collection contains a dataset of 55 CT scans collected on patients with various types of primary cancers at two different institutions. In addition to raw images, data include manual segmentations and contours, vertebral level labeling, vertebral lesion-type classifications, and patient demographic details. Our automated segmentation model uses nnU-Net, a freely available open-source framework for deep learning in healthcare imaging, and is made publicly available. This data will facilitate the development and validation of models for predicting the mechanical response to loading and the resulting risk of fractures and spinal deformity.

Introduction: Pathologic vertebral fractures are devastating for patients with spinal metastases. However, the mechanical process underlying these fractures is poorly understood, limiting physician's ability to predict which vertebral bodies will fail. Method: Here, we show the development of a damage-based finite element framework producing highly reliable pathologic vertebral strength and stiffness predictions from X-Ray computed tomography (CT) data. We evaluated the performance of specimen-specific material calibration vs. global material calibration across osteosclerotic, osteolytic, and mixed lesion vertebrae that we derived using a machine learning approach. Results: The FE framework using global calibration strongly predicted the pathologic vertebrae stiffness (R 2 = 0.90, p < 0.0001) and strength (R 2 = 0.83, p = 0.0002) despite the remarkable variance in the pathologic bone structure and density. Specimen-specific calibration produced a near-perfect prediction of both stiffness and strength (R 2 = 0.99, p < 0.0001, for both), validating the FE approach. The FE damage-based simulations highlighted the differences in the pattern of spatial damage evolution between osteosclerotic and osteolytic vertebral bodies. Discussion: With failure, the FE simulation suggested a common damage evolution pathway progressing largely localized to the low bone modulus regions within the vertebral volume. Applying this FE approach may allow us to predict the onset and anatomical location of vertebral failure, which is critical for developing image-based diagnostics of impending pathologic vertebral fractures.

Accurate and reliable registration of longitudinal spine images is essential for assessment of disease progression and surgical outcome. Implementing a fully automatic and robust registration is crucial for clinical use, however, it is challenging due to substantial change in shape and appearance due to lesions. In this paper we present a novel method to automatically align longitudinal spine CTs and accurately assess lesion progression. Our method follows a two-step pipeline where vertebrae are first automatically localized, labeled and 3D surfaces are generated using a deep learning model, then longitudinally aligned using a Gaussian mixture model surface registration. We tested our approach on 37 vertebrae, from 5 patients, with baseline CTs and 3, 6, and 12 months follow-ups leading to 111 registrations. Our experiment showed accurate registration with an average Hausdorff distance of 0.65 mm and average Dice score of 0.92.

OBJECTIVE: The aim of this study was to compare the ability of 1) CT-derived bone lesion quality (classification of vertebral bone metastases [BM]) and 2) computed CT-measured volumetric bone mineral density (vBMD) for evaluating the strength and stiffness of cadaver vertebrae from donors with metastatic spinal disease.

METHODS: Forty-five thoracic and lumbar vertebrae were obtained from cadaver spines of 11 donors with breast, esophageal, kidney, lung, or prostate cancer. Each vertebra was imaged using microCT (21.4 μm), vBMD, and bone volume to total volume were computed, and compressive strength and stiffness experimentally measured. The microCT images were reconstructed at 1-mm voxel size to simulate axial and sagittal clinical CT images. Five expert clinicians blindly classified the images according to bone lesion quality (osteolytic, osteoblastic, mixed, or healthy). Fleiss' kappa test was used to test agreement among 5 clinical raters for classifying bone lesion quality. Kruskal-Wallis ANOVA was used to test the difference in vertebral strength and stiffness based on bone lesion quality. Multivariable regression analysis was used to test the independent contribution of bone lesion quality, computed vBMD, age, gender, and race for predicting vertebral strength and stiffness.

RESULTS: A low interrater agreement was found for bone lesion quality (κ = 0.19). Although the osteoblastic vertebrae showed significantly higher strength than osteolytic vertebrae (p = 0.0148), the multivariable analysis showed that bone lesion quality explained 19% of the variability in vertebral strength and 13% in vertebral stiffness. The computed vBMD explained 75% of vertebral strength (p < 0.0001) and 48% of stiffness (p < 0.0001) variability. The type of BM affected vBMD-based estimates of vertebral strength, explaining 75% of strength variability in osteoblastic vertebrae (R2 = 0.75, p < 0.0001) but only 41% in vertebrae with mixed bone metastasis (R2 = 0.41, p = 0.0168), and 39% in osteolytic vertebrae (R2 = 0.39, p = 0.0381). For vertebral stiffness, vBMD was only associated with that of osteoblastic vertebrae (R2 = 0.44, p = 0.0024). Age and race inconsistently affected the model's strength and stiffness predictions.

CONCLUSIONS: Pathologic vertebral fracture occurs when the metastatic lesion degrades vertebral strength, rendering it unable to carry daily loads. This study demonstrated the limitation of qualitative clinical classification of bone lesion quality for predicting pathologic vertebral strength and stiffness. Computed CT-derived vBMD more reliably estimated vertebral strength and stiffness. Replacing the qualitative clinical classification with computed vBMD estimates may improve the prediction of vertebral fracture risk.