Solitary bone lesions: which ones to worry about?

Automated Bone Tumor Segmentation and Classification as Benign or Malignant Using Computed Tomographic Imaging

The purpose of this study was to pair computed tomography (CT) imaging and machine learning for automated bone tumor segmentation and classification to aid clinicians in determining the need for biopsy. In this retrospective study (March 2005-October 2020), a dataset of 84 femur CT scans (50 females and 34 males, 20 years and older) with definitive histologic confirmation of bone lesion (71% malignant) were leveraged to perform automated tumor segmentation and classification. Our method involves a deep learning architecture that receives a DICOM slice and predicts (i) a segmentation mask over the estimated tumor region, and (ii) a corresponding class as benign or malignant. Class prediction for each case is then determined via majority voting. Statistical analysis was conducted via fivefold cross validation, with results reported as averages along with 95% confidence intervals. Despite the imbalance between benign and malignant cases in our dataset, our approach attains similar classification performances in specificity (75%) and sensitivity (79%). Average segmentation performance attains 56% Dice score and reaches up to 80% for an image slice in each scan. The proposed approach establishes the first steps in developing an automated deep learning method on bone tumor segmentation and classification from CT imaging. Our approach attains comparable quantitative performance to existing deep learning models using other imaging modalities, including X-ray. Moreover, visual analysis of bone tumor segmentation indicates that our model is capable of learning typical tumor characteristics and provides a promising direction in aiding the clinical decision process for biopsy.

Prevalence of benign bone lesions of the lower extremity in the pediatric spinal disorders: a whole-body imaging study

There is a paucity of knowledge about benign bone lesions. The advances in imaging methods can screen bone lesions incidentally, and missing information can be provided. The aim of the study is to collect information about the prevalence and natural history of benign bone lesions with the use of whole-body biplanar slot-scanning imaging (EOS). EOS images acquired between 2015 and 2020 were retrospectively analyzed. Anatomical locations of lesions, number of lesions with polyostotic involvement and radiographic features of each were recorded. Fibrous lesions were further categorized according to the transition stages. The natural course was noted as remained in the same stage, progressed and disappeared during follow-up. A total of 1944 EOS images of 1378 (936 women and 442 men) patients were analyzed. The mean age was 12.3 (5-18) years. Bone lesions of the lower extremities were found in 278 scans (14.3%) of 196 (139 women and 57 men) patients (14.2%). Monostotic lesions were observed in 172 patients, and 24 had polyostotic lesions. The prevalence of lesions was 10.5%, 1.8%, 1.7%, 1.7% and 1.4‰ for fibrous cortical defect (FCD), nonossifying fibroma (NOF), osteochondroma, bone island and simple bone cyst, respectively. Among 145 FCDs, 55.2% of the lesions were stage A, 27.6% were stage B, 9.6% were stage C and 7.5% were stage D. EOS images acquired predominantly for spinal pathologies revealed a prevalence of 14% of benign bone tumors in the lower extremities. With the developments in imaging methods, the probability of encountering incidental lesion increases, and information about bone pathologies can be gathered.