Development of a Machine Learning-Based Model for Accurate Detection and Classification of Cervical Spine Fractures Using CT Imaging.

Cervical spine fractures represent a significant healthcare challenge, necessitating accurate detection for appropriate management and improved patient outcomes. This study aims to develop a machine learning-based model utilizing a computed tomography (CT) image dataset to detect and classify cervical spine fractures. Leveraging a large dataset of 4,050 CT images obtained from the Radiological Society of North America (RSNA) Cervical Spine Fracture dataset, we evaluate the potential of machine learning and deep learning algorithms in achieving accurate and reliable cervical spine fracture detection. The model demonstrates outstanding performance, achieving an average precision of 1 and 100% precision, recall, sensitivity, specificity, and accuracy values. These exceptional results highlight the potential of machine learning algorithms to enhance clinical decision-making and facilitate prompt treatment initiation for cervical spine fractures. However, further research and validation efforts are warranted to assess the model's generalizability across diverse populations and real-world clinical settings, ultimately contributing to improved patient outcomes in cervical spine fracture cases.

Optimal Image Gain Intensity of Point-of-care Ultrasound when Screening for Ocular Abnormalities in the Emergency Department.

INTRODUCTION: Point-of-care ultrasound (POCUS) plays a pivotal role in evaluating ocular complaints in the emergency department (ED). The rapid and non-invasive nature of ocular POCUS makes it a safe and informative imaging modality. Previous studies have investigated using ocular POCUS to diagnose posterior vitreous detachment (PVD), vitreous hemorrhage (VH), and retinal detachment (RD); however, there are few studies that assess image optimization techniques and how they impact the overall accuracy of ocular POCUS. METHODS: We performed a retrospective review of ED patients who received ocular POCUS examinations and ophthalmology consultations as part of their evaluation for eye complaints at our urban, Level I trauma center ED from November 2017-January 2021. Of 706 exams, 383 qualified for the study. In this study we primarily investigated how stratified gain levels impact the accuracy of ocular POCUS for detection of any posterior chamber pathology and, secondarily, whether stratified gain levels impact the accuracy of detecting RD, VH, and PVD specifically. RESULTS: The images were found to have an overall sensitivity of 81% (76-86%), specificity of 82% (76-88%), positive predictive value (PPV) of 86% (81-91%), and negative predictive value (NPV) of 77% (70-83%). Images acquired with a gain of (25, 50] had a sensitivity of 71% (61-80%), specificity of 95% (85-99%), PPV of 96% (88-99%), and NPV of 68% (56-78%). Images acquired with a gain of (50, 75] had a sensitivity of 85% (73-93%), specificity of 85% (72-93%), PPV of 86% (75-94%), and NPV of 83% (70-92%). Images acquired with a high gain (75, 100] had a sensitivity of 91% (82-97%), specificity of 67% (53-79%), PPV of 78% (68-86%), and NPV of 86% (72-95%). CONCLUSION: In the ED setting, high (75, 100] gain on ocular POCUS scanning has a higher degree of sensitivity for detecting any posterior chamber abnormality, as compared to low (25, 50] gain levels. Thus, incorporating the use of high gain for ocular POCUS exams produces a more effective tool for ocular pathologies in acute care settings and may be particularly valuable in resource-limited settings.