Article of the Month – November 2023
November 29, 2023FDA Updates – November 2023
November 29, 2023Zebra Medical Vision
Zebra-med was created to assist patients, physicians, and healthcare professionals in receiving and giving better, quicker, and more accurate treatment...
Overview
Zebra-Med aims to equip radiologists with the tools they need to improve patient care. The demand for medical imaging services is growing rapidly, surpassing the supply of competent radiologists and forcing them to work longer hours without sacrificing patient care. This issue can only be alleviated by implementing new technology that significantly improves radiologists’ skills. With its breakthrough AI solution, Zebra-Med allows radiologists to handle the ever-increasing workload without compromising quality. Zebra-Med is also using AI to develop drugs.
All-in-one artificial intelligence solutions
Zebra-Med offers an increasing list of AI technologies: One business, one customer service, and a wide variety of AI radiology solutions seamlessly integrated into the workflow. The Imaging Analytics Engine from Zebra-Med takes imaging scans from a variety of modalities and analyzes them automatically for a variety of clinical findings in a fast and seamless manner with radiology workflow. Zebra-Med creates software that analyzes data in real-time with human-level accuracy, using a proprietary library of millions of imaging scans and machine and deep learning techniques. This gives radiologists the help they need to handle ever-increasing workloads without losing quality. (1)
Based on value: Individuals who are at high risk of cardiovascular, lung, bone, and other diseases are identified by Zebra-Med. This simplifies putting preventative care programs in place, making the necessary risk adjustments and allocations, and satisfying quality standards. With Zebra-Med, attention can be focused on the correct patients at the right time, decreasing overall costs while improving care. Its Imaging Analytics Engine diagnoses brain, lung, liver, cardiovascular, and bone sickness in CT pictures, 40 different illnesses in X-ray scans, and breast cancer in 2D mammograms, thanks to an ever-growing pipeline. Zebra-All-in-One Med’s AI service allows radiologists to check out all Zebra-AI Med’s solutions for one low, transparent price. They claim to be able to scan for as little as $1 per scan and provide all their current and future AI-driven algorithmic capabilities while maintaining an All-in-One strategy, thanks to their growing number of solutions. (2)
The company leads the market with 7 FDA-cleared AI solutions incorporated into imaging modalities, Picture Archiving and Communication System (PACS), and Radiology Information System (RIS). The AI1 Imaging Analytics Engine from Zebra-Med can be integrated entirely into PACS/RIS systems to provide critical clinical insights to radiologists in two ways:
- In the viewer — When watching a study in the PACS viewer, the Radiology Assistant provides algorithmic insights immediately at the point of treatment.
- Workflow Integration – AI products for acute medical discoveries provide warnings that may be utilized to prioritize the radiology worklist. (3)
Simulating Dual-Energy X-Ray Absorptiometry in CT Using Deep-Learning Segmentation Cascade
Osteoporosis is considered an important burden for both patients and healthcare systems. Even today, the condition is surprisingly underdiagnosed, affecting the effectiveness of treatment and positive outcomes. In this study, researchers develop a method using a machine-learning algorithm to simulate lumbar DEXA scores from routine CT studies.
All images were taken between 2010 and 2014. These 610 CT studies of the abdomen and pelvis were taken to develop the spinal column and L1 to L4 multiclass segmentation. For DEXA simulation training and validation, 1,843 CT studies were paired with DEXA tests taken from the same patient within six months.
1,693 CT studies with their corresponding DEXA result were used for validation. Results show 1,144 true positives, 92 false positives, 245 true negatives, and 212 false negatives. The sensitivity for osteopenia and osteoporosis was 84.4%, and the specificity was 72.7%.(4).
Automated Opportunistic Osteoporotic Fracture Risk Assessment Using Computed Tomography Scans to Aid in FRAX Underutilization
Osteoporotic fractures represent an increasing medical concern. Currently, the methods utilized to identify patients at risk for osteoporotic fractures are underutilized. On the other hand, premature interventions have been shown to decrease osteoporotic fracture risks. In this study, researchers evaluated the viability of automatic, opportunistic fracture risk evaluation based on routine abdomen and chest CT scans.
They use three automatically generated biomarkers: vertebral compression fractures (VCFs), simulated DXA T-scores and lumbar trabecular density, and CT metadata for age and sex to create a CT-based predictor. A population of 48,227 patients between the ages of 50 and 90 with CT scans taken before 2012 were evaluated for fracture risk using FRAX with no bone mineral density input (FRAXnb) and the CT-based predictor.
Compared to FRAXnb, the CT-based predictor showed better receiver operating characteristics AUC (+1.9%), sensitivity (+2.4%), and PPV (+0.7%). (5)
Fully Automatic Deep Learning Used for Agatston Score Estimation
The cardiovascular complications and risk of death can be quantified by the proportion of coronary artery calcifications (CAC) through the Agatston score.
An advanced system that uses a fully convolutional deep neural network has been performed to estimate the Agatston calcium score in non-contrast CT scans.
1054 chest CTs were analyzed, and results showed a Pearson correlation coefficient (0.98); bland-altman analysis had a bias of 0.4 with 95% limits of agreement of [-189.9-190.7]. Lastly, the linearly weighted Kappa was 0.89 for the risk category assignment. (6)
Machine Learning Algorithms Used to Assess Risk for Cardiovascular Disease
Three automatic algorithms, such as CCS-Alg, Emphy-Alg, and LD-Alg, were proposed as potential cardiovascular disease occurrence and mortality predictors. Furthermore, all three of our algorithms can approximate disease severity and can be used by clinicians in better treatment decisions.
Compared to manual processing, these algorithms use machine learning and AI approaches to quantify calcifications in the coronary arteries. (6)