How IMIDEX uses V7 to Detect Pulmonary Nodules, Indicative of Lung Cancer
The Challenge
IMIDEX revolutionizes the way medical professionals detect subtle lung nodules in chest X-rays with cutting-edge AI technology.
The company’s goal was to achieve an FDA clearance for their nodule detection technology, which required great model results and a tool to set up a rigorous clinical study. With a vast amount of data and strict medical regulations to navigate, IMIDEX needed a training data platform that would allow them to customize data workflows, easily generate large DICOM datasets, and simultaneously employ multiple users from across the US.
The Solution
IMIDEX used V7 to annotate chest X-rays and execute multi-site clinical validation studies for product testing. V7’s user-friendly interface, advanced data management, DICOM labeling tools, annotation tracking, and support services allowed the team to generate high-quality ground truth data. The data formed the basis to train a lung nodule detection model, eventually becoming IMIDEX’s flagship product, VisiRad XR.
As a part of critical testing for VisiRad XR, IMIDEX used V7’s custom workflows and DICOM hanging protocol capabilities. The study results demonstrated that VisiRad XR improved every participating clinician’s ability to detect lung nodules and masses on chest X-rays. Utilizing these results, VisiRad XR was able to achieve FDA clearance in August 2023.
About IMIDEX
IMIDEX is on a mission to save lives lost to lung cancer by leveraging the power of AI.
The team of experts in data science, radiology, data curation, and regulatory compliance has created VisiRad XR. This revolutionary AI software helps radiologists identify lung nodules and masses in chest X-rays.
Lung nodules, often missed in routine care, are early indicators of lung cancer. VisiRad XR aims to close the gap in diagnosis and treatment, providing hospitals with the tools to catch cancer at an earlier stage when outcomes are much better.
VisiRad XR’s incredible performance led to receiving an FDA 510(k) clearance in August 2023. The model detects lung nodules and masses at a sensitivity of 83%, which is nearly 30% greater than radiologists are able to achieve in regular practice.
Annotating over ten thousand DICOM Images on V7
With the high stakes of the project, the steep quality demands of medical AI products, and the accuracy levels required to obtain FDA clearance, the IMIDEX team needed a tool that would let them make pixel-perfect annotations and set up highly complex workflows with airtight QA.
To train and test their models, IMIDEX prepared large datasets (tens of thousands of images) of accurately labeled chest X-rays. The team created custom data workflows tailored to specific projects to generate the highest quality outputs and adhere to rigorous QA processes. These workflows often included consensus stages and blind reads, where radiologists annotate the same images simultaneously without viewing each other's work.
IMIDEX also used V7’s workflows for their standalone and clinical studies needed for FDA clearance. These studies, which simulated the use of the product in clinical practice, involved hundreds of images and radiologists from across the USA. The clinical study data and workflows were among the most complex that IMIDEX undertook during the development process. In the team’s own words, "Without V7, we wouldn’t have been able to set up and execute these studies at the speed and scale we needed. As a small company that moves quickly, we’re incredibly grateful to have the partnership and support of V7."
Setting up complex annotation and clinical study workflows
The IMIDEX team needed a training data platform that would let them:
- Easily set up custom workflows—including QA processes with a consensus stage, blind reads, and a workflow for clinical studies
- Make pixel-perfect annotations
- Handle huge volumes of DICOM images
- Keep track of annotation progress
V7 met IMIDEX’s needs, providing a solution that balanced workflow complexity with ease of use. The IMIDEX team was able to set up workflows to coordinate the process of preparing training data and, and workflows to execute standalone and clinical studies needed to obtain FDA clearance. The workflow system provided a clear, high-level view of the annotation process, which was crucial in managing the remote team of annotators.
V7’s stability and intuitive interface, along with handy keyboard shortcuts for easier navigation, contributed to increased efficiency. Furthermore, IMIDEX has been continually impressed with V7's customer support—any emerging issues were solved within 24 hours.
Getting FDA clearance for a life-saving AI tool
By using V7 for data labeling, quality assurance, and custom workflows, IMIDEX improved the accuracy of their computer vision model, VisiRad XR, and executed studies needed for FDA clearance.
With high-quality, diverse training data, IMIDEX built a sophisticated medical tool with an astonishing accuracy level—proven by the improvement in area under the curve (AUC) and increased sensitivity across all readers in clinical validation studies. VisiRad XR positively impacted the radiologists’ success rate in detecting pulmonary nodules and masses within chest X-rays. The IMIDEX team estimates that across the US, VisiRad XR could identify up to 1.3 million additional lung nodules or masses annually.
Currently, half of the lung cancer cases in the US are identified in later stages, which brings down the five-year survival rates to 5-10%. With VisiRad XR, IMIDEX team can continue supporting radiologists in detecting lesions that may become early-stage cancer, contributing to better patient outcomes and, ultimately, more lives saved.
.jpg)
