UK-based computer vision platform Encord has rolled out a new purpose-built 3D annotation tool for healthcare AI, to automate medical-imaging annotation.

The new Digital Imaging and Communications in Medicine (DICOM) tool is said to be the first to facilitate 3D annotation for modalities such as CT, X-ray, and MRI.

It is capable of rendering more than 20,000-pixel intensities, while the existing tools can manage only 256 pixels, and features a label review functionality to support the FDA approval processes.

The new DICOM tool, which adds to its existing platform, will help automate the annotation process by delivering model inference models and interpolation, said the company.

Encord co-founder and CEO Ulrik Stig Hansen said: “Existing options rely on outsourcing data to human labellers, including clinicians. Human error arising from this process leads to clinicians wasting time reviewing and correcting labels.

“Instead, our tool allows healthcare AI firms to unlock the power of automation through deep-learning models. This reduces costs, increases efficiency and accuracy. We’re excited to see the positive impact this will have for our healthcare customers.”

Encord has worked with Kings College London, where it helped to annotate pre-cancerous polyp videos with superior efficiency, and automated 97% of labels.

In addition, the company has worked with Memorial Sloan Kettering Cancer Centre and Stanford Medical Centre, where it has cut the duration of the experiment by 80% and processed three times more images.

According to the company, current 3D annotation methods depend on human interaction to prepare training datasets for AI use.

Its DICOM annotation tool replaces manual processes by harnessing the power of automation through deep learning, which is otherwise costly and time-consuming.

The solution delivers 100% data privacy and security to its users, as the platform is deployed to its existing systems rather than transferring the data externally.

Its DICOM tool not only saves time and money, but also provides a data pipeline built around the images, said Encord.