RadNet subsidiary DeepHealth has received 510(k) clearance from the US Food and Drug Administration (FDA) for DeepHealth Breast Ultrasound, an AI-based tool developed for breast ultrasound imaging.
The technology is available for sale in the US, enabling healthcare providers to access reimbursement through an existing Category III CPT code for quantitative ultrasound tissue characterisation.
The DeepHealth Breast Ultrasound system automates the detection, characterisation, and reporting of breast lesions.
The company states that this aims to standardise workflows for sonographers and radiologists, potentially leading to improved consistency and efficiency during breast ultrasound examinations.
To support FDA clearance, the company referred to evidence from a multi-reader, multi-case study involving 16 board-certified radiologists at selected US imaging centres and hospitals.
DeepHealth also reported validating the tool in live clinical settings under regulated research protocols operated by RadNet.
RadNet California women’s imaging medical director Dr Jason McKellop said: “Breast ultrasound is an essential component of the breast care pathway, with approximately 40% of women undergoing the exam at some point in their lives. It is a highly complex, operator-dependent examination, which can lead to significant variability in image acquisition, interpretation and reporting.
“With DeepHealth’s breast ultrasound solution, we can achieve greater standardisation of workflows, improving consistency while saving time for patients, sonographers and radiologists. By streamlining the examination process, we can help reduce exam times, enhance efficiency and ultimately improve patient outcomes.”
DeepHealth’s AI technology is designed to assist clinicians by automating lesion localisation, characterising features in line with the ACR BI-RADS classification, and generating radiology reports.
The company cited results indicating more than 98% accuracy in lesion localisation and an 8% improvement in sensitivity for breast cancer detection, as well as a 37% reduction in radiologist interpretation time.
RadNet intends to implement DeepHealth Breast Ultrasound across its network by the end of the year, estimating that more than 700,000 annual breast ultrasound studies may qualify for reimbursement.
The DeepHealth portfolio also includes AI-enabled tools for mammography, density assessment, breast arterial calcification assessment, risk prediction, and operational analytics.
In September 2024, DeepHealth and HOPPR entered a partnership to commercialise a medical-grade generalised foundation model that will support the development of fine-tuned models for cancer detection.
