Israeli health technology company Aidoc has developed CARE1 (Clinical AI Reasoning Engine, Version 1), an advanced clinical-grade foundation model for CT imaging.
CARE1 represents the initial step in the company’s multi-year investment in its Clinical AI Reasoning Engine (CARE) framework.
The CARE framework was developed by training on millions of exams and using advanced multimodal capabilities, to establish a new benchmark in developing clinical AI solutions.
Aidoc designed the new CARE1 foundation model to provide clinicians with superior diagnostic accuracy and efficiency.
The solution unlocks the full potential of AI to address critical challenges in healthcare by minimising diagnostic delays, optimising workflows, and improving patient care, said the Israeli firm.
Aidoc CTO Michael Braginsky said: “Eighteen months ago, we embarked on an ambitious journey to redefine how clinical AI is developed and, in turn, how it serves healthcare.
“By unlocking new possibilities for AI, we aim to support clinicians in tackling complex challenges and help them set a new standard for patient care that the prior AI technologies couldn’t achieve.
“The CARE framework will gradually serve as the backbone of Aidoc’s AI solutions, enabling the development of versatile, high-performance AI capable of detecting a broad spectrum of critical conditions in real-time.”
Aidoc said that its CARE framework stands out from other foundation models building on its accuracy, and superior performance than the company’s existing AI algorithms.
The core strengths of the CARE framework include Aidoc’s nearly a decade of experience in clinical AI and a large-scale install base in the industry.
Aidoc has developed advanced AI infrastructure to address the requirements of clinical-grade AI, including a robust validation framework, which was refined over the years.
The company’s strong financial support enabled data processing, training, and testing to ensure the delivery of the CARE1 foundation model.
Aidoc achieved superior results for CARE1 by training the models on a superior data scale that spans millions of cases across various conditions and anatomies.
Aidoc CEO Elad Walach said: “As we’ve all learned over the past decade of advancing clinical AI, the greatest challenge is delivering meaningful outcomes in real-world clinical settings.
“With CARE1, Aidoc is just beginning to explore the immense possibilities that foundation models can unlock.
“Our mission has always been to empower providers to transform the core of diagnostic work, and CARE1 is only the tip of the iceberg in an exciting journey toward delivering even greater value across the healthcare ecosystem.”
Aidoc designed the CARE1 model to transform clinical AI and is well-positioned to set a new benchmark for accuracy, flexibility, and reliability in healthcare diagnostics.
The company has already submitted its first real-world derivative of the CARE1 foundation model for the US Food and Drug Administration (FDA) 510(k) review.
In the initial regulatory application, Aidoc used CARE1 in its retrained rib fractures triage module with the addition of a pre-training layer.
Braginsky added: “We decided to start this shift by transforming the rib fractures triage module to be CARE1-enabled, allowing us to begin learning how to measure, validate and ultimately bring foundation model-enabled algorithms to production in clinical settings.
“This marks a pivotal shift in AI development at Aidoc. We’ve fully committed to foundation model technology because it’s clear this approach surpasses anything we’ve seen before.
“At the same time, we are steadfast in ensuring this transition is done with precision, focusing on safety, rigorous validation and a comprehensive understanding of this groundbreaking model’s behaviour to ensure it delivers meaningful clinical impact.”
