Aidoc's AI Leap: Foundation Model Aims to Triage a Dozen Conditions

NEW YORK, NY – November 25, 2025 – In a move that signals a significant maturation of artificial intelligence in medicine, clinical AI leader Aidoc has filed for FDA review of a breakthrough device designed to detect and triage a dozen different acute abdominal conditions from a single CT scan. This submission for its CARE™ Foundation Model is not just another algorithm; it represents a fundamental shift from narrow, single-purpose AI to broad, multi-faceted diagnostic partners for radiologists.

The announcement is bolstered by a staggering operational milestone: Aidoc's underlying platform, the aiOS™ (AI Operating System), has now processed over 100 million patient scans. This massive real-world dataset provides the bedrock of safety and reliability the company argues is essential for deploying such powerful, wide-ranging AI in high-stakes clinical environments. For healthcare executives and investors, the move positions Aidoc to potentially capture a new frontier in medical imaging, moving beyond flagging a single condition to providing a comprehensive safety net for a multitude of time-sensitive findings.

The Dawn of Foundation Models in Radiology

For years, clinical AI has largely consisted of "narrow" algorithms, each meticulously trained to identify one specific abnormality, such as a pulmonary embolism or an intracranial hemorrhage. While effective, this approach creates a fragmented ecosystem of single-task tools. The introduction of foundation models, like Aidoc's CARE™, promises to change that paradigm.

Foundation models are large-scale AI systems pre-trained on vast, diverse datasets, which can then be adapted to a wide array of downstream tasks with minimal fine-tuning. In medicine, this means a single, robust model can be taught to recognize patterns associated with numerous diseases. Aidoc's pivotal study for its abdominal multi-triage solution reported impressive results: a mean sensitivity of 97% and a mean specificity of 98%. This high level of accuracy across multiple conditions is critical. As Elad Walach, CEO and Co-Founder of Aidoc, noted, "Broad foundation models are the path to expanding clinical AI across care delivery, but only if we raise the safety and quality bar beyond anything the field has seen."

The challenge with any broad AI, however, lies in avoiding the "noise" of excessive false positives that can lead to alert fatigue and erode physician trust. Walach emphasized that the CARE model "delivers precision that limits false positives and elevates only what matters." This focus on clinical utility is paramount. While the potential is immense, experts caution that the deployment of foundation models in medicine carries significant responsibility. Issues like algorithmic bias, a lack of transparency in the AI's decision-making process (the "black box" problem), and the potential for AI "hallucinations" must be rigorously managed.

More Than a Model: The AI Operating System

Perhaps the most critical piece of Aidoc's strategy is not the foundation model itself, but the platform it runs on. The company's aiOS™ has become the largest clinical AI deployment in healthcare, acting as the central nervous system for AI implementation in over 1,600 medical centers. This isn't just a technical backend; it's a comprehensive governance and orchestration engine.

"It's not just strong AI models - it's the orchestration, workflow intelligence, and performance tracking behind them," explained Neal Patel, MD, MPH, Chief Health Information Officer at Vanderbilt University Medical Center. "This empowers our care teams to routinely deliver the right care at the right time."

This "operating system" approach addresses the core challenges of operationalizing AI at scale. It handles intelligent orchestration, ensuring the right algorithm is applied to the right scan and that findings are routed to the correct physician workflow within existing systems like the PACS and EHR. More importantly, it provides continuous performance monitoring. AI models can experience "drift," where their accuracy degrades over time as scanner technology, patient populations, or clinical protocols change. The aiOS is designed to detect this drift, track overrides, and provide real-time analytics on AI performance, a practice known in the tech world as MLOps (Machine Learning Operations). Having validated its system on over 100 million real-world cases provides an unparalleled dataset for ensuring the models remain robust and generalizable across diverse clinical settings.

Navigating a New Regulatory Frontier

Aidoc's submission has been granted Breakthrough Device Designation by the FDA, a program designed to expedite the review of novel technologies that could provide more effective diagnosis for life-threatening conditions. This designation acknowledges the potential for a multi-triage AI to significantly improve patient care, but it doesn't guarantee an easy path to clearance.

The FDA's 510(k) pathway, through which most AI medical devices are cleared, was designed for products with "substantial equivalence" to existing devices. A first-of-its-kind, multi-condition foundation model presents unique regulatory questions. Regulators are increasingly focused on how manufacturers will ensure generalizability, mitigate bias, and manage the lifecycle of adaptive algorithms. Aidoc's extensive real-world data from its 100 million processed scans and the governance capabilities of its aiOS platform will likely be central to its case for demonstrating safety and effectiveness.

The competitive landscape is heating up, with companies like a2z Radiology AI recently gaining clearance for a device that flags seven abdominal findings. However, Aidoc's push for "double-digit" conditions, combined with its established enterprise platform, represents an ambitious play for market dominance. This submission will be a closely watched test case for how the FDA approaches the next generation of complex, powerful clinical AI.

As radiologists and hospital systems gather for the upcoming Radiological Society of North America (RSNA) annual meeting, the conversation will undoubtedly revolve around this shift from niche AI tools to integrated, enterprise-wide intelligence. While Aidoc's multi-triage device is not yet for sale pending FDA review, its development signifies that the era of foundational AI in clinical practice is no longer a distant vision, but an imminent reality. The industry is watching to see if this combination of a powerful model and a proven operating system can truly deliver on the promise of making healthcare faster, safer, and smarter.