Radiologist

The FDA has cleared multiple AI tools under the De Novo and 510(k) pathways that are explicitly authorized to make diagnostic decisions without requiring radiologist review of every output. IDx-DR (EyeDiagnosis) was the first FDA De Novo clearance for an autonomous AI diagnostic system in 2018 and operates without a specialist in the loop. Aidoc has 14+ FDA clearances across multiple imaging modalities and body regions. iCAD's ProFound AI for mammography and Hologic's Genius AI both hold FDA clearances for breast imaging decision support with varying levels of autonomy. The total number of FDA-cleared AI/ML medical devices has grown from approximately 100 in 2019 to over 950 by 2024, with radiology accounting for roughly 75% of all clearances — reflecting both the maturity of medical imaging AI and the regulatory pathway being well-understood.

Multiple peer-reviewed studies have documented 30–50% reductions in radiologist time-per-case when AI pre-reads are used as a starting point. A 2023 study in Radiology (Moy et al.) documented that AI-assisted chest X-ray reading reduced radiologist interpretation time by 42% with no significant change in diagnostic accuracy. If a radiologist previously read 80 chest X-rays per shift and now reads 130 with the same accuracy, health systems staffing models immediately register that they need 38% fewer radiologists to cover the same volume. This productivity gain is being actively quantified by health system CFOs and used in radiology department staffing models. The compression is structural: it does not require AI to replace radiologists — it only requires AI to make existing radiologists faster.