What happened
Imvaria of Berkeley, California, makes Fibresolve, a 3D deep learning model that classifies chest CT scans in suspected interstitial lung disease. It returns "Suggestive of IPF" or "Inconclusive" for idiopathic pulmonary fibrosis (IPF). The test cases came from the Lung Tissue Research Consortium, at sites independent of development, with diagnoses set by multidisciplinary discussion. On 137 scans with slices of 3 mm or less, FDA reviewers report a sensitivity of 55% and a specificity of 82%. An expert panel of five pulmonologists and thoracic radiologists, reviewing charts blind to invasive tests, reached 20% and 95%. FDA disagreed with the split endpoints but found a small probable benefit for early diagnosis. FDA granted the De Novo request on 12 January 2024. The cases date from 2005 to 2018, and the panel applied stricter criteria than current guidelines. FDA required a postmarket study with modern images.
Sources
-
Regulatory document, by authors with no tie to the system.
U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for Fibresolve (DEN220040): decision summary. Granted 2024-01-12.
The device appears to identify true positive IPF patients, at the expense of a high rate of false positives (specificity lower bound of 73%). This tradeoff offers a positive benefit-risk determination, considering the benefits provided by an early IPF diagnosis.
Details
AI system
- Workflow
-
The model analyzed a single stack of axial CT slices 3 mm thick or less. Results went to clinicians as a report without images.
Follow-up
- Conditions for replication
-
The result is used within a referral to multidisciplinary discussion or as part of one. The result is never used to rule out idiopathic pulmonary fibrosis.
Entry details
- Sector
-
Healthcare.
- Purpose
-
Imvaria built Fibresolve to classify lung CT scans in suspected interstitial lung disease as suggestive of idiopathic pulmonary fibrosis, as an adjunct before invasive testing.
Not yet checked: model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.
Cite this record
Peter H. Vartanian. (2026). Fibresolve flagged more lung fibrosis than experts in FDA review (AIBD-0773). Artificial Intelligence Benefits Database, edition of 31 October 2026. https://petervartanian.xyz/aibd/records/AIBD-0773.html (archived at https://doi.org/10.7910/DVN/AADSII, version 4.0)