Leveraging AI technology in sarcoidosis.

PURPOSE OF REVIEW: Sarcoidosis is a systemic, granulomatous disease of uncertain cause. Diagnosis may be difficult, prognosis uncertain and response to treatment unpredictable. The application of artificial intelligence to sarcoidosis may provide clinical decision support for these challenges. This review will provide an overview of current and potential future applications of artificial intelligence in sarcoidosis. RECENT FINDINGS: The predominant application of artificial intelligence in sarcoidosis is imaging. Imaging models may differentiate sarcoidosis from other pulmonary disorders. Models, which predict survival and identify key factors relevant to prognosis are also available. The application of cluster analysis to organize sarcoidosis patients into developmental phenotypes is underway. Machine learning algorithms to evaluate the treatment response of sarcoidosis patients do not yet exist but similar models may evaluate patients with other inflammatory disease. The potential applications of artificial intelligence to sarcoidosis is vast, but there are practical limitations that warrant consideration. These include: the accessibility of data, biases in data, cost and privacy. SUMMARY: The application of artificial intelligence in medicine is still in its early stages but models are poised to support the diagnostic and prognostic challenges in sarcoidosis patients. The predictive power of these artificial intelligence is likely to come from combining various models, trained on content-rich datasets from phenotypically heterogeneous sarcoidosis patients.

Diagnosis and detection of sarcoidosis: an official American Thoracic Society Clinical Practice Guideline

The diagnosis of sarcoidosis is not standardized but is based on three major criteria: a compatible clinical presentation, finding nonnecrotizing granulomatous inflammation in one or more tissue samples, and the exclusion of alternative causes of granulomatous disease. There are no universally accepted measures to determine if each diagnostic criterion has been satisfied; therefore, the diagnosis of sarcoidosis is never fully secure. Systematic reviews and, when appropriate, meta-analyses were performed to summarize the best available evidence. The evidence was appraised using the Grading of Recommendations, Assessment, Development, and Evaluation approach and then discussed by a multidisciplinary panel. Recommendations for or against various diagnostic tests were formulated and graded after the expert panel weighed desirable and undesirable consequences, certainty of estimates, feasibility, and acceptability. The clinical presentation, histopathology, and exclusion of alternative diagnoses were summarized. On the basis of the available evidence, the expert committee made 1 strong recommendation for baseline serum calcium testing, 13 conditional recommendations, and best practice statement. All evidence was very low quality.The panel used systematic reviews of the evidence to inform clinical recommendations in favor of or against various diagnostic tests in patients with suspected or known sarcoidosis. The evidence and recommendations should be revisited as new evidence becomes available.