Comprehensive Summary
Kakada et al write a systematic review evaluating the accuracy of AI in identifying pemphigus and pemphigoid disorders based on clinical images. Pemphigus and pemphigoid autoimmune disorders characterized by blister formation of the skin, eyes, and mouth. The authors searched major databases for AI models used in pemphigus/phemphigoid diagnosis and statistically pooled their diagnostic performance. They assessed the biases of the studies using QUADAS-2 and synthesized results with SROC modeling. The results showed that AI models were moderately accurate, correctly identifying the diseases about 86% of the time and correctly ruling them out about 84% of the time. Further statistical analysis also showed that AI models were consistently strong at both confirming and ruling out a diagnosis.
Outcomes and Implications
Pemphigus and pemphigoid are rare autoimmune diseases that are difficult to diagnose quickly and accurately. Early identification is crucial for preventing complications, and AI has the potential to reduce diagnosis time and differences between clinicians. This work showed the AI models can achieve moderate sensitivity and specificity, suggesting that they could support specialists in making faster and more reliable diagnoses. These AI tools can be especially useful in settings without specialists and where interpretation of current biopsy techniques are limited. The authors write that AI could be helpful in diagnosing these conditions, but currently the research varies in quality and needs to be more consistent. They explain that the AI tools will need better data and more real-world testing before it can be used in clinical settings.