Intraloop neoangiogenesis in an AI-classified scleroderma pattern: recognizing a morphological clue to suspected...

Area of Science:

• Dermatology
• Rheumatology
• Artificial Intelligence in Medicine

Background:

• Nailfold videocapillaroscopy (NVC) is a valuable tool for microvascular assessment in autoimmune diseases.
• AI algorithms are increasingly used to analyze NVC patterns, primarily trained on systemic sclerosis.
• Limitations exist in AI's ability to recognize patterns outside its training scope.

Purpose of the Study:

• To report a diagnostic near miss involving AI-assisted NVC.
• To highlight the importance of integrating AI pattern classification with expert morphological interpretation.
• To emphasize that AI's diagnostic accuracy is confined to its training framework.

Main Methods:

• A case study of a 50-year-old woman with episodic acrocyanosis.
• AI-assisted NVC analysis classified the pattern as 'early scleroderma'.
• Expert review identified intraloop neoangiogenesis and giant capillaries, prompting further immunological testing.

Main Results:

• AI correctly identified an 'early scleroderma' pattern but missed critical morphological details.
• The composite finding of neoangiogenesis and giant capillaries was not recognized by the AI.
• Immunological testing revealed anti-Mi-2 antibodies, suggesting dermatomyositis, not systemic sclerosis.

Conclusions:

• AI-assisted capillaroscopy excels at pattern classification within its training data but cannot replace integrative clinical reasoning.
• Clinicians must critically evaluate AI outputs, considering additional morphological findings like neoangiogenesis.
• Expert interpretation remains essential to avoid diagnostic errors stemming from AI limitations and over-reliance on pattern recognition.