The Role of Artificial Intelligence in Early Diagnosis of Rare Diseases: A Systematic Review

Abstract

Background: Rare diseases pose significant diagnostic challenges due to
their low prevalence, atypical presentations, and overlapping symptoms with
more common conditions. Artificial intelligence (AI) has emerged as a
promising tool for early and accurate diagnosis by leveraging pattern
recognition and predictive algorithms. This systematic review aims to
evaluate the role and effectiveness of AI in the early diagnosis of rare
diseases.
Methods: A systematic search was conducted across databases (PubMed,
Scopus, and Web of Science) to identify studies published between 2010 and
2023 focusing on AI applications in diagnosing rare diseases. Metrics
analyzed included sensitivity, specificity, positive predictive value (PPV),
negative predictive value (NPV), and overall accuracy.
Results: A total of 42 studies involving over 10,000 participants were
included. AI models demonstrated a pooled sensitivity of 89.7%, specificity of
92.5%, PPV of 88.4%, NPV of 93.6%, and accuracy of 91.2%. AI tools were
particularly effective in diagnosing genetic disorders, rare cancers, and
metabolic diseases. Neural networks and deep learning approaches showed
the highest diagnostic performance.
Conclusion: AI has demonstrated high accuracy in diagnosing rare diseases,
particularly in fields requiring complex data integration, such as genetics and
radiology. Continued development and integration of AI tools into clinical
practice could reduce diagnostic delays and improve patient outcomes.