Artificial Intelligence System for Diagnosing Rare Diseases: Design Principles and Clinical Validation Results

Introduction. Differential diagnostics of rare diseases at the pre-laboratory stage of patient examination is a significant challenge not only for pediatricians but also for geneticists. This fact is caused by limited physician experience in managing rare diseases, variability in symptom presentation, and ambiguity of clinical signs. AI-driven clinical decision support systems (CDSS) enable the generation and validation of diagnostic hypotheses.

Aim: To assess the architecture of the GenDiES (Genetic Diagnostic Expert System) CDSS for differential diagnosis of lysosomal storage diseases (LSDs) at the pre-laboratory stage and to present results from its clinical validation.

Material and Methods. The study included 30 clinical forms of LSDs, described using 35 clinical features based on three complementary expert-derived metrics: modality coefficient (diagnostic importance), manifestation certainty factor, and degree of expression certainty factor. Knowledge was extracted from literature and refined through expert input, where experts assigned their confidence to each feature across four age groups: ≤ 1 year, 1–3 years, 4–6 years, ≥ 7 years. This structured data formed the system’s knowledge base. Clinical aprobation utilized de-identified electronic health records (EHR) of pediatric LSD patients (mucopolysaccharidoses, mucolipidoses, gangliosidoses-conditions with broad overlapping phenotypic spectra). Validation cohort: 54 EHR extracts from a single Russian medical institution. Verification cohort: 38 EHR extracts from three institutions across different Russian regions. The system was built using knowledge engineering methods (for knowledge extraction and structuring), a matrix-based framework (to organize rules), and custom software for implementation.

Results. The updated GenDiES CDSS for differential diagnosis of rare hereditary diseases was deployed as a web application. Its knowledge base contains 12,600 expert confidence assessments for 35 clinical features across 30 LSD subtypes, categorized by age. A similarity-based algorithm compares patient profiles to expert-defined disease patterns. Accuracy for generating a differential diagnosis shortlist (top five hypotheses) reached 0.87 (95% CI [0.75; 0.95]) during validation and 0.90 (95% CI [0.75; 0.97]) during verification.

Conclusion. The GenDiES system demonstrated high diagnostic accuracy at the pre-laboratory stage, comparable to—and in some cases exceeding—the performance of limited existing international counterparts. Its web-based implementation ensures accessibility for physicians via any internet-connected device.

1. Richter T., Nestler-Parr S., Babela R., Khan Z.M., Tesoro T., Molsen E. et al. Rare diseases terminology and definitions – A systematic global review: Report of the ISPOR Rare Disease Special Interest Group. Value Health. 2015;18(6):906–914. https://doi.org/10.1016/j.jval.2015.05.008

2. Balkunova Y.N., Kochergina E.A., Bazanova N.A. Clinical case: type II/IIIA mucolipidosis in a child. Perm Medical Journal. 2023;40(4):135–140. (In Russ.). https://doi.org/10.17816/pmj404135-140

3. Gorbunova V.N. Congenital metabolic diseases. Lysosomal storage diseases. Pediatrician (St. Petersburg). 2021;12(2):73–83. (In Russ.). https://doi.org/10.17816/PED12273-83

4. Borges P., Pasqualim G., Giugliani R., Matte U. Estimated prevalence of mucopolysaccharidoses from population-based exomes and genomes. Orphanet J. Rare Dis. 2020;15:324. https://doi.org/10.1186/s13023-020-01608-0

5. Zakharova E.Yu., Baidakova G.V., Mikhailova S.V., Pchelina S.N., Krasnopolskaya K.D. Lysosomal Storage Diseases: A Guide for Clinicians. Moscow: GEOTAR-Media; 2021:424. https://doi.org/10.33029/9704-6321-5-LAD-2021-1-424. ISBN 978-5-9704-6321-5.

6. Nazarenko L.P., Nazarenko M.S. THE early symtoms of lysosomal storage diseases. Medical Genetics. 2013;12(9):20–23. (In Russ.). EDN: TJBVIV

7. Kravchuk ZP, Rumyantseva OA. Orphan diseases: diagnosis, problems, prospects. Health and Ecology Issues. 2013;(4):7–11. (In Russ). https://doi.org/10.51523/2708-6011.2013-10-4-1

8. Kang Q., Fang Y., Yang Y., Li D., Zheng L., Chen X. et al. Health service utilization, economic burden and quality of life of patients with mucopolysaccharidosis in China. Orphanet J. Rare Dis. 2024;19:324. https://doi.org/10.1186/s13023-024-03333-4

9. Shashel V.A., Firsova V.N., Trubilina M.M., Podporina L.A., Firsov N.A. Orphan diseases and associated problems. Medical Herald of the South of Russia. 2021;12(2):28–35. (In Russ). https://doi.org/10.21886/2219-8075-2021-12-2-28-35

10. Krishnan G., Singh S., Pathania M., Gosavi S., Abhishek S., Parchani A., Dhar M. Artificial intelligence in clinical medicine: catalyzing a sustainable global healthcare paradigm. Frontiers in Artificial Intelligence. 2023;6:1227091. https://doi.org/10.3389/frai.2023.1227091

11. Gavrilova T.A., Kudrjavcev D.V., Muromcev D.I. Inzhenerija znanij. Modeli i metody: Uchebnik. St. Petersburg: Lan'; 2023:324. (In Russ.). ISBN: 978-5-507-46580-4.

12. Kobrinskii B.A. Certainty factors triunity in the medical diagnostics tasks. Artificial Intelligence and Decision Making. 2018;(2):62–72. (In Russ.). https://doi.org/10.14357/20718594180205

13. Gribova V.V., Petryaeva M.V., Shalfeeva E.A. Сloud decision support service for diagnosis in gastroenterology. Medical doctor and information technologies. 2019;(3):65–71. (In Russ.).

14. Orekhova T.F., Kruzhilina Т.V., Neretina T.G. Matrix approach to the description of pedagogical processes in scientific pedagogical research. Business. Education. Law. 2020;(2):301–308. (In Russ.). https://doi.org/10.25683/VOLBI.2020.51.213

15. Blagosklonov N.A., Kobrinskii B.A. Decision making in conditions of incomplete or redundancy data. Integrated models and soft computing in artificial intelligence. Proceedings of the XII International Scientific and Practical Conference (IMSC-2024, Kolomna, May 14–17, 2024). In 2 vol. Vol. 1. Smolensk: Universum; 2024:55–63. (In Russ.). EDN: WKNXQX

16. Blagosklonov N.A., Kobrinskii B.A. Differential diagnosis of hereditary metabolic diseases using the expert knowledge-based system. Siberian Journal of Clinical and Experimental Medicine. 2020;35(4):71–78. (In Russ.). https://doi.org/10.29001/2073-8552-2020-35-4-71-78

17. Ronicke S., Hirsch M.C., Türk E., Larionov K., Tientcheu D., Wagner A.D. Can a decision support system accelerate rare disease diagnosis? Evaluating the potential impact of Ada DX in a retrospective study. Orphanet J. Rare Dis. 2019;(14):69. https://doi.org/10.1186/s13023-019-1040-6

18. Alves R., Piñol M., Vilaplana J., Teixidó I., Corella J., Comas J. et al. Computer-assisted initial diagnosis of rare diseases. PeerJ. 2016;4:e2211. https://doi.org/10.7717/peerj.2211

19. Pantel J.T., Zhao M., Mensah M.A., Hajjir N., Hsieh T.-C., Hanani Y. et al. Advances in computer-assisted syndrome recognition by the example of inborn errors of metabolism. J. Inherit. Metab. Dis. 2018;41(3):533–539. https://doi.org/10.1007/s10545-018-0174-3

20. Carrer A., Romaniello M.G., Calderara M.L., Mariani M., Biondi A., Selicorni A. Application of the Face2Gene tool in an Italian dysmorphological pediatric clinic: Retrospective validation and future perspectives. Am. J. Med. Genet. Part A. 2024;194(3):e63459. https://doi.org/10.1002/ajmg.a.63459