Clinical research on secure multi-party computation technology for auxiliary respiratory pathogen diagnosis and treatment design in preschool-aged children

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Background Privacy-preserving integration of respiratory pathogen, host and treatment data remains difficult in preschool-aged children. We evaluated secure multi-party computation (SMC) as clinician-reviewed decision support for respiratory pathogen diagnosis and treatment planning, while distinguishing technical validation from clinical efficacy. Methods This prospective observational study with a parallel in silico comparator was conducted at a tertiary paediatric centre in 2023 among children aged 3–6 years undergoing respiratory pathogen testing and treatment planning. A three-party SMC system using Brakerski-Gentry-Vaikuntanathan homomorphic encryption analysed pathogen markers, clinical features, immune status, inflammatory biomarkers and prespecified treatment rules. Diagnostic-support metrics were calculated against final clinician-documented diagnoses; the non-encrypted comparator assessed encryption overhead and output equivalence. Results Among 1,380 enrolled children, Mycoplasma pneumoniae immunoglobulin M (IgM), an atypical bacterial marker, had the highest positivity rate (41.30%), followed by influenza B (9.13%) and influenza A (7.39%) virus IgM. The SMC system showed 94.7% diagnostic-support concordance, macro-AUROC 0.962, precision 94.0%, recall 95.1% and F1-score 94.5%. Mean processing time was 2.3 ± 0.5 seconds, uptime was 99.97% and no cryptographic failure or unauthorised access event was detected. Treatment outcomes were interpreted descriptively. Conclusions SMC-based clinician-reviewed decision support was feasible for privacy-preserving auxiliary respiratory pathogen diagnosis and treatment planning in preschool-aged children at a single tertiary centre. The findings supported technical feasibility and reference-standard concordance but did not establish superiority over standard physician care. Broader multicentre, age-specific validation is required.