Diagnostic accuracy of artificial intelligence-based multi-spectrum analysis for molecular fingerprint detection of SARS-CoV-2

Reverse transcription-polymerase chain reaction (RT-PCR) is the reference standard for COVID-19 diagnosis, but the need for rapid, reproducible, and cost-effective diagnostic tools remains. This study investigated the diagnostic performance of a novel artificial intelligence-based multispectrum analysis (MSA, AP23) technique that detects the biomolecular fingerprint of severe acute respiratory syndrome coronavirus 2. A prospective, double-blinded observational design was used, involving 3614 volunteers. The artificial intelligence was trained with 2448 samples, validated with 816, and tested against RT-PCR using a blinded set of 350 samples. Sensitivity, specificity, positive predictive value, and negative predictive value were calculated. During validation, MSA achieved 88.4% sensitivity, 88.76% specificity, 86.77% positive predictive value, and 90.18% negative predictive value. In the blinded comparison phase, these values were 81.73%, 81.99%, 75.16%, and 87.81%, respectively, with an area under the receiver operating characteristic curve of 0.89. These findings suggest that MSA offers reliable diagnostic performance and may be a promising alternative to RT-PCR in COVID-19 diagnosis. The study was registered on ClinicalTrials.gov (NCT04860895).