Quality Assurance of Medical Mass Spectrometry with Artificial Intelligence

Abstract

Medical liquid chromatography-mass spectrometry (LC-MS) analysis requires technical quality assessment before results can be reported. Quality requirements involve technical validation of instrument parameters for each sample and analyte. Two main challenges limit this process: reliance on manual procedures using spreadsheets and macros, and the effectiveness of current procedures based on the Clinical and Laboratory Standards Institute (CLSI) guidelines C62-A. These guidelines suggest cutoff values for technical validation but fail to consider the dynamic interaction between instrument parameters. Therefore, the current quality assurance process is resource-intensive, prone to human error, challenging to scale, and limits both the number of technical parameters and the complexity of algorithms for their evaluation. As a direct consequence, a high degree of uncertainty with regard to analytical quality remains for a large proportion of analytical results. This constraint reduces the effectiveness of the technical. validation and the accuracy of the analytical quality predictions. The aim of this study was to improve the efficiency and reliability of technical quality assurance procedures in medical LC- MS analyses using artificial intelligence. Experimental data were obtained by repeated LC-MS analysis of six concentration levels of steroid hormones in serum under controlled conditions. The data were used to train and evaluate seven machine learning algorithms. The objective was to develop a model capable of accurately classifying the analytical quality of samples as either acceptable or not acceptable based on concentrations and patterns in instrument parameters. The model would be compared to the state-of-the-art performance for the experimental data of 11.7% recall, 9.0% precision and 94.3% specificity.