Akademik Veri Yönetim Sistemi
24th Triennial Meeting of the International Association of Forensic Sciences, Oblast Sofia-Grad, Bulgaristan, 25 - 30 Mayıs 2026, ss.1040-1041, (Özet Bildiri)
Background: In forensic psychiatry, particularly in postmortem cases, access to mental history is frequently limited, which results in diagnostic ambiguity. Emerging spectroscopic methods, however, may offer biochemical patterns revealing underlying psychiatric pathology. The current study investigates the potential of Machine Learning (ML) algorithms in distinguishing psychotic diseases’ molecular indicators based on Fourier Transform Infrared Spectroscopy (FTIR) spectra of serum, particularly in the absence of clinical data.
Methods: Serum samples from 20 patients diagnosed with psychotic disorders and 20 healthy individuals were analyzed using FTIR spectroscopy. Principal Component Analysis (PCA) was used for dimensionality reduction, followed by classification with Random Forest, Support Vector Machine (SVM), and Partial Least Squares Discriminant Analysis (PLS-DA). Model performance was assessed using confusion matrices, accuracy, sensitivity, specificity, and kappa statistics.
Results: Random Forest showed the highest overall accuracy (83.3%), with perfect sensitivity for the control group but reduced specificity for detecting psychotic disorders. SVM and PLS DA produced comparable accuracies (75%), with relatively higher specificity for psychosis but lower sensitivity for controls. Across all models, two spectral bands consistently emerged as discriminative: 1040 the 830-842 cm⁻¹, associated with lipid and membrane modifications, and the 1701-1705 cm⁻¹, corresponding to Amide I band, suggesting protein secondary structural changes. These findings align with existing literature linking psychotic disorders to inflammation, oxidative stress, and metabolic disorders, all of which exhibit as protein misfolding, and lipid peroxidation processes that manifest as detectable spectral changes of altered biochemical profiles.
Conclusion: FTIR-based multivariate classification shows promise as a supportive tool for detecting psychotic disorders using biochemical spectral fingerprints. In forensic psychiatry, this method can provide crucial diagnostic clues when mental history is absent, such as in postmortem investigations. The combination of FTIR and machine learning therefore provides a new viewpoint on forensic diagnosis, connecting molecular evidence with psychological evaluation.