Feature fusion technology based on serum FTIR spectra combined with chaos theory in the disease auxiliary diagnosis

Chaos theory is a mathematical theory that studies nonlinear dynamical systems and has found extensive applications in the disease auxiliary diagnosis. FTIR spectra is a technique based on infrared spectroscopy that provides information about molecular vibrations, rotations, and vibrational-rotational energy levels by recording the absorption spectrum of a sample in the infrared radiation range. This technology has gained attention for its extensive applications in the disease auxiliary diagnosis. However, due to the limited amount of molecular information captured by FTIR spectra and intricate clinical diagnostic scenarios, this study introduces an innovative approach by combining FTIR spectra with chaos theory. This novel method for disease prediction is proposed and validated using FTIR spectra datasets from various diseases, including glioma, non-small cell lung cancer (NSCLC), and systemic lupus erythematosus (SLE). The experimental results demonstrate that the proposed Low-rank Tensor Features Fusion-BiGRU (LTFF-BiGRU) model achieves competitive outcomes in three datasets. Comparing the spectral features, inputting spectral-chaotic fusion features into LTFF-BiGRU models can effectively improve the average Accuracy (Acc) by 3.5%, average Precision (Pre) by 3.30%, average Sensitivity (Sen) by 2.37%, average Specificity (Spe) by 4.07%, average F1 score by 3.10%, and average Area Under the ROC Curve (AUC) by 3.23%. Through low-rank tensor fusion, the correlations and interaction patterns between different feature data can be effectively captured, thus extracting a more comprehensive and enriched feature representation to enhance disease diagnosis results further. This research marks the first demonstration of chaotic characteristics in FTIR spectra and pioneers the exploration of employing low-rank tensor fusion between spectral features and chaotic features. The research signifies a crucial step in integrating FTIR spectra with chaos theory in the disease auxiliary diagnosis, paving the way for further exploration in this promising interdisciplinary field.