High-frequency waveform recordings of biological signals enable more detailed data analysis and deeper physiological exploration. However, the waveform data—like invasive arterial blood pressure (ABP)—are particularly susceptible to frequent contamination with artifacts that can devalue the subsequent calculations like pressure reactivity index (PRx). This study aimed to verify the ability of the short-time Fourier transform (STFT) based algorithm to detect artifacts in the ABP waveform. Four types of modeled artifacts (rectangular, fast impulse, sawtooth and baseline drift) with different durations and amplitudes were inserted into undisturbed ABP waveforms. Short-time Fourier transform with a 5-second time window is computed on artifact-polluted ABP signals to detect changes in the frequency domain caused by these artifacts. An algorithm with three decision-making rules based on the dominant frequency component, standardized power spectrum, and the value of the second harmonic of the dominant frequency was used. Only segments that passed all three rules were labeled as artifact-free. Results indicated high sensitivity (93.35% and 94.83%) in detecting rectangular and sawtooth artifacts, with specificity exceeding 99% for both. Baseline drift artifact was detected with a low sensitivity of 5.02%, and fast impulse was not detected. This study proposes the application of a short-time Fourier transform-based algorithm to enhance the detection of clinically significant artifacts in arterial blood pressure signals, particularly relevant for PRx and other secondary calculations.

Department of Anaesthesiology and Critical Care Bulovka University Hospital Prague Czech Republic

Department of Anaesthesiology Perioperative and Intensive Care Medicine Faculty of Health Studies J E Purkyne University in Usti nad Labem and Krajska zdravotni a s Masaryk Hospital Usti nad Labem Czech Republic

Department of Biomedical Technology Faculty of Biomedical Engineering Czech Technical University Prague Kladno Czech Republic

Citace poskytuje Crossref.org

Literatura