An Integrated Platform For Cardiac Health Diagnostics
Team: Instructors: Kuldeep Singh Rajput, Rohan Puri, Maulik Majmudar, M.D., Dr.Ramesh Raskar
Students: Harsha Vardhan Pokkalla, Aranya Goswami
Software required: MATLAB
Cardio24 Matlab scripts:
- afib_check : Function that takes ECG signal as input and outputs the observed morphologies which would help to identify abnormal parts of the input ECG signal.
- afib_input : Function that takes ECG signal as a file (txt file) and Sample Frequency as inputs, passes this data to 'afib_check' function and stores the outputs in a file (csv/json file)
Description of 'afib_check' Function:
This code is designed based on parameters of ECG signal (Fs, gain) given for MIT-BIH Database, PhysioNet. Adjust output parameters of this function based on need.
Inputs:
1. Input_ECG : Row/Column vector of the ECG signal (flag can be adjusted in the code)
2. Fs : Sampling Frequency of the ECG signal
Outputs:
1. x1 : Preprocessed ECG signal (DC subtracted, gain-reduced)
2. N : Length of the ECG signal
3. Morphs :
size: 1x12 vector
data: 12 Morphological values
i.Max HR
ii. Avg HR
iii. Min HR
iv. Total Number of QRS Complexes
v. Number of Irregular Beats
vi. Percentage of Irregular Beats
vii.Number of Episodes with Consec Irregular Beats
viii. Average PR Interval
ix. Average QRS Interval
x. Average QTc Interval
xi.Number of P wave absences
xii.Number of Episodes that has >4 consec P wave absences
4. Morphs_c :
size: 1x12 vector
data: Normal/Abnormal condition based on above morphologies respectively
Morphs_c(1,i) = 'Normal' or 'Abnormal'
5. time_markRint :
size: 1xN vector
data: marks locations of Irregular Beats with HIGH/LOW. Here: 0.9/0
time_markRint(1,i) = HIGH if Beat at 'i'th sample point is Irregular;Otherwise, LOW.
6. time_RR :
size: 1xN vector
data: Array of Beat to Beat Intervals
time_RR(1,i) = Interval from Beat at 'i'th sample point to Beat at 'i+1'th sample point.
7. P_abs :
size: 1xN vector
data: marks locations of P wave absences with HIGH/LOW. Here: 0.3/0
8. time_RR2:
size: 1xN vector
data: marks the episodes of consecutive irregular beats with value of R_peak.
9. R_loc:
size: 1xnQRS vector (nQRS is number of observed QRS complexes )
data: locations of R peaks
10. R_value:
size: 1xnQRS vector (nQRS is number of observed QRS complexes )
data: values of R peaks
11. S_loc:
size: 1xnQRS vector (nQRS is number of observed QRS complexes )
data: locations of S peaks
12. S_value:
size: 1xnQRS vector (nQRS is number of observed QRS complexes )
data: values of S peaks
13. Q_loc:
size: 1xnQRS vector (nQRS is number of observed QRS complexes )
data: locations of Q peaks
14. Q_value,
size: 1xnQRS vector (nQRS is number of observed QRS complexes )
data: values of Q peaks
15. T_loc:
size: 1x(nQRS-1) vector (nQRS is number of observed QRS complexes )
data: locations of T peaks
16. T_value:
size: 1x(nQRS-1) vector (nQRS is number of observed QRS complexes )
data: values of T peaks
17. P_loc:
size: 1x(nQRS-1) vector (nQRS is number of observed QRS complexes )
data: locations of P peaks
18. P_value:
size: 1x(nQRS-1) vector (nQRS is number of observed QRS complexes )
data: values of P peaks
FLOWCHARTS:
QRS Detection :
Rhythm Detection:
RESOURCES:
-
ECG Data
i) Physionet ATM Toolkit : https://physionet.org/cgi-bin/atm/ATM. This code is tested on 1 hour ECG signals from MIT-BIH Arrhythmia Database and QT Database
RESULTS:
On QT Database, Physionet: (Overall Accuracy)
QRS Complex : 98.9%
P Wave : 96.5%
Q Wave : 94.3%
PQRST Detection:
Irregular Episode Detection on AFIB ECG Signal
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent