A Fast Machine

Learning Model
for ECG-Based
Heartbeat
Classification
and Arrhythmia
Detection
Aim/ objective

To make advancements in the field of Biomedical science for

a better tomorrow.
 Introduction

● What is an ECG?

● What is Arrhythmia?

● How to diagnose?

● Old ways vs new ones

● Features and factors to be considered

● How to optimize a classifier?

The Databases

Here, we make use of two bases, which are as follows:

● MIT- BIH AR
● AHA AR
Performance Metrics

▪ Sensitivity (Se)

▪ Positive Predictive Value (PPV)

▪ Specificity (Sp)

▪ Accuracy (Acc)
The Heartbeat Classifier

● Stage 1–Processing of the ECG recordings

● Stage 2–Classification between SVEB+ and VEB+ classes

Processing of the ECG and Feature Extraction
Processing of the ECG and Feature Extraction (contd.)

The steps involved in the processing of ECG recordings-

➢ ECG re-sampling
➢ ECG filtering
➢ Heartbeat detection
➢ RR calculation
➢ Heartbeat segmentation
➢ Heartbeat normalization
Reservoir Computing

● “Reservoir” means network stores past inputs.

● We don’t compute the input or hidden weights.
● Training the reservoir involves only training the output weights.
● Reservoir Computing can be used to train only networks satisfying
the echo state property.
 Reservoir Computing

A reservoir network A general neural network

Applications of Reservoir Computing

● Reservoir computing is used to solve chaos problems due to the

input forgetting aspect of the echo state property.
● Neuroscience
● Solve multi-objective problems in machine learning
● Speech Processing
Neural
Networks
Echo State Networks

● What are ESNs?

● Why and when should you use ESN?
● Classification Algorithm…
Classification Algorithm : Echo state
Network
Classification Algorithm : Echo state
Network

In this process, dimensions must change from d × Hb to N × Hb, where d, Hb, and
N are the number of input features,heartbeats, and network neurons, respectively.
The mapping from the input into the reservoir is done through a random input
matrix Win generated from a uniform distribution 2 [−1, 1].

Hence, the ECG data original features vector ud×Hb is modified according to:

XN×Hb = (Win N×d × ud×Hb).

Classification Algorithm : Echo state Network
(contd.)

The reservoir matrix response r for the nth heartbeat for the standard ESN is obtained as
follows:

r(n) = F(gX(n) + hWr(n − 1)),

where W is the random connection square matrix, with dimensions N × N,

F is the ESN activation function and
g and h are the input and connection scaling parameters, respectively.
 Implementation of ESNs

● ReservoirPy

● Steps

1.Provide a random RNN

2.Harvest reservoir states

3. Compute output weights

Parameter Optimization of the ESN for the SVEB+
and VEB+ Classification

● SVEB+
● VEB+
● Optimization
Advantages of ESN over other NN

ESN presents two main advantages over other neural networks.

1. Lesser Computation time- The classical methods involve

relatively time consuming complex computations in the training
phase that are computed relatively easily in ESN.

2. Advantage of recurrent connections- Past heartbeats play a role in

the classification task in the case of ESN due to its intrinsic
memory, having a positive impact on the performance.
THANK

YOU!