Faisal Ahmad Siddiqui

Academic

Career

Verbatim text

Group Members: Adan Nazir i19-1680 Arsalan i19-1885 Faisal Ahmad Siddiqui i19-1674

SeizureDL
SeizureDL is an end user tool to analyze and diagnose Epileptic Seizure in patients by using Electroencephalography (EEG) signals fetched from brain using EEG device. We used CHB-MIT dataset, which is a collection of recordings of 22 pediatric subjects with intractable seizure. We trained a custom Neural network using Hyper parameter tuning and Sequential connections to predict the onset of seizure in a target input patient. SeizureDL has a training accuracy of 90.18% while a test accuracy of 90.72% and has a precision, recall and F1 score of 0.87,0.92 and 0.89. We compared our evaluation metric with the current SOTA available models, and we gained on-par results for our research. We used Z-scoring and weight quantization technique to normalize our data in the pre-processing phase of our pipeline and avoid overfitting while training our model.
Features included:
- Providing an end user model which will be hosted on web and generates real time results, with a graph
Real-time analysis of Seizure data
A multi-model approach for providing multiple prediction for the same live data
Empirical investigation of the available SOTA techniques

Objective
An EEG based web application that will predict Seizure using multimodal based approach.

Workflow

Timeline
Sept-Oct
- Literature review of processing EEG signals
- Literature review of seizure prediction techniques
- Literature review of types of epileptic seizures, their causes and effects
Nov-Dec
- Dataset front
- Preprocessing of Dataset
- Model implementation
Jan-Mar
- Website and UI design
- Integrating forward and interface with backend algorithm
- Further improvement in model efficiency
- Deployment of models
Apr-May
- Final evaluation of results
- Finalize Research Paper for publishing

Tools

Technology Used:
Python, Google Colab, Tensorflow, Keras, HTML, CSS, Flask
Supervisor Name:
Ms. Humera Sabir
Group Members:
Adan Nazir (i19 - 1680)
Faisal Ahmad Siddiqui (i19 - 1674)
Arsalan (i19 - 1885)

AI enrichment