Publications

Investigating Population-Specific Epilepsy Detection from Noisy EEG Signals Using Deep-learning Models

HELIYON, 6 November, 2023, [Q1, IF:4.00]

Removed noise, EMG artifacts from raw EEG signal. Utilized statistical features and correlation to select channels to construct an image dataset. Applied transform analysis methods, with ensemble averaging.
Investigated the performance of CNN models and conducted population analysis based on the results.
Investigated different methods of epilepsy detection using EEG signals and deep learning models.

Performance Investigation of Epilepsy Detection from Noisy EEG Signals Using Base-2-Meta Stacking Classifier

[Under Review]

investigated 32 features, the accuracy of single base models, and seven classifier models that exhibited SF-95 (Our new proposed metrics is called SF-A). In these metrics, accuracy implies the cut-off accuracy for the selected model.
Utilized disparate cross-validation values, selected the best-performed models as base classifiers, and used Bagging Classifier as meta classifier which achieved significantly higher accuracy
Utilized a novel Base-2-Meta stacking model for the detection of epilepsy after feature extraction and noise rejection of EEG signal.

Automatic Classification of COVID-19 from Chest X-Ray Image using Convolutional Neural Network

In 5th International Conference on Electrical Information and Communication Technology, IEEE, 16 March, 2022

Images from different publicly available databases were collected.
ResNet18, ResNet50V2, DenseNet121, DenseNet201, modified DenseNet201 and VGG16 were used to detect COVID-19. From the experimental results, modified DenseNet201 showed the best performance with 99.5% mean accuracy, 99.5% mean F1 score, and 100% mean sensitivity in binary (COVID-19 and normal) classification, and 98.33% mean accuracy, 98.34 mean F1 score, and 98.34% mean sensitivity (98% sensitivity for COVID-19) in 3-class (COVID-19, pneumonia, normal) classification.

Glaucoma Detection by Designing a Web Application Based on Color Fundus Images by Using Modified DenseNet201

Grade A+ (4.00/4.00)

Conducted performance analysis of multiple CNN models, utilized data augmentation and dropout to recognize the subtle elements involved in the classification task, such as microaneurysms, exudate, and retinal hemorrhages, and enhanced the performance
Modified DenseNet201 architecture. Generated a web application with an acquired accuracy of 94.55 percent to make Glaucoma detection automatic, accessible, and user-friendly for clinicians.

Page updated

Google Sites

Report abuse