PulseTrendingJournal ClubResearchersJournalsExplore
Instagram
HomeTrendingJournal ClubExplore
Synapse
⌘+K
Synapse
November 14, 2022PLoS ONEOpen Access

Textural feature based intelligent approach for neurological abnormality detection from brain signal data

View Full Paper
Ask AI
Bookmark
Share

Why the study?

Visual analysis of EEG data is time-consuming and subjective, and existing classification models typically focus on single diseases rather than a unified framework for multiple neurological disorders.

Population

Four real-time EEG datasets

Comparison

Four machine learning-based classifiers applied to textural features of spectrogram images

Design

Machine learning development and validation study

Key result

A machine learning framework utilizing tCENTRIST feature extraction and a Support Vector Machine classifier achieved an overall accuracy of 88.78% in classifying four neurological abnormalities and healthy controls from EEG data.

Authors

MTMd. Nurul Ahad TawhidSSSiuly SiulyKWKate Wang

Discussion

Loading...

Member takes

Overview

May aid automated EEG interpretation in neurology; leaves open external validation before clinical adoption.

Structured PICO

P
Population
86 subjects from four public datasets, including patients with autism spectrum disorder, epilepsy, Parkinson's disease, schizophrenia, and healthy controls, used to evaluate an EEG-based machine learning classification framework.
I
Intervention
Machine learning framework using cCENTRIST and tCENTRIST textural feature extractors on short-time Fourier transform (STFT) spectrogram images, classified using Support Vector Machine (SVM), k-Nearest Neighbor (kNN), Random Forest (RF), and Linear Discriminant Analysis (LDA).
O
Outcome
Classification performance measured by sensitivity, specificity, precision, F1 score, and accuracy using five-fold cross-validation.

A unified machine learning framework using spectrogram images and textural features can effectively classify multiple neurological abnormalities from EEG data with high accuracy.

Limitations

  • Requires downsampling and channel reduction to match the dataset with the minimum number of channels
  • Relatively small sample sizes in the individual public datasets used for evaluation

Cite This Study

Tawhid et al. (2022) studied Neurological abnormalities (Autism Spectrum Disorder, Epilepsy, Parkinson's disease, Schizophrenia) (n=86). tCENTRIST feature extraction with Support Vector Machine (SVM) classifier vs. Other machine learning classifiers (cCENTRIST, kNN, RF, LDA) was evaluated on Classification accuracy for five-class categorization (ASD vs EP vs PD vs SZ vs HC). A machine learning framework utilizing tCENTRIST feature extraction and a Support Vector Machine classifier achieved an overall accuracy of 88.78% in classifying four neurological abnormalities and healthy controls from EEG data.

synapsesocial.com/papers/6a2086a747fdc8d429f427c3https://doi.org/10.1371/journal.pone.0277555
View Full Paper
Ask AI
Bookmark
Share