Table 1 Comprehensive Overview of Studies in Epilepsy Seizure Recognition

Gaowei Xu et al. [12]

CNN-LSTM 1D

Epileptic Seizure Recognition dataset: 500 patients with 5 health conditions.

EEG

ACC on binary: 99.39% ACC on the five class: 82%

The article does not provide interpretability of the model used.

Mengnan Ma et al. [13]

RCNN

Epileptic Seizure Recognition dataset: 500 patients with 5 health conditions. Privated Dataset: 15 patients with 3 health conditions.

EEG

ACC on the three class: 100%

The article uses a database with few patients.

Rubén San-Segundo et al. [14]

CNN-1D

Bern-Barcelona EEG dataset: 5 patients with 2 states. Epileptic Seizure Recognition dataset: 500 patients with 5 health conditions.

EEG

ACC: 98.9% between focal and non-focal signals; 99.5% for classifying non-seizure vs. seizure; 96.5% between healthy, non-focal and seizure; 95.7% when considering healthy, focal and seizure.

The article uses a database with few patients.

Amirmasoud Ahmadi et al. [17]

SVM

Epilepsy Centre at the Bonn University: Five health patients and Five epilepsy patients.

EEG

Performance varies between 94.38% to 99.64% each corresponds to the cases A versus E and D versus E respectively.

The article uses a database with few patients.

Lina Wang et al. [18]

DWT, multi-domain feature extraction and nonlinear analysis

Epilepsy Centre at the Bonn University: Five health patients and Five epilepsy patients.

EEG

ACC: 99.25%

The article uses a database with few patients.

Shen et al. [19]

TQWT and CNN

Database CHB-MIT: 5 males and 17 females

EEG

ACC: 97.57%

The article uses a database with few patients.

Chen et al. [20]

RF + CNN

Epilepsy Centre at the Bonn University: Five health patients and Five epilepsy patients New Delhi EEG dataset: Ten epilepsy patients

EEG

ACC: 99,9%

The study does not mention any limitations related to the generalizability of the proposed model to different patient populations or EEG recording conditions.