Page 120 - 2024-Vol20-Issue2

116 | Assim & Mahmood

ing architectures used in previous research, including prevalence [6]. Patients can benefit from early detection of

convolutional neural networks, recurrent neural net- epileptic seizures since it improves their quality of life and

works, and hybrid models. By emphasizing their strengths lowers their risks [7–9]. Epilepsy is a clinical diagnosis based

and limitations to provide insights into the technical as- solely on a patient’s medical history, as healthcare providers

pects of these methodologies. rarely view the patients seizure activity.

3. Preprocessing Techniques:The review addresses pre- III. EPILEPSY DIAGNOSIS TECHNIQUES
processing techniques used in previous research to im-
prove EEG data quality, such as noise reduction, artifact Effective therapy for epilepsy and seizures depends on a cor-
removal, and feature extraction. This practical informa- rect diagnosis. Diagnostic testing can assist in identifying
tion is crucial for researchers and practitioners working whether and where a brain injury produces seizures. Exam-
in the field. ples of epilepsy diagnosis techniques include:

4. Dataset Challenges:The problem of limited labeled A. Electroencephalogram EEG:
epileptic EEG datasets and present potential solutions. EEG is the most important diagnostic tool for epilepsy di-
Addressing this challenge is crucial for the advancement agnosis [10–14]. It measures brainwaves dynamics and the
of research in this area. brains electrical activity [15–17]. Electrodes are placed on
various brain areas to record EEG signals, as shown in Fig.1.
5. Anticipation of Future direction:By highlighting Different types of seizures are associated with specific EEG
challenges like dataset size and diversity, model in- patterns:
terpretability, and integration with clinical decision
support systems, this review not only discusses cur- 1. Interictal Spikes:
rent issues but also anticipates future challenges in the Brief bursts of high-frequency activity between seizures,
field. This forward-looking approach is valuable for indicating an increased risk of seizure occurrence.
researchers.

6. Contributing to Healthcare Advancements:Ultimately, 2. Ictal Activity:
the papers comprehensive review and analysis con- EEG patterns during a seizure vary based on the seizure
tribute to ongoing efforts to advance the quality of type and location.
epilepsy management. By showcasing the potential
of deep learning techniques, it paves the way for future 3. Slow Waves:
research and innovations in epileptic seizure detection. Low-frequency waves indicating decreased brain activ-
ity after a seizure.
II. EPILEPSY TYPES
4. High-Frequency Oscillations (HFOs):
Epilepsy might be one of four kinds: focal, generalized, un- Fast EEG oscillations associated with epileptic activity
known, or unclassified. A focal seizure begins with a single are often observed with interictal spikes.
point of attention [2]. The term ”focal” has replaced the
phrase ”partial”. The name ”focal” was used since it was Fig. 1. EEG recording system [12].
deemed more accurate and natural than seizures that begin
with a focus [3]. When both brain hemispheres are active
at once, it leads to a ”generalized” seizure. A seizure is
categorized as having an unknown onset if the history and
supporting research do not provide enough information to
classify it as focal or generalized. Including ”unknown onset”
in the classification has the advantage that it” allows” clas-
sification of the remainder of the seizure ”even ”if the onset
is unknown [4]. An unclassifiable category is still used in
seizure classification; although ”unknown” has been included
as a seizure onset type, it is anticipated that it will be used less
frequently [5]. Epilepsy affects people of all ages differently,
with one peak around the age of 5 to 9 years and the other
around the age of 80. There is no gender difference in epilepsy