Abstract:
Electroencephalogram (EEG) can be used for the diagnosis of neurologist disorders: Alzheimer’s
disease, depression, dementia, and epilepsy. Manual interpretation of EEG is time consuming and
resource hungry process. An automated diagnosis system would help neurologist to interpret EEG in less
time. EEG data collected from a local hospital along with channel wise annotations of anomalies created
a unique opportunity for the proposed research problem. A hybrid model is proposed to localize
anomalies in each channel of EEG record. Proposed architecture is divided into two steps. First, Deep CNN is trained for detecting abnormal channels. Furthermore, to detect anomaly time from abnormal channels Long Short-Term Memory (LSTM) network is trained.
