Titulo Estágio
Epileptic seizure detection and prediction based on topographic maps and deep machine learning (EPI-DEEP)
Áreas de especialidade
Sistemas Inteligentes
Local do Estágio
CISUC - Centro de Informática e Sistemas da Universidade de Coimbra
Enquadramento
Humans are able to fuse sensory data from several sources simultaneously over time, and are able to capture critical aspects of data streams. The ability to reproduce artificially this human capacity has been a core challenge in artificial intelligence. In fact, mimicking the human cerebral capacities involves several areas, and is an exciting multidisciplinary field of research. As an example, neuroscientists found that neocortex is organized in a modular hierarchy of modules that overtime learn to represent observations based on their regularities1,2. Aiming to model this behaviour, and thus produce artificial learning, a class of machine learning methods emerged, called deep machine learning. State-of-art results have been obtained by deep machine learning methods in several spatial-temporal tasks such as speech and image recognition, and are already been used in real-life applications by companies such as Google, Microsoft and Facebook , .
Despite available drug and surgical treatment options, more than 30% of patients with epilepsy continue to experience seizures. It would be of great benefit to improve the quality of life of these patients if a method existed to warn the patient of an impending seizure or trigger an antiepileptic device to prevent seizure occurrence. Traditional machine learning algorithms, based on electroencephalogram (EEG) features, have been applied to discriminate the different brain states, namely feed-forward artificial neural networks (FF-ANN) and support vector machines (SVM). The results obtained with FF-ANN and SVM yielded some success on predicting seizures, but far from the performance required for a real clinical application. Thus, new approaches are needed to achieve the quality levels required for the clinical use of seizure detection and prediction algorithms. An unexplored aspect is the long-term analysis of spatial-temporal patterns obtained from topographic maps and that are expected to bring encouraging results. Topographic maps are spatial representations of a given electroencephalogram feature over the scalp, and that may change over-time, exhibiting the cerebral dynamics.
Objetivo
This master dissertation is aimed at applying deep machine learning networks to identify patterns related to seizure events based on topographic maps and successfully predict seizures. Topographic maps will be extracted from EEG data. The resulting maps will then be used to train deep machine learning techniques. In particular, the student is expected to compare state-of-the-art networks, such as Convolutional Neural Networks (CNN), Deep Belief Networks (DBNs) and/or Long Short-term Memory networks (LSTM).
Plano de Trabalhos - Semestre 1
1. Literature review
• Objective: In the first phase the student should understand the problem of seizure detection and prediction by computational intelligence methods, and understand the main deep-learning approaches.
• Start: September 2018
• End: November 2018
• Expected Result: A report describing the actual formulations of seizure prediction by computational intelligence methods, and the main deep learning approaches suitable for the epileptic seizure detection and prediction problem.
2. Implementation of deep learning methods
• Objective: Development of a toolbox for seizure detection and prediction by deep learning and topographic maps.
• Start: October 2018
• End: December 2018
• Expected Result: A toolbox that should enable the development of detectors/predictors based on topographic maps and deep-learning.
Plano de Trabalhos - Semestre 2
2. Implementation of deep learning methods (Cont.)
• Objective: Development of a toolbox for seizure detection and prediction by deep learning and topographic maps.
• Start: January 2019
• End: March 2019
• Expected Result: A toolbox that should enable the development of detectors/predictors based on topographic maps and deep-learning.
3. Analysis of the algorithms in a large set of patients
• Objective: Using the developed toolbox, predictors for 100 patients from the EPILEPSIAE database will be developed.
• Start: February 2019
• End: June 2019
• Expected Result: A report describing the results obtained in the 100 patients considered.
4. Thesis writing
• Start: May 2019
• End: July 2019
Condições
The work is supposed to be developed on the Adaptive Computation Group Laboratories-CISUC, under normal logistic conditions.
Observações
Depending on the work quality, publications in international conferences may occur.
Orientador
César Alexandre Domingues Teixeira
cteixei@dei.uc.pt 📩