Algorithm for the Parkinson’s disease behavioural models characterization using a biosensor

• Author(s): Pimentel, Angela Bairos
• Date: 2012
• Persistent ID: http://hdl.handle.net/10362/8443
• Origin: Repositório Institucional da UNL
• Subject(s): PD; Zebrafish; MOBS; Behaviour; Machine learning; Zero crossing rate
• Document Type: Master thesis
• Language: English
• Advisor(s): Gamboa, Hugo; Correia, Ana
• Contributor(s): RUN

Description

Dissertação para obtenção do Grau de Mestre em Engenharia Biomédica

The neurodegenerative disease, Parkinson’s Disease (PD) constitutes a major health problem in the modern world, and its impact on public health and society is expected to increase with the ongoing ageing of the human population. This disease is characterized by motor and non-motor manifestations that are progressive and ultimately refractory to therapeutic interventions. The degeneration of dopaminergic neurons emanating from the substantia nigra is largely responsible for the motor manifestations. Thus, understanding the behaviour related to this disease is an added value for the diagnosis and treatment of PD. Also, in vivo models are essential tools for deciphering the molecular mechanisms underpinning the neurodegenerative process. Zebrafish has several features that make this species a good candidate to study PD. In particular, the occurrence of behavioural phenotypes of treated animals with neurotoxin drugs that mimic the disease has been investigated. And, an electric biosensor, Marine On-line Biomonitor System (MOBS) is being used for the real-time quantification of such behaviour. This equipment allows quantifying the fish movements through signal processing algorithms. Specifically, the algorithm is used for the evaluation of fish locomotion detected by a series of bursts in the domain of MOBS that correspond to the zebrafish tail-flip activity. In this thesis we proceeded to the development of an algorithm affording a electrical signal discrimination between "healthy" and "ill" zebrafish and consequently improving the detection of parkinsonism-like phenotypes in zebrafish. The first approach was the improvement of the existent algorithm. However, the first analysis failed to distinguish between different behavioural phenotypes when fish were treated with the neurotoxin 6-hydroxydopamine (6-OHDA). Consequently, we generated a new algorithm based on Machine Learning techniques. As a result, the novel algorithm provided a classification over the health condition of the fish, if the same is "healthy" or "ill" with its respective probability and the level of activity of the fish in number of tail-flips per minute. The method Support Vector Machine (SVM)was useful for the classification of the fish events. The zero crossing rate parameter was used for the characterization of the swimming activities. The algorithm was also integrated in the platform Open Signals, and for a faster evaluation of the signals, the algorithm implementation included parallel programming methods. This algorithm is a useful tool to study behaviour in zebrafish. Not only it will allow a more realistic study over the PD research area but also test and assess new drugs that use zebrafish as animal model.