Algorithm would help identify Parkinson’s disease
The mathematical formula helps identify basal ganglia, subcortical structures located on the base of the brain and which would help identify the disease.Manizales, 05 de agosto de 2016 — Agencia de Noticias UN-
The main function of the basal ganglia is controlling motor, emotional and cognitive areas in human beings. Therefore diseases related to any of these areas are generally linked to lesions or alterations of these structures.
This inspired a Universidad Nacional de Colombia (UNal) in Manizales Digital Signal Processing and Control Research Group master’s and doctorate candidates to design a brain structure segmentation algorithm for medical imagery using computer vision and machine learning techniques.
“The segmentation algorithm which is ready and working identifies and reports the position, volume and form of basal ganglia, important features of the analysis and for understanding of several pathologies. It also helps discover biomedical markers for neurological disorders which help in the diagnosis of diseases or treatment planning,” said Engineering-Industrial Automation master’s candidate Mauricio Orbes Arteaga, one of its creators.
The algorithm was designed to segment basal ganglia given the importance of following-up and surgical planning of Parkinson’s disease patients. This is a combined effort with participation of the Universidad Tecnológica de Pereira, the Universidad del Quindío and the Neurocentro del Café.
The division of these types of structures is a challenge mainly due to the low contrast of the images and high anatomic variability, making identification difficult, inclusively for clinical experts.
Due to the complexity of the task and the accuracy level required in medical applications, the algorithm is based on a multi-atlas segmentation technique which uses a set of atlases, correctly divided by experts.
The idea is to use machine learning techniques with the purpose of extrapolating structure segmentation of a new patient from a set of atlas based images.
During the development of this new methodology they had two important discoveries such as a new image magnetic resonance representation strategy, which allows a demographic classification of patients of a determined population. Furthermore it was the beginning of a development of other applications such as identifying Alzheimer patients.
Moreover, this project has been awarded in different scientific events. For instance it was nominated as best article for a young researcher at the “International Conference on Image Analysis and Processing 2015”, in Genoa (Italy). Furthermore the method was ranked high in two international segmentation challenges, one with 86% success in subcortical structure segmentation and 89% in neck and head.
The algorithm is part of the segmentation model of the Neuroplanner software which was the result of the research project which originated the study and whose results are currently being used at the Neurocentro del Café.
In this manner, physicians now have a platform that will enable them to segment brain structures with great accuracy, while reducing the human resources involved in this task.
“In the world 65% of the patients which have Parkinson’s disease need to be operated, therefore they need surgical planning and segmentation is also important,” said Orbes Arteaga.(Por: Fin/JDMP/DMH/APBL