How it’s generates a brain disease and to seek a solution, is the aim of deciphering the human brain connectome study. The scientific challenge is even more complex than the study of the genome, as explained the Biomedical Engineer, Nicole Labra. In her report of title, Nicole improved an algorithm classification of white matter fibers, work led by Professor Miguel Figueroa, in Optoelectronics Division of our Center.
The brain functioning remains a mystery, and in our country is a research line virtually unknown. “the field is still in its infancy in Chile”, Nicole Labra recognized. The classification of brain white matter fibers by non-invasively way allow in vivo study of human brain connectome, “it’s something that carries out many years of research in other countries, because since the 2000′ are no studies on the matter, however there are many professionals health in Chile don’t know the issue”, she said.
The algorithm optimized by Nicole can segment and classifies brain fibers identified in a set of data, to know the trajectory and grouping these fibers. It was developed by Professor of the Faculty of Engineering, Pamela Guevara during his PhD. in the Université Paris-Sud 11 in France. “One of the complications that had that classification algorithm, which delivers very good results, is its slowness in implementing the data, because it is based on cluster hierarchical in Python and C++ languages, and the problem is that the data set used are very large, which leads to have a very slow algorithm”, said Nicole.
The student’s work was to accelerate the algorithm, using a general-purpose graphics processing unit (GP-GPU), more efficient and economical than using a traditional CPU, to program the algorithm again and to search paralyzed parts, occupying all resources of the GP-GPU. The work improved 119.8 to 286.7 times time of the original algorithm, depending on the amount of data processed. Finally, the student included the optimized algorithm in a graphic application to visualize the fibers listed in 3 dimensions.
Nicole’s work was accepted at two conferences, one in U.S. and one in Japan. The researcher had to decide to attend one and chose the latter instance, the 35th International Conference of the Society of Medicine and Biology. “We are seeking funding for to travel in different institutions, and CEFOP has been a great help, because through their support, now that I’m in the Master in Engineering Science, mention in Electric Engineering, I can devote myself to study and research. Otherwise would have to pay for the trip myself and could not spend as long as necessary”.
Optimize algorithms is a very attractive work for new professionals like Nicole. “I’ve talked to physicists and chemists who have many problems, because they have algorithms that are very good but very slow, and for optimize them were complicated because they don’t have the knowledge and time to devote to this, and then it is an area where work it’s a real need”.
“With CEFOP I learned new things and also I met to researchers developing in depth topics that I job now. They have been a fundamental contribution because they were available to help me and guide me in areas that in Biomedical Engineering don’t exist because my career is more oriented to the management”, she said. With a Masters in Electronic Engineering in course, Nicole’s idea is make faster and didactic the application.
In the long term, scientists, doctors and other professionals will use the application in a user-friendly interface to visualize the fibers. “Now most of the treatments are based on drugs and this is very detrimental because they fix something but they have side effects. In particular in the brain, it is difficult to give a treatment, because the symptoms are confusing, so I think this will be a good tool to diagnose”, he said.