An algorithm developed by IST researchers won the third edition of the ISLES challenge 2017, an international competition that aims at developing methods for the identification of brain areas affected by ischemic stroke lesions from magnetic resonance data.
The method won the first place (ex aequo) among the completely automatic methods, which perform the diagnostic without human intervention. “We used fully convolutional neural network (a deep learning algorithm), which has several uses in the field of computer vision”, explains engineer Miguel Monteiro. Deep learning techniques were used to infer classifiers that determine the affected brain areas by ischemic stroke lesions.
This method was developed by engineer Miguel Monteiro, researcher at INESC-ID, in collaboration with professor Arlindo Oliveira within the PRECISE project and was presented at MICCAI 2017 (Conference on Medical Image Computing and Computer Assisted Intervention) held in Quebec. The development of methods for medical diagnostic based on deep learning techniques have the potential to create a revolution in the area, since they reach human level performance at a fraction of the cost.
“It’s always good to be recognised for a good work, especially when you are competing with top international institutions,” points out the researcher. “The research is still at a very early stage. In the future we look forward to continue working in the area of computer-assisted medicine for both ischemic stroke lesions and other diseases,” adds engineer Miguel Monteiro.
The project PRECISE, funded by FCT, is developed by researchers from Técnico, INESC-ID, IMM and Faculdade de Medicina, all from Universidade de Lisboa. Accelerating progress on Portuguese healthcare system toward a new era of precision medicine, applying personalised treatment and prevention strategies to each patient is the main goal of this research project. The development of methods for medical diagnostic based on deep learning techniques have the potential to create a revolution in the area, since they reach human level performance at a fraction of the cost. “Collaboration between medicine and engineering isn’t something new. As far as I know collaborations of this kind have always happened,” says the researcher. “These techniques can be useful to help health care professionals in clinical decision-making” he adds.