About one-third of the total disease burden in developing countries results from brain disorders such as epilepsy, schizophrenia, and depression. These neuropsychiatric disorders are prevalent in low- to middle-income countries due to various factors, e.g. difficult births, malnutrition, and exposure to infectious diseases and toxins. Around 80 percent of the world’s epilepsy occurs in low- to middle-income countries, but only 20 percent of people get treatment. The physician-to-patient ratio can be as low as one for every 20,000 people in those countries, with even fewer psychiatrists and neurologists, causing a so-called treatment gap.1 However, timely diagnosis and treatment of epilepsy is possible and can make a difference.2
Last fall, partnering with the Nanyang Technological University (NTU) of Singapore, we took the first steps in tackling this challenge in our Science for Social Good program. Our team included a Social Good Fellow from Columbia University, several machine learning and cloud computing researchers from IBM Research, and collaborators from NTU. Together, we came up with a cloud-based automated machine learning approach to provide decision support for non-specialist physicians in electroencephalography (EEG) analysis and interpretation.
EEG is a useful tool in epilepsy diagnosis. It is a non-invasive, portable, and relatively inexpensive technique that detects tiny electrical impulses at electrodes placed on the scalp as neurons communicate with each other. EEG bands have been found to be useful in detecting, monitoring, and treating a variety of brain disorders, including epilepsy. An abrupt change from the typical pattern of EEG is often associated with epileptic seizure. However, a highly trained neurologist is required to analyze EEG signals as they are easy to misinterpret. This problem can be addressed by teleEEG, which uses remote EEG analysis by experts or by local computer processing of EEGs. However, both of these options are prohibitively expensive and require abundant computing or health infrastructure, which is scarce in developing countries.
This situation urgently calls for development of an accurate machine learning algorithm, which can be used for EEG-based detection of brain disorder and monitoring of brain function with minimal human intervention.
In the proof-of-concept study performed last year and presented at the NIPS 2017 Workshop on Machine Learning for the Developing World, we developed a pipeline that includes moving EEG data to the cloud and getting optimal deep neural net models for various classification tasks. Named neurology-as-a-service, the approach requires almost no manual intervention in feature engineering and in the selection of an optimal architecture and hyperparameters of the neural network.
This project is an example use case for IBM’s recently announced Deep Learning as a Service (DLaaS). We demonstrate the performance of our proposed approach using an EEG dataset for motor imagery movement in a brain-computer interface experiment. Our service attains 63.4 percent accuracy on classifying real vs. imaginary activity, which is significantly higher than what is obtained with a shallow approach such as support vector machines. The proposed approach can therefore be used as a clinical decision support system by providing real-time feedback to aid in diagnosis. Our initial prototype has been tested only in developed world environments to-date, but NTU aims to test it in developing world environments soon.
- NIH Fogarty Newsletter, 13:1 (2014): 8.