Monitoring devices help to detect signs of impending medical crisis
Katie Carr, Coordinated Science Laboratory
- Recognizing the warning signs of an oncoming medical attack is essential to providing quick medical attention.
- ECE student Homa Alemzadeh, ECE Professor Ravi Iyer, and other researchers in the DEPEND group used neural networks to create a reconfigurable and low power device that could detect an oncoming seizure by extracting different features from electroencephalogram (EEG) readings from the brain.
With many life-threatening medical conditions, such as a seizure or heart attack, early detection and treatment can often be the difference between life and death. Recognizing the warning signs of an oncoming attack is essential in providing quick medical attention.
Researchers in the DEPEND group at the Coordinated Science Laboratory (CSL) have been working on this problem and have developed methods that can monitor and analyze vital biomedical signals from individuals with the objective of an early detection of abnormalities, such as brain trauma or warning of an epileptic seizure.
Homa Alemzadeh, a PhD candidate in electrical and computer engineering, and her advisor, ECE Professor Ravishankar K Iyer, were recently recognized for their work in this area in the Fall 2012 Biomedical Computation Review, an NIH-funded magazine published by Simbios, the National Center for Physics-Based Simulation of Biological Structures at Stanford University. Iyer is a researcher in CSL.
“They came across our paper and interviewed us about the use of machine learning algorithms in biomedical devices,” Alemzadeh said. “It is important that our work is recognized along with others doing similar work, such as Medtronic. We are continuing in this direction to develop medical monitoring devices that are adaptable and configurable to different patients’ characteristics and diagnostic needs.”
The Simbios article looked at using embedded medical devices to detect patient-specific data to help detect and treat diseases. They referred to Alemzadeh and Iyer’s work on development of embedded hardware to detect epileptic seizures. The paper, "An Efficient Embedded Hardware for High Accuracy Detection of Epileptic Seizures," was published at the 2010 IEEE International Conference on Biomedical Engineering and Informatics.
Alemzadeh, Iyer, and other researchers in the DEPEND group, including CSL Research Professor Zbigniew Kalbarczyk, used neural networks to create a reconfigurable and low power device that could detect an oncoming seizure by extracting different features from electroencephalogram (EEG) readings from the brain.
“We had previously done work that detected traumatic brain injury for soldiers on the battlefield by analyzing EEG signals, oxygen saturation and body movements,” Alemzadeh said. “The seizure research was a follow-up to that work to see how we can use the same measurements to predict seizures. We’re now making the device reconfigurable so that we can support different types of medical conditions. We’re working on employing the same computations to detect other medical conditions, such as the clinical symptoms leading to heart attacks.”
Alemzadeh is currently focusing on cardiac events, such as arrhythmia. She is looking at how tracking multiple patterns, such as electrocardiogram (ECG) signals, heart rate and blood pressure, will lead to earlier and more accurate detection of abnormalities.
“Our approach is that we want to detect as early as possible,” Alemzadeh said. “If you detect a little earlier, then you can actually prevent a seizure or at least provide an alert to the patient. Even if it’s just one minute before, it’s helpful to be able to provide some medication or receive help from a doctor nearby.”
In addition to developing hardware designs that will be applicable in multiple medical areas, they plan to create portable devices that could be configured for a certain patient and could be dynamically changed to measure different signals depending on the patient conditions.
Previous studies using current generation cell phones to run real-time monitoring software showed that the added software consumed the majority of the phone’s power. Alemzadeh and Iyer are working to solve that problem by designing hardware architecture for real-time biomedical monitoring with low power consumption.
“One important thing is that the medical devices that are out there for detection of critical events now are usually big and costly, are only available in the hospitals and intensive care units and are not affordable for everyone,” Alemzadeh said. “The future direction of health monitoring is going to personalized health care, so everyone could have medical devices in their home or while traveling. We’re wanting to create a device that is cheap, adaptive and reliable."