How is Artificial Intelligence changing lives of epileptic patients?
Epilepsy damages lives of suffering patients and their families and if we can diagnose, predict, and prevent it, that would be a game changer for mental health. Anti-epileptic medicines have several side-effects. Imagine a world where epileptic patient had to take fewer medications or simply could have an implant which could prevent epileptic seizure. All of that would require few tiny sensors on skull, hidden underneath patients’ hair and a smart phone which receives and processes the information, either offline or online, and activates a probe to prevent seizure. Well we have obviously oversimplified here but we want you to get the picture without getting lost in technical jargon first. We at Khurana group have a team working on it day and night out. While we are very proud of our efforts but there are other friends (or if you prefer to call them competitors) who are doing commendable job in the same direction for years. In this article, we would take you through what products are already out there but before that let us take you through some basics of epilepsy.
Epilepsy varies from a brief loss of awareness to longer periods of a loss of sense combined by the muscle stiffening and jerking. General symptoms of epilepsy are unresponsive and uncontrollable movements such as repetitive jerking, loss of consciousness which might include loss of bowel or bladder control and unusual behavior such as continuous mood swings. According to WHO, about 50 million people worldwide are suffering from epilepsy. Based on our knowledge of causes, we can classify epilepsy into Idiopathic epilepsy, where the underlying causes are unknown and symptomatic epilepsy, which are due to known causes such as, head injury, genetic disorders, infection, strokes, and tumors.
Based on the complexity of seizures, epilepsy seizures are of two types: Focal Seizure and Generalized Seizure.
Focal Seizure:
If the electrical surge happens in a limited area of the brain, it is Focal Seizure. There are two types of Focal Seizure:
- Simple Partial Seizure: Patients have strange movements and uncontrollable jerky movements but the patient remains aware of surroundings.
- Complex Partial Seizure: Besides strange and uncontrollable jerky movements, patients face changes in awareness, responsiveness, and consciousness.
Generalized Seizure:
If the electrical surge happens in the full brain, it is Generalized Seizure. Generalized Seizure is further classified into 6 categories:
- Absence Seizure: It is generally found in children. In this type of seizure, patients suffer a brief loss of awareness often manifested by a blank stare with or without body movements like blinking, lip smacking etc. The patient may not be aware if something is wrong with them.
- Tonic Seizure: In this type of seizure patients generally suffer from stiffening of muscles. This may cause the patients to fall, generally backward.
- Atonic Seizure: In this type of seizure patients suffer from a sudden loss of muscle tone. This may cause the patients to collapse or fall down.
- Clonic Seizure: In this type of seizure patients suffer from a continuous jerking of muscles. It is a rarely found seizure. The muscles that are generally affected are of neck, face, arms, and legs.
- Myoclonic Seizure: In this type of seizure patient suffer from twitch or sudden jerk of an order of 0.1 seconds. In Clonic Seizure the jerking occurs for a longer duration of time.
- Tonic-Clonic/Convulsive Seizure: It is the most common and dramatic type of seizure. Patients suffer from a combination of muscle stiffening and jerking. It involves a sudden loss of consciousness and bladder control. This type of seizure lasts longer than 5 min, thus requiring an immediate treatment.
Diagnosis and Treatment:
For the diagnosis of Epilepsy, the doctor will review the symptoms and the medical history of the patient. Besides this, there is a test called Electrical Encephalogram (EEG) in which the electrical activities of the brain are recorded and tested to look for any abnormal brain waves. When we talk of our small probes underneath hair, we are riding on the shoulder of this giant (and not a new) development.
Standaard treatments for epilepsy include taking anti-epileptic medications like phenytoin, carbamazepine, and valproate, which reduce seizures or their effects. Other than this, people also follow a ketogenic diet that is less in carbohydrate. If the seizure is not controlled by above methods, doctors recommend the surgical approach which might involve removing affected part from the brain.
According to data provided by Epilepsy Foundation, one-third of the people suffering from epilepsy are suffering from uncontrollable seizures because no available treatment works for them. Understandably, these patients face a major risk factor due to the unpredictability of seizures. According to 2016 survey by Epilepsy Foundation, regardless of seizure frequency and type, selected unpredictability of seizures as a top issue. Thus, Epilepsy prediction is one of the most intriguing problems for the researchers today.
On the basis of temporal analysis of brain activity, it can be divided it into four states: Interictal (between seizures, or baseline), Preictal (prior to seizure), Ictal (seizure), and Postictal (after seizures). The main aim of prediction system is to detect Preictal stage.
For some of the kinds of the seizure, when they occur, changes in heart rate, breathing patterns, and skin is being reported. So besides EEG one can also look into these parameters.
Our friends in academia and industry are using various technologies related to Big Data, Artificial Intelligence, and IOT to predict epilepsy. Most recently, In December 2017, IBM Research Australia and University of Melbourne have made an important contribution towards personalized seizure detection as described in their research paper titled “Epileptic Seizure Prediction using Big Data and Deep Learning: Toward a Mobile System”. According to IBM research, their AI algorithm is able to predict an average of 69 percent of seizures across patients, including patients who previously had no prediction indicators. This should raise high hopes and we think one of us, whether our friends or us would soon get to accuracy in high 90%. Anyway, before you get carried away with just one wonderful piece of work from friends at IBM, have a look at what others have been up to.
- myCareCentric Epilepsy: It is a wearable technology collaboratively developed by Graphnet Health, Shearwater Systems, and the University of Kent. They describe their product as a combination of a patient portal via a mobile app, integrated digital care records, wearable technology, clinical communication and workflow, data analysis and machine learning tools. They use both patient data from hospitals as well as user data via smart band using various sensors such as heart rate, skin conductance, gyroscope, and accelerometer as well information related to mood, stress levels, alcohol intake, diet and medication compliance. This data is passed to their Microsoft Azure Machine Learning and Microsoft Stream Analytics based algorithms. Users can visualize the results on the patient portal.
- Embrace: It is a smart band technology created by Empatica headed by a professor Rosalind Picard from MIT. Embrace uses an accelerometer to pick up motion data while also using electrodes to monitor skin conductance. A huge spike in the skin conductance indicates an upcoming seizure. This band also has an alerting system which instantly alerts to a list of people whose details are specified by the patient. This is inspiring bit of technology, which should be combined with better prediction system than it has.
- Neuropace RNS System: This is a small matchbox-like device that is being fitted in the brain through the surgical procedure. The RNS system monitors the brain wave activity and once the seizure is detected, it releases inhibitory signals to reduce the wave activity. It is a great start and has all our respect but we think either Neuropace or other companies would soon come out with much smaller implants. It is gigantic compared to what should be possible soon.
- Sami Alert: It is a sleep activity monitor to detect seizures while sleeping. It consists of a remote infrared camera which passes the information to an IOS device. The Sami app records and analyses for seizure activity. When the event is detected it alarms and informs the associated persons. We think this should be an add on to an overall package.
- Smartwatch: SmartWatch is an easy-to-use wristwatch developed by Smart Monitor that continuously monitors movements and instantly alerts connected people when there is any repetitive, irregular shaking motion. Their research partners include names like Stanford School of Medicine and John Hopkins Medicines. This is great wearable tech option and we think such an approach should be combined with EEG monitors.
- EpiWatch: It is an app collaboratively created by Apple and John Hopkins University aimed to predict seizures. This app collects heart rate and movement patterns via Apple Watch. By collecting data before, during, and after a seizure, researchers hope to be able to predict seizures. This app is in early stages. This Apple Watch has inspired us to think, whatever solution we can come up with, should also be integratable in an Apple watch. What better than to to have add ons to existing products patients already like to use.
- T-Jay: This is an Indian product which was showcased at Innovate Digital India challenge 2015 by Intel and DST. It is a wearable glove that senses 11 types of signals such as temperature, respiration, and blood pressure. This device compares the data gathered via sensors with the threshold values, which provides an indication of an epileptic seizure. T-Jay was one of the top 10 finalists in the contest. We like the technology here but we are not sure of using something so obvious as a glove would be a preferred solution by patients, who can have access to more inconspicuous technology.
There are many more technologies that are emerging for seizure detection and prediction. With constantly evolving technology, there will be better solutions to approach this problem in the future. We hope for better future of our friends and ourselves to soon have prediction rates approaching hundred percent and false alarm rates diving close to zero.
— —
About:
Manav Bagai is an engineer in Dr. Khurana’s epilepsy diagnosis and prediction team.
Dr. Sukant Khurana runs an academic research lab and several tech companies. He is also a known artist, author, and speaker. You can learn more about Sukant at www.brainnart.com or www.dataisnotjustdata.com and if you wish to work on biomedical research, neuroscience, sustainable development, artificial intelligence or data science projects for public good, you can contact him at skgroup.iiserk@gmail.com or by reaching out to him on linkedin https://www.linkedin.com/in/sukant-khurana-755a2343/.
Here are two small documentaries on Sukant and a TEDx video on his citizen science effort.
