System Bits: July 11

An algorithm to diagnose heart arrhythmias with cardiologist-level accuracy
To speed diagnosis and improve treatment for people in rural locations, Stanford University researchers have developed a deep learning algorithm can diagnose 14 types of heart rhythm defects better than cardiologists.

The algorithm can sift through hours of heart rhythm data generated by some wearable monitors to find sometimes life-threatening irregular heartbeats, called arrhythmias. The algorithm reportedly performs better than trained cardiologists, and has the added benefit of being able to sort through data from remote locations where people don’t have routine access to cardiologists, the researchers said.

Stanford researchers say their algorithm could bring quick, accurate diagnoses of heart arrhythmias to people without ready access to cardiologists. (Source: Stanford University/Getty Images)

Awni Hannun, a graduate student and co-lead author of a paper on this subject said, “One of the big deals about this work, in my opinion, is not just that we do abnormality detection but that we do it with high accuracy across a large number of different types of abnormalities. This is definitely something that you won’t find to this level of accuracy anywhere else.”

People suspected to have an arrhythmia will often have an electrocardiogram (ECG) performed in a doctor’s office but if it doesn’t reveal the problem, the doctor may prescribe the patient a wearable ECG that monitors the heart continuously for two weeks, and then the resulting hundreds of hours of data would then need to be inspected second by second for any indications of problematic arrhythmias, some of which are extremely difficult to differentiate from harmless heartbeat irregularities.

Andrew Ng, an adjunct professor of computer science — who is also VP and Chief Scientist at Baidu — saw this as a data problem. He led researchers in the Stanford Machine Learning Group in the development of a deep learning algorithm to detect 14 types of arrhythmia from ECG signals. They collaborated with the heartbeat monitor company iRhythm to collect a massive dataset that they used to train a deep neural network model. In seven months, it was able to diagnose these arrhythmias about as accurately as cardiologists and outperform them in most cases.

The researchers said they believe this algorithm could someday help make cardiologist-level arrhythmia diagnosis and treatment more accessible to people who are unable to see a cardiologist in person. Ng thinks this is just one of many opportunities for deep learning to improve patients’ quality of care and help doctors save time.

Lab-on-a-chip finds sepsis in a single drop of blood
A team of researchers from the University of Illinois and Carle Foundation Hospital in Urbana, Illinois have completed a clinical study of a portable device that can quickly find markers of deadly, unpredictable sepsis infection from a single drop of blood.

University of Illinois researchers and physicians at Carle Foundation Hospital developed a rapid test for sepsis that counts white blood cells and certain protein markers on their surface to monitor a patient’s immune response.
(Source: University of Illinois)

The team said this is the first lab-on-a-chip device to provide rapid, point-of-care measurement of the immune system’s response, without any need to process the blood. This can help doctors identify sepsis at its onset, monitor infected patients and could even point to a prognosis, according to research team leader Rashid Bashir, a professor of bioengineering at the U. of I. and the interim vice dean of the Carle Illinois College of Medicine.

The small, lab-on-a-chip device counts white blood cells in total as well as specific white blood cells called neutrophils, and measures a protein marker called CD64 on the surface of neutrophils. The levels of CD64 surge as the patient’s immune response increases.

‘Brain training’ app improves memory in people with mild cognitive impairment
A ‘brain training’ game developed by researchers at the University of Cambridge could help improve the memory of patients in the very earliest stages of dementia.

Amnestic mild cognitive impairment (aMCI) has been described as the transitional stage between ‘healthy aging’ and dementia, which is characterized by day-to-day memory difficulties and problems of motivation. At present, there are no approved drug treatments for the cognitive impairments of patients affected by the condition, according to the team. However, cognitive training has shown some benefits, such as speed of attentional processing, for patients with aMCI, but training packages are typically repetitive and boring, affecting patients’ motivation.

To overcome this problem, researchers from the Departments of Psychiatry and Clinical Neurosciences and the Behavioral and Clinical Neuroscience Institute at the University of Cambridge developed ‘Game Show,’ a memory game app, in collaboration with patients with aMCI, and tested its effects on cognition and motivation.

The researchers randomly assigned forty-two patients with amnestic MCI to either the cognitive training or control group. Participants in the cognitive training group played the memory game for a total of eight one-hour sessions over a four-week period; participants in the control group continued their clinic visits as usual.

In the game, which participants played on an iPad, the player takes part in a game show to win gold coins. In each round, they are challenged to associate different geometric patterns with different locations. Each correct answer allows the player to earn more coins. Rounds continue until completion or after six incorrect attempts are made. The better the player gets, the higher the number of geometric patterns presented – this helps tailor the difficulty of the game to the individual’s performance to keep them motivated and engaged. A game show host encourages the player to maintain and progress beyond their last played level.

The results showed that patients who played the game made around a third fewer errors, needed fewer trials and improved their memory score by around 40%, showing that they had correctly remembered the locations of more information at the first attempt on a test of episodic memory. Episodic memory is important for day-to-day activities and is used, for example, when remembering where we left our keys in the house or where we parked our car in a multi-story car park. Compared to the control group, the cognitive training group also retained more complex visual information after training. 

In addition, participants in the cognitive training group indicated that they enjoyed playing the game and were motivated to continue playing across the eight hours of cognitive training. Their confidence and subjective memory also increased with gameplay. The researchers say that this demonstrates that games can help maximize engagement with cognitive training.

Screenshot from Game Show. (Source: University of Cambridge/Sahakian Lab)