Artificial intelligence uncovers diseases by analyzing physiological signals

Electrocardiogram and electroencephalogram tell us nothing if they are not analyzed. The Physiological Signal Analysis Group in the University of Oulu is looking for novel ways to interpret various measurements. In addition to studying the signals of the heart, respiration, and the nervous system, they also study expressions, gestures and speech.

Electrocardiogram (ECG) and electroencephalogram (EEG) are well-known examples of physiological signals. But it is not enough just to measure signals. If we want to know what signals tell us about health, they must also be analyzed.

Means for doing this are being developed in the Physiological Signal Analysis Group in the University of Oulu. Their team consists of 15-20 researchers from the fields of medicine and technology. Their task is to write algorithms for calculating and deducing medical facts from signals.

One of the main research subjects is ECG-based heart diagnostics. “One of the most recent breakthroughs is the early risk marker for repolarization, studied by professor of cardiology Heikki Huikuri. The marker reveals an atypical normalization after a beat, as well as the risk for sudden death”, says the director of the Physiological Signal Analysis Group, professor of medical technology Tapio Seppänen.

”This observation has been published on the leading international forums. Heart signal analysis in Oulu has been the foremost in Finland for a long time now. Its basis is the triangle of the University’s technological field, the faculty of medicine, and the unit of sports and exercise medicine.”

The smart mirror reads and analyzes heartbeat and respiration from the face

In heart signal analysis, novel possibilities in home measurement are emphasized: the yesterday’s fruit of cooperation between research in Oulu and the business life – the heart rate monitor – is reaching medical precision.

”Previously, heart rate analysis was only possible to do in a hospital, but recently even home equipment have been able to produce medical information. They can, for example, assist in the rehabilitation program of a heart patient, or use heart rate fluctuation to deduct breath frequency and sleep apnea”, says Seppänen.

Even in that case, it is still the doctor’s responsibility to diagnose diseases. “The patient sends the signal via a cloud service or a similar channel to a doctor, and that way there is no ‘white coat effect’ in the result of the measurement. Wireless transfer of measurement data is one of our research subjects.”

In the future looms signal analysis with a video camera, or even with the patient’s own mobile phone via a ”smart mirror”. The mirror works by looking at the alteration of facial color and the width of the face in time with heart beat and breathing.

”It is already possible to calculate heartbeat and respiratory data with machine vision algorithms”, says Seppänen. “We have an ongoing Academy Project with machine vision group professor Guoying Zhao, where our aim is to find an algorithm so accurate that it can be used for medical purposes and for analyzing atrial fibrillation, for example.”

Researchers in the machine vision group and the physiological signal analysis group have succeeded in developing a way to measure and analyze the heart signal wirelessly with a video camera or a smartphone. Transferring the measurement data wirelessly makes it possible for a doctor to analyze the patient’s condition irrespective of time and place.

Automatic recognition of feelings as resource for psychiatry and physiatrics

The smart mirror idea is connected with another central research field of the analysis group: affective calculation. The term stands for automatic interpretation of feelings and nonverbal messages from expressions, gestures, speaking voice and nervous system signals, for example.

”We have modeled the way voice changes with different emotional states, and the machine vision group have done the same with facial expressions. In calculating emotions, we can now reach the accuracy of an average human listener or viewer. We are also combining different analyses, and we are the first in the world to have connected expressions with the EEG”, says Seppänen.

The field is still close to basic research, and it has not been commercialized in Finland. But the applications are already raising interest. They could be useful in psychiatry, for example, to identify psychoses or depression, or to help autists to communicate.

”Physiatrists and cardiologists are also interested in this. One or our Tekes projects is the automatic recognition of pain, in which speech, expressions and autonomic nervous system signals reveal, for example, the pains of a back patient.”

Nervous system signals and brain signals constitute their own field of research. Among other things, the research group have found a phenomenon in the EEG, which will help evaluate the depth of anesthesia better than before. The method is suitable for use in an operating room, and it offers a basis for an ongoing Tekes project which uses EEG to study the evaluation of brain damages.

”The intensive unit in the Oulu University Hospital is participating in the study. The method has already been patented, and the goal is to commercialize it”, says Seppänen.

The world’s most accurate method for measuring airway resistance

The fourth central field of the Physiological Signal Analysis Group is cardiorespiratory research, ie. research on heart and lung functions. It has applications in diagnosis of allergies, asthma, sleep apnea and chronic obstructive pulmonary diseases.

”Professor of otorhinolaryngology (study of ear, nose and throat diseases), Olli-Pekka Alho would like to have a device to measure nose breathing automatically”, says Seppänen, as he lists the stumbling blocks of previous measurement methods - susceptibility to interference, weak repeatability, and lack of analysis of dynamic response in functional tests.

The group cleared the obstacles by combining different signal analyses. “We developed and patented a technology which measures airway resistance in the upper respiratory tract as a continuing signal. That way, even quick changes can be measured accurately.”

The analysis group are using equipment which is already on the market. The only new things are how they are used, and how analysis is conducted. As examples there are the two mathematic models. The first model combines the movements of the diaphragm and the rib cage in the standard equipment measuring the air current in breathing. The other combines upper respiratory tract air pressure alteration in the respiratory air current, and forms a signal to describe airway resistance.

The result is a novel method for measuring airway resistance by physiological signals. “I can tell you that it is the most accurate method in the world. Moreover, we are the first in the world to have combined it with heart rate analysis. It could be immediately used in hospitals”, says the professor excitedly.

”This has quite a huge business potential. Respiratory measurement has not been utilized in home equipment, but it can be transferred to them whenever companies so wish.”

Text: Jarno Mällinen

Pictures: Juha Sarkkinen

Last updated: 24/2/2017