If the Heart Could Speak
Start-ups Heart sound analysis
A new Siemens spin-off has developed diagnostic algorithms that can listen to heart sounds, identify where they were produced, and interpret their probable causes. Spin-off partner SPEEDUS is working on an Internet-accessible database for archiving the sounds and harvesting information from them.
Raymond Watrous, developer of an algorithm that can identify heart sounds
It's time for your annual check-up. Your shirt is off and you're lying on the examination table with its crinkly paper cover. Finally, the doctor arrives, and the metallic head of his stethoscope is pressed against your chest. He listens carefully for a few moments: contraction, relaxation; contraction, relaxation. Four chamberstwo atria and two ventriclesplus four valves, all working together to pump about five liters of blood throughout your body seventy to eighty times a minute. There's a lot to listen to. And the sounds are inherently hard to decipher because they are faint and lie at the lower end of the audible frequency range. Furthermore, the sounds of the heart are difficult to discriminate because they are separated from one another by less than 30 msan exceedingly short time for human analysis.
How, then, do doctors distinguish innocent from pathological heart murmurs? The answer is that many of them don't. Numerous studies have shown that extremely high percentages (as much as 87 %) of patients referred to cardiologists for evaluation have benign murmurs. In other words, the general practitioner (GP) heard something, but a follow-up examination showed that it wasn't anything to get excited about. That may be good news for patients, but it certainly isn't what efficient medical care is all about. Typically, a visit to a cardiologist, including ultrasound work-up and the cost of a technician, runs from $300 to $1,000 in the United States; and there are probablyno statistics are as yet availablehundreds of thousands of such avoidable referrals each year. Not to be forgotten, on the other hand, are cases that go undetected. Many forms of heart disease may be asymptomatic for years. Left untreated, however, they may require surgery.
Against this backdrop, Princeton, New Jersey-based Siemens Corporate Research (SCR) has developed unique, intelligent algorithms that can analyze heart sounds and deliver clinical findings. At the patient end of what SCR developed is Tricorder, a prototype system that uses advanced signal processing algorithms to interpret acoustic patterns. "Tricorder analyzes heart sounds and derives important clinical information from them." says Raymond Watrous, Ph.D., a former Distinguished Member of the Technical Staff at SCR's Signal Processing Department and now Chief Technology Officer of Zargis Medical Corporation. Normal heart sounds, explains Watrous, are generated by the opening and closing of valves as the heart contracts and relaxes and by the free flow of blood through vessels and chambers. But if the valves stiffen as a result of disease they can produce a snapping sound that has a unique acoustic signature. Similarly, constrictions will produce distinctive sounds caused by turbulent blood flow at given moments. "These abnormal sounds are analyzed in terms of their duration and amplitude," says Watrous. "Based on all this information, Tricorder's patent-pending algorithms can identify the most likely source of the anomalies by determining which phase of heart motion they were produced in."
With this technology in mind, the algorithms' developers tapped experts within SCR's own business and technical team and at Brooklyn, N.Y.-based telecom company SPEEDUS.Com, Inc. to investigate the possibility of establishing a start-up company. In January, 2001, SCR and SPEEDUS founded Zargis Medical Corporation, which is now located in Princeton. "Our goal," says Shahram Hejazi, a Member of the Zargis Board of Directors, "is to develop advanced diagnostic products and services to enable primary care physicians to detect heart abnormalities." Based on Tricorder technology, the company plans to develop a product* that captures heart sounds and sends them to a nearby PC. The PC will then analyze the sounds and display a detailed but easy-to-understand graphic representation of its findings (see figure). Armed with the resulting information, the average general practitioner will then be in a position to make an informed decision regarding the need for referral to a cardiologist. Should there be any doubt, the informationincluding associated sound filecould be forwarded to the cardiologist prior to the consultation. Furthermore, plans call for an archive to be established for each patient. This will allow specialists to compare older heart sounds with new ones, thereby providing an objective way of monitoring previously diagnosed conditions. Eventually, the database will become an information reservoir from which cardiologists will be able to interrogate expert systems and extract new data pertinent to their patients.
Tricorder's phonocardiogram shows the results of an examination in four sections.Top: complete waveform showing collective cardiac activity;S1: systolic phase (contraction);S2: diastolic phase (relaxation); and murmur detection. At bottom left is the system's interpretation
In a nutshell, that's the idea behind Zargis; and marketing studies indicate that the idea will fly. "What GPs really want is an inexpensive tool that will help them interpret heart sounds in a simple way," says Silvano Dall'Asta, Chairman of the Zargis Board. "On the other hand, cardiologists will welcome a technology that allows them to concentrate on those patients who really need their help while giving them the ability to refer to a specialized database." With all that it has to offer, Tricorder and its connection to a central archival database may be literally just what the doctor ordered. "The arithmetic looks good," says Dall'Asta, who is also Chief Financial Officer at SCR. "Given the number of annual physical examinations in the U.S. and a reasonable fee for service, we see a $2 billion market. Worldwide, it could be several times that figure."
Arthur F. Pease