| Japanese | English |

Biomedical sound analysis and diagnostic system using Internet

Research proposal to TAO (Telecommunication Advancement Organization of Japan). Submitted in May, 2003. 

Introduction

When a patient is examined, physical assessment is very important. Basic physical assessment techniques are Interview, Inspection, Palpation, Percussion, and Auscultation. Inspection means to examine by looking visually. Palpation is examination by touch. The texture, consistency and tenderness of tissue are determined by palpation. Percussion is the technique of tapping the surface of a body to make a determination by sound. Auscultation means examining by listening to the sounds coming from organs. See here for more information. All of these assessments are important. It seems that exact diagnosis cannot be performed unless these information are not provided. 

Recently the medical consultation by the Internet is available, though a clear diagnosis is impossible because no objective information about physical condition is not offered. It seems that medical consultation is the service most desired on the Internet. To help such a service by providing physical data of patients is the main purpose of telemedicine. 

In this project, we especially focus to the sound technology in telemedicine. Diagnosis by sound, such as auscultation and percussion, is lagging behind in research and development compared with image diagnosis. The reason is the transfer network of sound data has not been established. As described below, the sound coming from internal organs gives us much information very effective in diagnosis. The purpose of this research is to carry out the maximum use of the information of sound to contribute to the remote medical treatment.

What does auscultation show us?

Here, the relationships between auscultation sound and disease are described. Auscultation sound was analyzed by DSSF3. 

1. Auscultation of neck

What is heard by the neck auscultation? Much information is contained in the sound of a neck. Many disease can be diagnosed by the auscultation of the neck. 

1. It is said that the air passing a tracheal gill would make unusual wind sound if a malignant neoplasm is made in a bronchus. So the bronchus neoplasm is found by auscultation of the Adam's apple. This sound is called rhonchi.  Rhonchi are low pitched, snore-like sounds. They are caused by airway secretions and airway narrowing. 

Rhonchi sound Measurement of abnormal lung sound

 

1. About the abnormalities of a blood vessel, if the sound of the blood vessel of a carotidartery can be heard around a neck, the head artery (it is the thick blood vessel which sends blood to a brain from the heart) may have started arteriosclerosis. Cerebral infarction may be started as complications of an operation or inspection. It is said that sound like a car running a highway can be heard around the Adam's apple in the case of arteriosclerosis. Moreover, the grade of hardening can be judged by the sound.

2. Auscultation of chest

Auscultation of chest is very common. The organs which make sound are lungs and the heart. Other disease may be diagnosed. The sound of lungs and the heart is very complicated. So the qualities of diagnosis are very different between an experienced doctor and an inexperienced doctor. 

1. Pneumonia 
   This is the state in which bacteria and the virus were infected and bred in lungs. From the state of respiratory sound, it can be diagnosed about the grade and cause bacillus of pneumonia. The feature of sound is coarse crackle, which was analyzed in Lung sound measurement 2. More precisely, it is heard by bronchitis, pneumonia, pulmonary tuberculosis, pulmonary infarction, pulmonary suppuration, a lung congestion, the lung blister, bronchiectasis. Crackle sound is described as fine, popping, crackling, discontinuous, non-musical noises. It lasts about 10-25 ms. 

    

   Coarse crackle  Lung sound measurement 2

   

   Fine crackle Lung sound measurement 3

   

2. Emphysema
   Emphysema is the disease in which respiratory bronchioles and alveoli extend and are destroyed. Consequently, effective lung capacity falls and the respiratory function becomes bad. Emphysema is caused by smoking addictions and is increasing in recent years. 
3. Bronchial asthma
   A characteristic sound can be heard. So it can be found easily. Auscultation may be the most effective diagnostic method. The feature of sound is wheezing. Wheezes are caused by narrowing, constriction, or spasm in the very small airways. They can occur because of asthma, congestive heart failure, fibrosis, pneumonia, and tuberculosis, also. The lasting time of the wheezing sound is about 250 ms. 

   Wheezing sound Lung sound measurement 3

   

4. Pneumothorax
   This is a disease in which a hole opens suddenly in lungs, air leaks in the thorax, and lungs are pressed and shrunk. It can be found immediately since respiratory sound stops. If time is taken by the X-ray examination, dyspnea may advance suddenly and may cause shock condition. This is the illness which should be diagnosed only by auscultation. 
5. Pleurisy
   Pleural membrane is a film covered with lungs. Pleurisy is the state in which this film is infected with bacteria and inflammation arises. This can be diagnosed by auscultation. 
6. Cardiac insufficiency
   This is in the state where the function of the heart is weakened and the flow of the blood of the whole body is overdue. Blood accumulates in lungs and the skin. It can be diagnosed by the medical examination of whole body including auscultation.
7. Valvular heart disease
   The heart has the function which continues sending blood to the one direction. The structure for adverse current prevention called a valve is in each part of heart so that blood may always flow in the fixed direction. If this valve breaks or it becomes hard, it will become that blood cannot flow or flow backwards, finally it will become heart failure. There are four valves in the heart. It can be diagnosed by auscultation which one is abnormal, whether it is broken or harden. 

   Mitral stenosis

   
   

   Mitral regurgitation

   
   

   Both analysis can be seen at Heartbeat measurement 8

8. Congenital cardiopathy
   Many of congenital cardiopathy has a characteristic heartbeat. It can be diagnosed by auscultation.
9. Arrhythmia
   The stroke of a pulse becomes less regular. In many case you may leave it. However, there is a dangerous arrhythmia which causes sudden death. 
10. Pericarditis
    A pericardium is a film like the bag covering the circumference of the heart. If inflammation happens here, a motion of the heart may become bad and may cause heart failure. It is said that pericarditis can be diagnosed by auscultation. 

The belly is a very busy place when a stethoscope is applied. From the state of the sound, the state of abdomen internal organs including the stomach and intestines is guessable. Even if not disease, the function of the internal organs of the belly can be grasped. For example, it can be seen that this man is constipation or this man has arteriosclerosis, etc. 

1. Ileus
   A motion of intestines is understood very well by auscultation. Ileus has big influence on a motion of intestines. In diagnosis of ileus, auscultation is the most important method.
2. Enteritis
   Enteritis is also found by the auscultation. 
3. Coarctation of the aorta
   It is in the state where the abdominal aorta (the thickest blood vessel in the belly) is narrowed. 
4. The abnormalities of other abdomen blood vessels
   By auscultation, an unusual sound called blood vessel noise can be heard. 

Background and concept

Sound is effective also in the broad diagnosis other than medical assessment. Remember that sound is used for investigation of the crack of concrete, or that the appearance of internal organs is drawn by using the ultrasonic sound. The animal in the dark worlds, such as a bat and a owl, uses an ultrasonic sound to understand the external situation. Also, a dolphin and a killer whale in the sea are applying sound for a radar or a communication. It is because a position, a motion, and a state of an object can be understood from the strength of sound, the interval of sound, and a timbre of sound. For such analyses, all the information of sound is needed. To use much information of sound effectively, high time resolution analysis and exact transfer of timbral and spatial nature of sound are indispensable. 

A big problem is being unable to transmit sound information correctly in the present Internet. It is simply considered that the problem of sound quality is solved by accelerating a network. That is, it is supposed that optical fiber and a broadband network will solve all. However, it is the greatest misunderstanding. Transmission of a real time signal is fundamentally impossible in the communication system of the present Internet. Because of the fundamental structure of the relay system, neither delay of sound data nor mixing of interval and noise is avoided. Delay of signal has the error more than number of two figures to the accuracy which biomedical sound analysis needs. Therefore, even if it can transmit language and music somehow, neither precise spatial information nor time information can be transmitted. We can transmit the compressed data, but very important information is lost and it does not become useful. 

Then, it is necessary to newly make the communicative protocol which covers the sound data for medical diagnosis. Otherwise, we have to improve all equipment and standards of Internet. But it is not an actual way out in respect of expense or efficiency. In order to use the structure of the Internet which has spread widely and cheaply as it is, a transmission protocol needs to be developed. We offer the protocol for analyzing sound and transmitting a required sound parameter. It is the idea of sending sound data after analyzing, since the sound data itself sent in real time is not helpful to analysis. The network of the sound for medical diagnosis can be realized by extracting only the required portion of data instead of all data. Data size will be compressed even to an E-mail.

Work plan 

Contents of research and development

The project is divided into four subunits. The first is the development of the portable measurement device shown in the left of the figure. This device will be equipped with the data recorder, the calculation program of sound parameters, and the data transmission and reception program. It is made so that anyone can use easily. The control commands of a calculation program or a transmitting program of operation are defined in the form of the template of every object. Calculation software and templates are maintained at the newest by the on-line update using the Internet. The second is the highly efficient automatic surveillance system. It works on PC, acquires information of sound by using the autocorrelation and the cross correlation analysis in high time resolution, analyzes the important acoustical parameters, and creates templates for controlling efficient communication with the measurement devices. The third is the server system that contains the sound analysis and diagnostic system and templates for controlling the measurement devices. Details of sound analysis such as the high time resolution analysis and the correlation analysis are described later.  The last one is the network system to connect the measurement system, the surveillance system, and the server system. By extracting the necessary information of sound and clarifying the contents, the amount of transmissions is reduced and an efficient network is established. This is the big purpose of this project. Portable measurement device Automatic surveillance system   (Especially for research purpose) Low cost Secure

High time resolution analysis

Conventional sound analysis was biased in favor of a frequency domain too much, and analysis in a time domain was underdeveloped. However, the auscultation sound is an unsteady signal with a large temporal fluctuation. The signal of various time cycles, such as a impulsive sound of valve closure and a breathing sound changing gently, are included.  To catch such a time change of signal, detailed time domain analysis is required. Actually, the sound analysis differs by the time resolution.  For example, the following figure shows time course of a sound level of heartbeat analyzed by the time resolution of 0.1 s (left) and 0.01 s (right). It turns out that the typical opening and closing valve sound and heart murmur are caught exactly by the high time resolution analysis. Thus, new technology aims at catching time change of a more detailed sound signal by the high time resolution analysis with an order of 1ms.  To raise the time resolution of sound analysis, it is necessary to process a lot of data at high speed. For this purpose, the proposed technology processes the signal by the correlation analysis. This analysis technique has advantages in the rate of data compression and the simplicity of the acoustical parameters.  During the project period, basic research will be done for performing automatic measurement in the respectively optimal accuracy about various diagnostic parts, such as the heart, lungs, and belly. Data acquisition of the auscultation sound at the time of normal and abnormal, and sound analysis are performed in detail, and finally the program of automatic measurement is developed. Analysis example of heartbeat with different time resolution (left: 0.1 s, right: 0.01 s).

Autocorrelation and cross-correlation analysis

In the conventional sound analysis, the main purpose was to find the feature of the frequency spectrum. However, many not important features on hearing sensation are included in the parameter set expressing the shape of the spectrum. Moreover, there is a limit in analyzing the spectrum shape that is changed a lot by the background noise and the reverberation.  On the other hand, we use the correlation analysis which is based on the model of the human auditory system. By this analysis, it is possible to catch the feature of sound which people hear as it is. It becomes possible to analyze time change of the information about tonal quality such as loudness, pitch, reverberation by the autocorrelation analysis, and the spatial information such as a position and a movement of sound source by the cross-correlation analysis.  Furthermore, the most important point is that this method can express the sound feature with a small number of parameter. It becomes advantageous in transmission of measurement data or construction of a sound database. It is necessary to always continue sending sound data in the case of "in-home" telemedicine. Even if the data-transmission capacity becomes large, it becomes impossible considering a situation which examines many patients at once. By using the proposed technology, it can analyze auscultation sound and extract only an important portion.  So far, we analyzed sample of heartbeat and lung sound and examined the validity of this method. We aim at acquiring still large-scale data this time, and improving the analyzing system to the level of utilization.

Construction of sound database using Internet

A skill of auscultation depends on practice and experience. It is necessary to hear sound and memorize the feature of the characteristic sound related to some disease. A young doctor with an insufficient experience often misdiagnose. By using the proposed technology, the characteristic sound components contained in auscultation sound can be caught correctly. It can support diagnosis exactly. Furthermore, by applying the technique established by the speech recognition, it becomes possible to discover abnormalities by comparing the measured data with the database of auscultation sound. The database is regarded as a kind of medical sound dictionary. Finally, it is possible to construct a database of all the information about auscultation sound. It is also possible by constructing this database on a network to contribute to support the research and the remote medical examination. We especially aim at the development of a measurement device and the construction of a sound diagnostic network by the server system.

Economic and social impact of proposed technology