Computer Aided Auscultation as a Service

A look at the underlying technology:

Real-time monitoring of acoustic heart sounds and ECG signals

Signal acquisition is done in a digital fashion. By digitalising the heart sound signal, advanced signal processing can be applied to enhance and extract the required information. This process of information extraction is done in much the same way that a doctor would listen to a heart sound.

Identification of S1 and S2 sound pulses

Each time you hear a heart beat two distinct sounds are audible, the lub-dub sound. This lub-dub sound corresponds to the oxygen rich blood being pumped from the left side of the heart, and de-oxygenated blood flowing in from the body. The process is repeated continuously to produce  lub-dub sounds also labeled as the S1-S2 intervals of the heart beat signal.

Segmentation of the heart sound

When a doctor is listening to a patients heart, the doctor needs to identify each heart beat, and  determine when the sound intervals S1-S2 sounds start and stop. In this way the doctor segments the heart sound and  obtains useful information which may be used to confirm a diagnosis. In the same way signal processing can be applied to segment the heart sound.

Extracting timing, frequency and duration information

A human heart is an amazing organ, and is able to pump blood at a rate the body requires. Naturally this means that a person’s heart rate may be very high or low. At low heart rates, below 120 bpm, segmentation and obtaining timing information is relatively easy to do with a practiced ear. When heart rates rise above 120 bpm, which is very common for young children, the task may become impossible. In these cases a computer can easily aid doctors to capture the quick heart beats, and extract the information required.

Artificial Neural Networks

Artificial Neural Networks is a mathematical model used to simulate the way the brain learns. In its simples form the neurons of the brain can be seen to function as a network of interconnected nodes sharing information based on inputs from the environment. These neural pathways are made stronger every time the same inputs are experienced. Similarly, an Artificial Neural Network learns from the inputs it is given. The ANN used for Sensi was trained using hundreds heart signals of both healthy and abnormal sounding hearts.

Algorithms for embedded platforms

Technologies such as mobile devices and medical equipment are in most cases much slower and require expert skills to program than do today’s desktop computers. Digital signal processing algorithms have been specifically designed for an embedded system and can be ported to different platforms.


The development of computer aided decision algorithms are done in corporation with and validated by medical specialists.

Designing mobile monitoring systems

In designing mobile monitoring systems Diacoustic addressed the following technology obstacles:

Device size is important

The overall size should be compatible with the required form factor. The usefulness of a device, especially a mobile device, is mainly in determining the correct size and designing the technology in such a way as to fit properly without negatively impacting the devices look and feel.

Energy usage

The energy autonomy of current battery-powered systems is limited. The technology and power output of batteries have not changed drastically over the last decade. Although, the power usage of devices have changed drastically, with the addition of more powerful processors and touchscreens.

Embedded Design

Embedded design for mobile monitoring devices are done in-house.  These devices are able to execute complex decision support algorithms built on our digital signal processing libraries.

Our in-house expertise includes:

• Housing and industrial design
• User interface development
• Firmware development
• File and memory management development
• Bluetooth interfacing
• Battery and power management
• EMC and EMI testing