Medical Diagnosis

iNovAITec Medical Diagnosis

Background

Lung and heart diseases have been leading causes of mortality worldwide for many years. Since the onset of the coronavirus disease 2019 (COVID-19), the number of deaths related to them has increased rapidly.

Various clinical methods have been developed to diagnose and evaluate lung and heart health conditions, including computed tomographic scans (CT), chest X-rays, and pulmonary function tests. However, these methods are complex due to the massive amount of data and their high costs, often limiting their accessibility to high-end clinics.

Solution

iNovAITec utilizes the transformative power of deep learning to create a groundbreaking healthcare solution.

iNovAITec’s advanced AI software is designed to provide real-time, on-the-spot classification and identification of various diseases cases. For example the software can be detected the COVID-19 and brain tumor using CT scan images. Also the Classification of lung conditions (e.g. asthma, chronic obstructive pulmonary disease (COPD), and pneumonia) and heart respiratory sounds, revolutionizing the way medical professionals diagnose pulmonary and cardiac conditions. The respiratory sounds are recorded by an electronic stethoscope that can be used with any device. Imagine a world where doctors receive immediate and precise sound identifications, enabling them to make critical decisions swiftly and accurately. This technology empowers healthcare providers, making a tangible difference in patient care. iNovAITec’s deep learning and real-time sound identification and classification converge to ensure a healthier future for all.

Benefits

With iNovAITec software, more than 10.000 lung sound files were recorded using an electronic stethoscope device. These sounds were transformed into spectrogram images using the mel frequency cepstral coefficient features (MFCC) in Deep Learning.

The patient’s lung respiratory sounds include categories such as asthma, pneumonia, chronic obstructive pulmonary disease (COPD), and sounds from healthy patients.

As an experiment result, it was found that spectrogram image classification using Deep Learning algorithms in addition to in-house developed Image Processing algorithms showed excellent results (97% accuracy). Given the substantial amount of data, the used algorithms can accurately classify and pre-diagnose respiratory audio.

Classification of lung sounds using Deep Learning