Liver Disease Detection using Neural Network approach via Raspberry Pi

Project Title

Project Area of Specialization

Biomedical Engineering

Project Summary

Liver is the most vital organ of human body. Liver diseases are detected with the aid of clinical information, including blood tests, imaging techniques such as magnetic resonance image (MRI), and magnetic resonance elastography (MRE). However, such methods have limitations, and a liver biopsy is usually required to diagnose liver diseases like cirrhosis, hepatocellular carcinoma, and hepatitis. The MRI and MRE technique are the earliest noninvasive techniques. The related procedures are time-consuming and expensive. However, these techniques usually do not detect a disease at its early stages. To address these issues, we propose a solution in the form of a device that detects liver stiffness and scarring using ultrasonic sensors. This can help to detect diseases such as fibrosis, cirrhosis etc. The device classifies a person’s liver as healthy or disease- affected by processing the reflected pulse of an ultrasonic sensor. For this purpose, it uses a trained neural network. In this way, an initial warning is generated for a person so that he/she can consult a physician for a detailed checkup and diagnosis.

Project Objectives

The main objective of this project is to design a scanner that can detect liver diseases and classify them using a trained neural network.

Project Implementation Method

An ultrasound scanner uses high frequency waves that penetrate into the body to detect problem in liver and other body organs. In our project the scanner will scan the liver using ultrasonic sensor and check its stiffness.

Then these results are classified by Respberry Pi in order to detect the type of disease.

Benefits of the Project

The proposed non-invasive scanner can easily warn a patient about any abnormalities in liver and issues a warning to get a thorough medical checkup as soon as possible. It detects proteins that get affected by the diseases, with the help of an ultrasonic sensor.

The device eradicates waiting period usually assigned to get lab results and the diagnosis paints a general picture for the doctor. The doctor can ask for a comprehensive test result, as the warning issued will be a precursor for a much deadlier disease that goes unchecked.

Technical Details of Final Deliverable

We will take dataset of 530 patients and will train our neural network using MATLAB software. Then we will classify the results of liver patients as affected or non-affected, using data of healthy and unhealthy livers from hospitals. Then we will classify the type of disease while training the network using elasticity values.