International Journal of Advanced Research, Ideas and Innovations in Technology
© 2021, www.IJARIIT.com All Rights Reserved Page| 135
ISSN: 2454-132X
Impact Factor: 6.078
(Volume 7, Issue 5 - V7I5-1200)
Available online at: https://www.ijariit.com
A study of Machine Learning algorithms to predict liver
cirrhosis and its stage
Prakash Aryan
2019btechaiprakash7275@poornima.edu.in
Poornima University, Vidhani, Rajasthan
Abstract: In early stages, cirrhosis usually doesn’t cause
symptoms. Only through routine blood tests or liver biopsy does
a doctor diagnose damage to the liver. Using Machine learning
we develop a model that can assist doctors in diagnosing the
early stages of liver cirrhosis before it gets fatal. In this study
we use various machine learning algorithms to determine the
liver cirrhosis stage. Data is collected from the Mayo Clinic
trial, USA, in primary biliary cirrhosis (PBC) of the liver
conducted between 1974 and 1984. Performance of the
algorithms were evaluated using ROC-AUC curve which is a
very practical method of model evaluation for classification
problems. Results showed that by applying Logistic Regression
to predict the cirrhosis stage we get a ROC-AUC score of 0.74
which is considerable in view of the instance we have.
Keywords: Supervised Machine Learning, Disease Prediction,
Python
1. INTRODUCTION
Liver is the second largest organ in our body after our skin. It
helps the body in removing toxins that are present in the blood
supply, breaking down drugs, making the digestive fluid "Bile''
(helps in digestion and removing wastes from the body), storage
and release of glucose as required and various other processes
and it is thus called as the "body's chemical processing plant" [1].
Fibrosis of the liver is the early stage when a healthy tissue is
replaced with non-living scarred tissue (Figure I) and obstructs
the functioning of the liver. Cirrhosis is the late or end stage of
liver fibrosis that causes permanent and serious scarring of the
liver. Liver cirrhosis makes the functioning of the liver very
difficult. End-stage liver disease (ESLD) leads to a liver in which
the functioning of the liver has drastically deteriorated such that
it causes disruptions in flow of blood to the liver which leads to
pressure build up in the portal vein. It is a very severe case of
cirrhosis.
According to The Lancet (Gastroenterology & Hepatology) the
age - standardized death rate globally, due to liver cirrhosis in
2017 was 16.5 per 100,000 population. In the sub-Saharan Africa
super-region it was 32.2 per 100,000 population as compared to
10.1 per 100,000 population in the high-income super-region [2].
According to The Global Health Observatory of the World
Health Organization, for countries such as Cambodia, Egypt,
Nigeria and Sao Tome and Principe the age-standardized death
due to liver cirrhosis is over 100 per 100,000 male population for
the year 2016 [3].
In the initial stages around 40 percent of those affected with liver
cirrhosis are asymptomatic [5]. The Most prevailing causes of
liver cirrhosis are alcohol abuse, Hepatitis A, Hepatitis B and
Nonalcoholic fatty liver disease (NAFLD) [6]. Global Alcohol-
attributable fraction(AAF) for liver cirrhosis deaths(%) is 44
percent [7].Patients with liver cirrhosis have a number of
complications for examples, ascites, spontaneous bacterial
peritonitis, hepatic encephalopathy, portal hypertension, variceal
bleeding, and hepatorenal syndrome [8]. Various important
parameters such as drug, age, ascites, hepatomegaly, spiders,
edema, bilirubin, cholesterol, albumin, copper, alkaline
phosphatase, SGOT, triglycerides, platelets and prothrombin are
collected for each individual patient.
Figure I. Left: Healthy liver Right: Liver cirrhosis[4]
Artificial Intelligence in healthcare has helped in the reduction
of the burden on doctors and their patients. An artificial
intelligent system is a system that can perform tasks that would
otherwise require human intelligence. In this study machine