Overview of artificial neural network in medical diagnosis - Pubrica
                     OVERVIEW OF 
ARTIFICIAL  NEURAL 
NETWORK IN  MEDICAL 
DIAGNOSIS
An Academic presentation by
Dr. Nancy Agens, Head, Technical Operations, 
Pubrica  Group: www.pubrica.com
Email: [email protected]
Today's Discussion
Outline In-Brief  
Introductio
n
Artificial Neural 
Network  Architecture
Overview of Artificial Neural Network in Medical 
Diagnosis  Cardiovascular Diseases
Cancer  
Diabetes  
Conclusio
n
In-Brief
A massive volume of clinical data is produced daily that possess minute and critical  
information as well as varied, in-depth concepts of biochemistry and the results of  
imaging devices. Information provided byeach kind of data must be evaluated and  
assigned for diagnostic processes. To simplify the diagnostic process and evade
errors in that process, artificial intelligence techniques can be adopted like computer-  
aided diagnosis and artificial neural networks. The b iostatistical services machine
learning algorithms can deal with a broad set of specific data and produce  
categorized outputs by checking the blogs in Pubric.
Introduction The artificial neural network has been widely used in 
 the fields of science and technology. It is used for the  
optimization of data.
It predicts the outputs using the input data in fields like 
 chemical engineering, biotechnology, healthcare,  
agriculture, etc., which all handles varied sets of data.
The artificial neural network can be used for 
modelling
non-linear systems with a complex system of 
variables.
Thus, most of the chemical engineering and biological 
 processes are modelled using Artificial neural 
network  with the help of b iostatistical consulting 
services.
Artificial 
Clinical biostatistics services state that Artificial neural  
 Neural  network is the simulation of human neural architecture.
Network
The learning and generalization potentials of human 
 neural network inspired for the development of an  
artificial neural network.
It works by taking the 70% of input data to build a  
network then takes the remaining 15% data to train itself 
 and at last utilize the remaining 15% data to test itself  
and eventually produce the optimized outputs.
Architecture
The a rtificial neural network is made up of 
three  layers, viz., – (i) input layer, (ii) hidden 
layer, (iii)  output layer.
The schema of the neurons built inside the 
network
is based upon the complexity of the system.
The input layer collects the input data and transfers 
 to the hidden layer where the data is processed to  
produce optimized results with s tatistical  p
rogramming services.
Every Artificial neural network has an activation function that is used for 
 determining the output.
Each neuron is interconnected, and each connection has a weight attached 
 possessing either positive or negative value which tends to change upon  
the training the network.
Overview  
of Artificial Seeking various uses in various fields of science,
 Neural  medical diagnosis field also has found the application  of artificial neural network using biostatistics in  
Network in clinical services.
 Medical  
Diagnosis It is used in the diagnosis of cancer, sclerosis,  diabetes, heart diseases, etc.
An adaptive algorithm is developed and applied to  
yield maximum accuracy in outputs with the s 
tatistics  i n clinical trials.
Cardiovascula
r  Diseases It is the collection of diseases affecting the heart,
cardiac muscles, blood vessels, veins.
National centre of health statistics reported  
that leading cause of death in united states of 
 America is these c ardiovascular diseases.
In the past, the data collected from the patients 
 were used to develop an Artificial neural  
network model with the backpropagation  
algorithm was developed.
Contd..
This model was able to achieve 91.2% accuracy in the diagnosis of these diseases 
 from the d ata collected.
There were other models with less than 90% accuracy also used to
diagnosespecific types of heart diseases.
Contd.
.

Cancer In 2012, reports of American cancer society said 
 that more than 1.6 million newly diagnosed cases 
 were found.
Hence, there was the need to develop a rapid and 
 appropriate diagnosis for clinical management.
The pertinent information for diagnosis was  
collected from the advanced analytical methods like 
 mass spectrometry and applied in the clinical  
diagnosis of breast and ovarian cancer.
Contd..
Artificial neural network is also used to develop in diagnosing the different types of 
 brain tumours, lung carcinoma.
Ultimately, Artificial neural network was seen using the ground-level data that 
 ranges from clinical data to results of biochemical assays and providing  
maximum diagnostic accuracy for different types of cancer.
Diabetes
Diabetes has become a severe health risk issue 
 in both developed and developing countries 
that  reaching an estimate of 366 million 
diabetes  cases globally.
Type ii diabetes is the standard type of this 
 disease which is due to the improper 
cellular  response to insulin which leads to  
hyperglycemia.
Contd..
The information of parameters like age, gender, weight and glucose level were  
collected and used as input data for building an Artificial neural network which could 
 able to produce results with 90% accuracy.
Artificial neural networks are used to track the level of glucose as well as
diagnosing diabetes according to b iostatistical research for clinical trials.
Conclusion
The artificial neural network can be inferred as a powerful 
 tool in clinical management of diseases with several  
advantages like the capability of processing a vast set of  
data, reducing the processing time, ability to produce  
optimized results with maximum accuracy.
Nevertheless, Artificial neural network can be used only as 
 tool aiding in diagnosis done by the clinical physician,  
says b iostatistical CRO, who is responsible for critical  
evaluation of the results.
Pubrica helped to understand the role of ANN tool in the 
 medical field.
Contact 
Us
UNITED KINGDOM
+44-1143520021
INDIA
+91-9884350006
EMAIL
[email protected]