Title: Prediction of Hepatitis C using Artificial Neural Network
1 Prediction of Hepatitis C using Artificial
Neural Network
- Rinki Jajoo, Dinesh Mital, Syed Haque
2- SUMMARY - RESEARCH
- The main objective of this research project is
develop an expert systems system for the
diagnosis of Hepatitis C, based on a back
propagation feed forward artificial neural
network (ANN), - Compare its performance with other existing
computer based decision support systems. - The ANN based system is developed with the
commercially available software package (Brain
Maker, California scientific Software). - Two different types of ANN models, unsupervised
and supervised, are developed, compared, and
tested. - The predictive accuracy and the training for the
supervised model is significantly better. - The model is able to predict the Hepatitis C in
patients accurately.
3- The performance of the proposed ANN model is
not significantly better than the traditional
computer model based techniques. - Further investigations are needed to understand
the impact of this methodology on the outcome
analysis. - An existing database of Hepatitis C infected
patients is used. Data of 15 infected and 20
non-infected individuals are used in the model.
This could be one of the possible reasons that
the results were not significantly better. - Continuous variable are recorded for patient
age, ethnicity, and patient number. The results
have been very interesting, however, some further
research work is required to fine-tune the
results. - The main advantage of the developed system is
that it is adaptive and self- learning type.
4- ARTIFICIAL NEURAL NETWORKS
- Artificial Neural networks are composed of a
series of computational nodes structured into
several layers. - Back propagation feed-forward network is used for
this application. - Each node is connected to all nodes in the
previous layer. - The Artificial Neural network structure used for
in this project consists of a three layered feed
forward design. -
5Neuron Structure
6(No Transcript)
7- The first layer is an input layer in which each
node represent an input variable. The second
layer is a hidden computational layer and the
third layer is a single output node representing
the outcome for a given data record. (i.e.,
Hepatitis C) - The weighted sum of the input data ( nodes) is
calculated and presented to each node in the
hidden layer. -
- A transfer function then takes the weighted sum
and calculates a value that numerically scales
the strength of each nodes output. - The output of each succeeding layer becomes the
input of the next layer.
8(No Transcript)
9- There are two user-dependent parameter that
affects the training. - a) The learning rate and
- b) The momentum.
- The learning rate defines the amount by which
weights are changed during each iteration. A
larger learning rate leads to larger weight
change. - Momentum allows the weight change to be
proportional to the previous weight change.
10- THE MODEL
- The Artificial Neural networks in our study used
a sigmoid transfer function and the network was
trained with Feed forward-back propagation
learning algorithm. - The structure of the network is a three layered
with one hidden layer. - During training, the outcome is known for each
input data record. That is system is trained on
the known patient data taken from the database. - After the training is completed, the system is
validated on the known patient database. - During the training, the output value for the
Artificial Neural network is calculated, the
value is compared with the known output from the
training data.
11- If there is a discrepancy between the calculated
and the actual output, the error is propagated
through the Artificial Neural network and weights
are adjusted until the error discrepancy is
within the allowed tolerance. - The outcome of the proposed model shows that the
supervised model is better than unsupervised
system. -
- The training of the network is done using a
number of facts (input data).
12- There are 8 hidden neurons used by the hidden
layer of the model. -
- In each training run there are multiple number of
inputs (facts) to be trained. - When the system is trained using the available
facts it tries to learn after each iteration.This
system was trained using 1218 facts in 87
training runs. There were 14 facts in each run. - During the training period, the RMS error
decreased (0.079). - When system started learning there are no good
output(s). All the outputs are bad. After the
extensive training, there should be no bad
outputs - most of the outputs will be good.
13(No Transcript)
14(No Transcript)
15(No Transcript)
16(No Transcript)
17(No Transcript)
18(No Transcript)
19(No Transcript)
20- CONCLUSION
- We have designed, implemented and tested two
different artificial neural network models. - Both the models are tested with normal and
infected patient data. - The classification performance of the Supervised
ANN model is superior than either rule based
model or Unsupervised network. - The ANN based systems are self adaptive and learn
with each additional input data set provided to
it.
21- CONCLUSION ( contd. )
- Currently, the models have been tested on a very
limited and small patient database. However, we
plan to test it rigorously on large patient
database. - We believe that this proposed model for
predicting the Hepatitis C would assist doctors
in the diagnosis of disease. - Accuracy of the model for the prediction of test
cases is over 80 for the supervised network.
22- Learning
- An ANN learns from its experience. The usual
process of learning involves three tasks - Compute output(s).
- Compare outputs with desired results and feedback
the error. - Adjust the weights and repeat the process.
- The learning process starts by setting the
weights by some rules ( or randomly). The
difference between the actual output (y) and the
desired output(z) is called error (delta). - The objective is to minimize delta ( to zero).
The reduction in delta is done by changing the
weights
23- Supervised
-
- Uses a set of inputs for which the desired
outputs results / classes are known.The
difference between the desired and actual output
is used to calculate adjustment to weights of the
NN structure. - Unsupervised
- Only input stimuli (parameters) are presented to
the network. The network is self organizing, - No knowledge is supplied about the classification
of outputs. However, the number of categories
into which the network classifies the inputs can
be controlled by varying certain parameters in
the model.
24- Developing NN Applications
- Important step is the selection of network
structureThe available network structures are - Associative Memory Systems
- It refers to ability to recall complete
situations from partial information. Such systems
correlate input data with information stored in
memory, - Information can be recalled even from incomplete
or noisy inputs. - Associative memory systems can detect
similarities between new inputs stored input
patterns. - Hidden Layer Systems
- Complex practical applications require one or
more hidden layers between inputs and outputs and
and a corresponding large number of weights.
25Inputs Each input corresponds to a single
attribute. For example for diagnosing a disease,
each symptom, could represent an input to one
node. Input could be image of skin texture, if
we are looking for cancer cells. Outputs The
outputs of the network represent the solution to
a problem. For diagnosis of a disease, the answer
could be yes or no. Weights A key element of
ANN is weight. Weight expresses relative strength
of the entering data from various connections
that transfers data from input point to the
output point.
26The Procedure for developing NN applications is
1. Collect Data. 2. Separate the data into
training and test sets. 3. Define the network
structure. 4. Select a learning
algorithm. 5. Transform data to network
inputs. 6. Start training and revise weights
until the error criterion is satisfied. 7. Stop
and test the results with test data. 8. Implementa
tion use network for the new case. Transformatio
n Function One of the common function is sigmoid
function where YT is the transformed value of Y.