17
York University Instructor: Professor N.Cercone Noada Lugaj Jason Panas 1

York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

Embed Size (px)

Citation preview

Page 1: York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

York University

Instructor: Professor N.Cercone

Noada LugajJason Panas

1

Page 2: York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

Introduction Medical Diagnosis Background Challenges and Problem Definition Past Work Decision Theory Artificial Neural Network Our Approach and Algorithm Description Limitations

2

Page 3: York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

Medical Diagnosis is one of the hardest fields of medicine.

and

Automating the process In this project we aim to investigate the

application of artificial neural networks in medical diagnosis and propose a simple and applied method for that.

3

Page 4: York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

4

• Medical Diagnosis Background

Medical Diagnosis has been always seen as art. Throughout the history we remember famous doctors as well as famous painters or composers

Hippocrates (c. 460–370 BCE) — Greek father of medicine

Elliott P. Joslin (1869–1962) — pioneer in the treatment of diabetesTheodor Kocher thyroid surgery and first surgeon to win the Nobel

Prize

It is undeniable the contribution of these doctors and many other famous physicians for their role in the advancement of medicine.

Page 5: York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

5

Diagnosis and a treatment

Page 6: York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

6

A sufficient number of experienced cases only in the middle of his career.

More complications for new diseases where experienced physicians are in the same position as newcomers.

Humans resemble pattern recognition systems and not statistic computers

A good diagnosis

= a talented and experienced physician

Performance of a doctor can be degraded from fatigue and emotional situation

Education to become a doctor requires long time and is expensive

Page 7: York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

The situation is fairly complicated and includes many problems so it would be nice if computers could help.

Computers have served well the medical sector for a few decades now.

They have been used widely in medicine and patient databases: local and global; digital archives and emergency networks.

It is unrealistic to expect a fully automated computer based medical diagnosis system because of the complexity of this task. But it is quite obvious why a system for automated medical diagnosis would enhance medical care and reduce costs.

7

Page 8: York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

8

• Decision TheoryMedical Diagnosis is a vital task that should be performed as accurately and efficiently possible.

Decision Theory can be used in order to implement a computer based medical diagnosis system.

In mathematics and statistics Decision theory is concerned with identifying the values, uncertainties and other issues relevant in a given decision, its rationality, and the resulting optimal decision. Decision theory can be used to make optimal choices based on probabilities and utilities. On one hand probability theory tells us the probability of future states and how to represent uncertainty events. On the other hand utility theory values different possible events so they can be compared to each other. Decision theory can be used during information gathering in a diagnosis session to determine which new evidence will be most efficient to acquire next or which further evidence will no longer improve the accuracy of diagnosis .Hence Decision theory can be used to decide what to do next in order to maximize the information gathered or when to stop gathering information.

Page 9: York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

9

• Artificial Neural Network(ANN)

Neural Network is a very popular topic in Artificial Intelligent field.

ANN are designed to simulate the behavior of biological neural networks for several purposes.

It is an attempt to simulate within specialized hardware or sophisticated software, the multiple layers of simple processing elements called neurons. Each neuron is linked to certain of its neighbors with varying coefficients of connectivity that represent the strengths of these connections. Learning is accomplished by adjusting these strengths to cause the overall network to output appropriate results.

Page 10: York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

Neural Networks are ideal in recognizing diseases since there is no need to provide a specific algorithm on how to identify the disease.ANN is highly able to derive meaning from complicated and even imprecise data. It can be used to recognize patterns or trends that are almost impossible to be detected by humans or other computer-based techniques.One of the most important problems of medical diagnosis, in general, is to perform pattern recognition activities. This is when ANN comes to help. A  study  in 1971  showed  these  basic  facts  in  the medical area. This study had shown that human have many limitations in diagnosis. The results of this experiment were as follows:•Best human diagnosis (most experienced  physician): 79.7%•Computer with expert data base: 82.2%•Computer with 600 patient data: 91.1%

From these results we can see that humans cannot analyze complex data without errors.

10

Page 11: York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

Neural Network approach has been used in diagnosing diseases using patient medical data such as breast cancer, heart failure, medical images, acidosis diseases, and lung cancer.

•Breast cancer

It is the second largest cause of cancer deaths among women. The automatic diagnosis of breast cancer is an important, real-world medical problem.

This has given rise, over the past few decades, to computerized diagnostic tools, intended to aid the physician in making sense out of the confusion of data. General Regression Neural Network and Probabilistic Neural Network can be effectively used for breast cancer diagnosis to help oncologists.

Hence, ANN has been proven of their capabilities in many domains such as medical application. Neural network with ability to learn by example

makes them very flexible and powerful in medical diagnosis.

11

Page 12: York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

Programmed in Java

Algorithm uses 1 linked list which holds custom objects called “disease”

Disease objects are composed of 3 parts:Disease nameLinkedList of symptomsInteger “match” variable

12

Page 13: York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

The combination of these 2 data structures allows us to use a neural network kind of approach to this problem.

With each symptom that is listed, that symptom is run through the linkedlist of diseases, and the match variable of each disease is updated.

After each object in our knowledge base is checked and matched, then each item that has a match is returned to the user with a weight that is the probability that that disease is the one the patient will have.

13

Page 14: York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

Weighting is done as follows: A disease has a symptom listed multiple times if it is a more prominent

symptom. Once a symptom is given, the match level is increased by the number

of times the symptom appears. When all the symptoms that have a match we get the highest match #

of the number of symptoms given, whichever is higher. For each match a disease doesn’t have, the % for the disease is

reduced by a factor of 2n, where n is the difference in match values, with the rest being distributed over the other diseases of higher match levels.

14

Page 15: York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

Demo

15

Page 16: York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

The program aims to show the implementation of ANN in medical diagnosis. It does not focus on how to setup proper medical data. This is why the results produced do not have any medical meaning or are irrelevant to any real life medical diagnosis process. But it shows how important and how accurately the process of Medical Diagnosis can be done using ANN techniques.

In order to get more medically accurate results it would require to collect real data from real cases and maybe add a more elaborated manipulation process upon this data.

16

Page 17: York University Instructor : Professor N.Cercone Noada Lugaj Jason Panas 1

17

Thank youNovember,29,2011