Page 197 - DCAP208_Management Support Systems

Management Support Systems

Notes 15. ................... of an ANN often requires interfaces with other computer based information
systems and user training.

Case Study Neural Networks for Breast Cancer Diagnosis

NN have proven to be a useful tool in pattern recognition and classification tasks
in diverse areas, including clinical medicine. Despite the wide applicability of
AANN, the large amount of data required for training makes using them an
unsuitable classification technique when the available data are scarce. Magnetic Resonance
Spectroscopy (MRS) plays a pivotal role in the investigation of cell biochemistry and
provides a reliable method for detection of metabolic changes in breast tissue. The scarcity
of data and the complexity of interpretation of relevant physiological information impose
extra demands that prohibit the applicability of most statistical and machine learning
techniques developed.
Knowledge-based artificial neural networks (KBANN) help to prevail over such difficulties
and complexities. A KBANN combines knowledge from a domain, in the form of simple
rules, with connectionist learning. This combination trains the network through the use of
small sets of data (as is typical of medical diagnosis tasks). The primary structure is based
on the dependencies of a set of known domain rules, and it is necessary to refine those
rules through training. The KBANN process consists of two algorithms.

One is the Rules-to-Network algorithm, in which the main task is the translation process
between a knowledge base containing information about a domain theory and the initial
structure of a neural network. This algorithm maps the structure of an approximately
correct domain theory, with all the rules and their dependencies, into a neural network
structure. The defined network is then trained using the backpropagation learning
algorithm.
Feedback mechanisms, which inhibit or stimulate the growth of normal cells, control the
division and replacement of cells in normal tissues. In the case of tumors, that process is
incapable of controlling the production of new cells, and the division is done without any
regard to the need for replacement, disturbing the structure of normal tissue. Changes
observed in phospholipid metabolite concentrations, which are associated with differences
in cell proliferation in malignant tissues, have served as the basic inputs for the
identification of relevant features present in malignant or cancerous tissues but not in
normal tissues. The abnormal levels of certain phospholipid characteristics are considered
indicators of tumors. These include several parameters, such as PDE, PME, Pi, PCr, γATP,
αATP, and βATP. KBANN produced an accurate tumor classification of 87 percent from a
set of 26, with an average pattern error of 0.0500 and a standard deviation of 0.0179.
Question

Discuss the role of Magnetic resonance spectroscopy (MRS) in the investigation of cell
biochemistry.
Source: http://www70.homepage.villanova.edu/matthew.liberatore/Mgt2206/turban_online_ch06.pdf

11.3 Summary

Neural networks represent a brain metaphor for information processing.
The resulting model from neural computing is often called an artificial neural network
(ANN) or a neural network.

190 LOVELY PROFESSIONAL UNIVERSITY