An Intelligent System Model for Diagnostic of Human Pancreatic Cancer

Pancreatic Cancer is one of the lethal cancers in the world with vague symptoms and low survival rates. It is the seventh leading cause of cancer-related deaths in the world and will become the second within 2030 in USA. The five (5) years survival rate for all stages diagnosed is estimated to be 9%. More than 60% of Pancreatic Cancer patients are diagnosed at an advanced stage due to a lack of highly sensitive and specific screening tests. However, surgical resection is one of the best treatments for Pancreatic Cancer; but only 5% patients are eligible for this treatment. Early detection and treatment are the only options to reduce fatality. The proposed study is intended to reduce the fatal effect of late detection of Pancreatic Cancer by estimating the risk before exposure. The parameters are extracted by reviewing the literature review, where biographical, clinical, and lifestyle information has thoroughly discussed and effects of those inputs are identified. The input variables are Diabetes, Smoking Habits, Alcohol Consumption, Obesity, Family History Of Pancreatic Cancer, Chronic Pancreatic, Gallbladder Diseases, Blood Group, Dietary Factor, Age, Sex, and Race. This Thesis provides a Fuzzy Logic approach to develop a risk evaluation Intelligent System for Pancreatic Cancer. Fuzzy Logic is a modern approach to deal with uncertainty. Pancreatic Cancer is the most uncertain disease due to its asymptomatic nature and Fuzzy Logic will be the appropriate method to deal with the uncertainty. Mamdani Fuzzy Inference system is used to develop through MATLAB Fuzzy Logic Tool Box. The Mamdani Fuzzy Inference system provides an opportunity to use the linguistic variables to develop the system. The proposed system is made up of six Fuzzy Inference Systems, consisting of one primary system and three sub-systems. Biographical, Clinical and Lifestyle Habit risk evaluation are the input for the primary system. In turn, the Diabetes risk evaluation FIS, and Smoking risk evaluation FIS are the sub-subsystems of Clinical and Lifestyle risk evaluation, respectively. Total 1105 If-Then rules are generated based on domain knowledge. A Multiple Input Single Output (MISO) fuzzy decision structure is implemented. The shape of the membership functions is triangular for all inputs and output of each system. The entire system is considered a set of subsystems to reduce the complexity of rules. However, a user-friendly interface has been developed to create the acceptability of the system in real life. The MATLAB App Designer tool used to create the user friendly interface. The performance evaluation has been done with the Confusion Matrix. 100 randomized data verified by physicians have been used to measure the system performance. The accuracy, sensitivity, and specificity of the system achieved 93%, 93.65% and 91.89%, respectively.The results of the system is higher than the results of other researchs of Pancreatic Cancer Diagnosis where different methodologies have been used.This Thesis is a noble research in terms of applying Fuzzy Logic in the diagnosis of Pancreatic Cancer. The introduction of Fuzzy Inference Systems as subsystems and their connection with primary Fuzzy Inference system to avoid an exponential growth of fuzzy rules represents also a uniqueness aspect of the Thesis. Moreover, the application of the user-friendly App designer is also a highlight of the Thesis which separates the research from other research papers.

Name thesis:degree_name

Master of Applied Science (MASc)

Level thesis:degree_level

Master's

Discipline thesis:degree_discipline

Engineering - Industrial Systems

Grantor dc:publisher

Faculty of Graduate Studies and Research, University of Regina

Year dc:date.issued

2021

Author dc:creator

• Bhuiyan, Md Tuohidul Alam

Advisor dc:contributor.advisor

• Mayorga, Rene

Committee member dc:contributor.committeemember

• Peng, Wei

Language dc:language.iso

en

DOI

https://doi.org/10.82465/4376

OAI identifier oai:identifier

oai:uregina.scholaris.ca:10294/14319