Studies show that fuzzy logic has different approaches for enhancing personal health care delivery in the health care sector. Currently, breast cancer rated as the second leading cause of death among women, according to the World Health Organisation. Previous studies relating to breast cancer prognosis using fuzzy logic approach were directed at reoccurrence of the disease as well as the survivability of individual. However, there is need for early identification of the predisposing risk factors to breast cancer growth and its elimination. Consequently, this study focuses on developing an efficient Mobile-based Fuzzy Expert System (MFES) for initial breast cancer growth prognosis that can predict the individual risk level and thus, reduce the high incidence rate.
Facts relating to the predisposing factors of breast cancer were elicited from four domain experts through direct contact; this was used to generate the appropriate fuzzy rules. The fuzzy inference approach was employed to formulate the membership functions and fuzzy rules to design the MFES. Mamdani approach was used for the fuzzification of input and de-fuzzification of the output. The system accommodates imprecision tolerance and uncertainty to achieve tractability, robustness and least solution cost. Java expert system shell running on Android operating system was used to achieve the mobile technology aspect of the system. For the purpose of system evaluation, 2500 data were collected from two health care centers in Nigeria using random sampling technique. The data were stratified into twenty-five different strata. Each stratum contained 100 dataset and four individual data were selected at random.
The result indicated that the fact elicited from the experts serves as range values for the 12 risk factors used for the fuzzification of the input and thus, 36 rules were generated. The rules were used as basis for the development of the MFES. The developed MFES for breast cancer growth prognosis recorded 96% accuracy for all dataset picked from the 25 different strata. The system showed that with higher number of fuzzy rules focusing on pre-tumour growth and detailed predisposing risk factors; the prediction of risk level was reliable.
This work provided the resource for an individual to personally examine the breast cancer risk level, showing that the predisposing risk factors can be reduced by personal health monitoring. Though, MFES employed higher number of fuzzy rules unlike the existing systems with less number of rules; it supports pre-tumor growth instead of post-tumor growth which was incapable of handling the high incidence rate. It is therefore recommended that MFES be used to detect predisposing breast cancer risk levels early enough. The main contribution of this work is the reduction of the incidence rate in contrast to the existing methods currently applied in the diagnosis of breast cancer.
Keywords: Soft Computing, Fuzzy Set, Breast Cancer, Risk Factors, Membership Functions
Word Count: 448