Differential microcyte anemia diagnosis using hierarchical soft computing

Abstract

Anemia is the most common hematological disorder. It’s difficult to discriminate either thalassemia or Iron deficiency anemia, due to the two subtype of microcytic anemia, which has the feature similarly with mean cell volume less than 80 fL (fluid ounces). The CBC is objective for physician to discriminate anemia between iron deficiency anemia with thalassemia. The disorder will be more serious when physicians cannot identify to therapy adequately. Applied soft computing to solve problem of classification such as Fuzzy C-means, Competitive learning gain for more attentions, and Adaptive neural-fuzzy inference system, that parallels the human mind to process information under imprecision and uncertain circumstance. Soft computing is fitting to discriminate based on CBC result of the microcytic diagnosis under imprecision and uncertain. After ANFIS pruning rule, this paper find: (1) There is 98% accuracy inferred in 50 confirmed cases by ANFIS reasoning, which is more accurate than traditional experience. (2) Under sensitivity and specificity, sensitivity is 90%, and specificity is 95.8% higher than others discriminant function when employ ANFIS with 13 rules, and inference value is 13.6.