Hemodynamic Analysis for Olfactory Perceptual Degradation Assessment Using Generalized Type-2 Fuzzy Regression

• Published in: IEEE transactions on cognitive and developmental systems Vol. 14; no. 3; pp. 1217 - 1231
• Main Authors: Laha, Mousumi, Konar, Amit, Rakshit, Pratyusha, Nagar, Atulya K.
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.09.2022; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Brain modeling; Computational modeling; Confidence intervals; Degradation; Hemodynamic analysis by functional near-infrared spectroscopy (f-NIRs); Hemodynamic responses; Hemodynamics; Model testing; Olfactory; olfactory perceptual degradation (OPD); Optimization; Predictive models; Regression models; Statistical analysis; Stimuli; Training; type-2 fuzzy reasoning and regression; Uncertainty
• ISSN: 2379-8920; 2379-8939
• DOI: 10.1109/TCDS.2021.3101897

Olfactory perceptual degradation (OPD) refers to the inability of people to recognize the variation in concentration levels of olfactory stimuli. This article attempts to assess the degree of OPD of subjects from their hemodynamic response to olfactory stimuli. This is done in two phases. In the first (training) phase, a regression model is developed to assess the degree of concentration levels of an olfactory stimulus by a subject from her hemodynamic response to the stimulus. In the second (test) phase, the model is employed to predict the possible concentration level experienced by the subject in [0, 100] scale. The difference between the model-predicted response and the oral response (the center value of the qualitative grades) of the subject about her perceived concentration level is regarded as the quantitative measure of the degree of subject's olfactory degradation. The novelty of the present research lies in the design of a general type-2 fuzzy regression model, which is capable of handling uncertainty due to the presence of intrasession and intersession variations in the brain responses to olfactory stimuli. The attractive feature of the article lies in adaptive tuning of secondary membership functions to reduce model prediction error in an evolutionary optimization setting. The effect of such adaptation in secondary measures is utilized to adjust the corresponding primary memberships in order to reduce the uncertainty involved in the regression process. The proposed regression model has good prediction accuracy and high time efficiency as evident from average percentage success rate (PSR) and runtime complexity analysis, respectively. The Friedman test undertaken also confirms the superior performance of the proposed technique with other competitive techniques at 95% confidence level.