An adaptive classification model based on the Artificial Immune System for chemical sensor drift mitigation

Adaptive classification model based on artificial immune system for breast cancer detection

“…Immune system is a multi-level defense architecture within a living being that protects the organism against external agents [4]. Basically, the immune system consists of two parts namely, Innate Immune System (IIS) and Adaptive Immune System (AIS).…”

“…Basically, the immune system consists of two parts namely, Innate Immune System (IIS) and Adaptive Immune System (AIS). These mechanisms identify and neutralize antigens and pathogens that infect the organism [4]. Each aspect in the system differs with respect to how immediately it responds and how long it allows the pathogens.…”

“…The computerized systems looks for abnormal areas of density, i.e., a mass, or clustered calcifications that may indicate the presence of breast cancer. Screening mammography detects breast cancer in an asymptomatic population [2,4]. Generally the screening mammography consists of four views with two projections of each breast namely, cranio-caudal (CC) and mediolateral oblique (MLO) projections.…”

Identifying volatile organic compounds by determining their diffusion and surface adsorption parameters in microfluidic channels

“…Such compact size, light weight, durable, rugged, user friendly [4,5] and low cost systems are expected to discriminate among odors and complex gas mixtures [6,7] without time consuming and costly systematic analysis. Sensor arraybased e-nose systems [8,9] satisfy most of the required operational conditions and are, in principle, suitable for these applications, but they generally suffer from the unpredictable [10,11,12] and predictable [13,14] drifts of the array components [15,16], which render the system unreliable. These sensor arrays need frequent recalibrations and/or costly sensor array replacements [17][18][19].…”

An On-line Reconfigurable Classification Algorithm Improves the Long-term Stability of Gas Sensor Arrays in Case of Faulty and Drifting Sensors

“…Feature extraction is typically optimized during the training [2][3]; however, accidents like drift and fault can alter the information content of descriptors making the initial feature selection inadequate [4]. Although several algorithms have been introduced in literature to counteract the chemical sensor drift, few papers have investigated efficient solutions to reduce the decrease of performances when a sensor fault occurs in the online functioning [5][6][7]. Here, we introduce an algorithm that dynamically updates the selection of features considering the evolution of the information content.…”