Analysis and classification of delay-sensitive cortical neurons based on response to temporal parameters in echolocation signals.
Echolocating bats generate an acoustic image of their target by processing target-reflected echoes of their emitted biosonar pulses. Efforts in building computational models of auditory processing in the bat auditory system, using extensive neurophysiological data from cortical studies are challenged by the intrinsic complexity and the significant variability in neural response to stimuli. In this paper, we use a computerized method for the analysis and classification of delay-sensitive neurons to classify neurons from the auditory cortex of Myotis lucifugus, a species that echolocates with FM signals. The coefficients of the bi-linear fit to the best delay response surfaces (mean R2 = 0.01) were used in classifying the neurons. Six classes were derived that corresponded to the four previously characterized neurophysiologically. The first class corresponded to delay-tuned neurons which exhibited a constant best delay at different pulse repetition rates and pulse durations. Three other classes corresponded to the different subtypes of tracking neurons which changed their best delay to one or both of these stimulus temporal parameters. Two additional classes were differentiated although their best-delay response were similar to either the delay-tuned or the duration and pulse-repetition rate sensitive class. Artificial delay-sensitive neurons built from the parameters of the centroid of each class, will serve a key role in the FM bat auditory system model that we are building.