A Caltech Library Service

Quantification and classification of neuronal responses in kernel smoothed peristimulus time histograms

Hill, Michael R. H. and Fried, Itzhak and Koch, Christof (2015) Quantification and classification of neuronal responses in kernel smoothed peristimulus time histograms. Journal of Neurophysiology, 113 (4). pp. 1260-1274. ISSN 0022-3077. PMCID PMC4422346.

Full text is not posted in this repository. Consult Related URLs below.

Use this Persistent URL to link to this item:


Peristimulus time histograms are a widespread form of visualizing neuronal responses. Kernel convolution methods transform these histograms into a smooth, continuous probability density function. This provides an improved estimate of a neuron's actual response envelope. We here develop a classifier, called the h-coefficient, to determine whether time-locked fluctuations in the firing rate of a neuron should be classified as a response or as random noise. Unlike previous approaches, the h-coefficient takes advantage of the more precise response envelope estimation provided by the kernel convolution method. The h-coefficient quantizes the smoothed response envelope and calculates the probability of a response of a given shape to occur by chance. We tested the efficacy of the h-coefficient in a large dataset of Monte Carlo simulated smoothed peristimulus time histograms with varying response amplitudes, response durations, trial numbers and baseline firing rates. Across all these conditions, the h-coefficient significantly outperformed more classical classifiers, with a mean false alarm rate of 0.004 and a mean hit rate of 0.494. We also tested the h-coefficient's performance in a set of neuronal responses recorded in humans. The algorithm behind the h-coefficient provides various opportunities for further adaptation and the flexibility to target specific parameters in a given dataset. Our findings confirm that the h-coefficient can provide a conservative and powerful tool for the analysis of peristimulus time histograms with great potential for future development.

Item Type:Article
Related URLs:
URLURL TypeDescription CentralArticle
Fried, Itzhak0000-0002-5962-2678
Koch, Christof0000-0001-6482-8067
Additional Information:© 2014 by the American Physiological Society. August 2014. We thank Hideaki Shimazaki for valuable advice on optimization algorithms and Ueli Rutishauser for insightful comments on signal detection theory. This work was supported by the Swiss National Science Foundation (PBSKP3-124730) and the G. Harold & Leila Y. Mathers Foundation (09212007).
Group:Koch Laboratory (KLAB)
Funding AgencyGrant Number
Swiss National Science Foundation (SNSF)PBSKP3-124730
G. Harold and Leila Y. Mathers Charitable Foundation09212007
Subject Keywords:Peristimulus time histogram, kernel convolution, signal detection, action potentials, spike trains, firing rate, human data
Issue or Number:4
PubMed Central ID:PMC4422346
Record Number:CaltechAUTHORS:20141215-111758754
Persistent URL:
Official Citation:Quantification and classification of neuronal responses in kernel-smoothed peristimulus time histograms Michael R. H. Hill, Itzhak Fried, Christof Koch February 15, 2015 : 1260-1274 DOI: 10.1152/jn.00595.2014
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:52812
Deposited By: Ruth Sustaita
Deposited On:15 Dec 2014 20:14
Last Modified:03 Oct 2019 07:44

Repository Staff Only: item control page