This site needs JavaScript to work properly. Please enable it to take advantage of the complete set of features!
Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

NIH NLM Logo
Log in
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2013 Oct 7:14:114.
doi: 10.1186/1471-2202-14-114.

The discrimination of interaural level difference sensitivity functions: development of a taxonomic data template for modelling

Affiliations

The discrimination of interaural level difference sensitivity functions: development of a taxonomic data template for modelling

Balemir Uragun et al. BMC Neurosci. .

Abstract

Background: A major cue for the position of a high-frequency sound source in azimuth is the difference in sound pressure levels in the two ears, Interaural Level Differences (ILDs), as a sound is presented from different positions around the head. This study aims to use data classification techniques to build a descriptive model of electro-physiologically determined neuronal sensitivity functions for ILDs. The ILDs were recorded from neurons in the central nucleus of the Inferior Colliculus (ICc), an obligatory midbrain auditory relay nucleus. The majority of ICc neurons (~ 85%) show sensitivity to ILDs but with a variety of different forms that are often difficult to unambiguously separate into different information-bearing types. Thus, this division is often based on laboratory-specific and relatively subjective criteria. Given the subjectivity and non-uniformity of ILD classification methods in use, we examined if objective data classification techniques for this purpose. Our key objectives were to determine if we could find an analytical method (A) to validate the presence of four typical ILD sensitivity functions as is commonly assumed in the field, and (B) whether this method produced classifications that mapped on to the physiologically observed results.

Methods: The three-step data classification procedure forms the basic methodology of this manuscript. In this three-step procedure, several data normalization techniques were first tested to select a suitable normalization technique to our data. This was then followed by PCA to reduce data dimensionality without losing the core characteristics of the data. Finally Cluster Analysis technique was applied to determine the number of clustered data with the aid of the CCC and Inconsistency Coefficient values.

Results: The outcome of a three-step analytical data classification process was the identification of seven distinctive forms of ILD functions. These seven ILD function classes were found to map to the four "known" ideal ILD sensitivity function types, namely: Sigmoidal-EI, Sigmoidal-IE, Peaked, and Insensitive, ILD functions, and variations within these classes. This indicates that these seven templates can be utilized in future modelling studies.

Conclusions: We developed a taxonomy of ILD sensitivity functions using a methodological data classification approach. The number and types of generic ILD function patterns found with this method mapped well on to our electrophysiologically determined ILD sensitivity functions. While a larger data set of the latter functions may bring a more robust outcome, this good mapping is encouraging in providing a principled method for classifying such data sets, and could be well extended to other such neuronal sensitivity functions, such as contrast tuning in vision.

PubMed Disclaimer

Figures

Figure 1
Figure 1
The data classification procedure. The three step data classification process resulted with seven ILD sensitivity functions data groups. In this three-step process; (1st) several data normalization techniques tested for our data and the UTPM (unit total probability mass) data normalization method was most suitable one, (2nd) first three principal components of PCA were selected and these were good enough to present entire data by the 97.6 % variance explained, (3rd) Cluster Analysis based on the Ward linkage and Cosine pairwise-distance algorithms those selected algorithms helped to compose a dendrogram where Inconsistency Coefficient determines the number of clustered data.
Figure 2
Figure 2
The raw and normalized data compared. The raw as unprocessed data consists of 208 vectors, and each vector (panels) has got varying number of spike counts between zero and 188 for 13 ILD levels from −30 dB to +30 dB with the increment of 5 dB (A), and (Additional file 1). The normalized data by the UTPM function perfectly preserves the shapes of raw-data while scales the number of spike counts down by %92.38 (B), Table 1. Therefore, there is not much differences observed between normalized data in (B) and raw data in (A).
Figure 3
Figure 3
The Scree-plot used for determining the number of principal components. The Scree-plot (the lines above the bar plots) and variance explained by the percentage bar plots, are both used for the number of principal component selection towards PCA for seven normalization techniques. Raw (A) and seven different normalized data (B-H) all applied for PCA. In a result, the variances information of each set of principal components (PC1, PC2, PC3 ... and PC13) is extracted from the PCA to show the significance. Either higher variance values of principal components, or prior to bending point "elbow" in the Scree-plot, they both indicate necessary number of principal component usage for the reduced data dimension representation.
Figure 4
Figure 4
First three principal components are depicted in 3D. The selection of first three principal components is decided by the Scree-plot (Figure 3G), and expressed for 208 normalized data (circles) in three-dimensional plot, (A). These transformed values are viewed by pairs in two-dimensional: First and second principal components (B), first and third principal components (C), and second and third components in (D). All zero origins are marked "⊗" for a reference point with the line axes.
Figure 5
Figure 5
The selection of inconsistency coefficient. The Inconsistency Coefficient varies by the number of clusters (as a depth value in Dendrogram) for number of linkage distributions. This distribution becomes in more compact form around the maximum value of Inconsistency Coefficient of seven. This value suggests the cut-off point for the Dendrogram or the number of cluster selection.
Figure 6
Figure 6
Seven clustered data represented by the Dendrogram. This Dendrogram is extracted from Figure 12 with the cut-off point of seven. The homogenous distribution of seven clustered data emphasized by their number of objects (bold parenthesized) for each cluster, i.e. Cluster-4 contains 25 similar ILD functions, which are also close relative of Cluster-6, which consists of 11 similar ILD functions.
Figure 7
Figure 7
Seven types of ILD functions observed. Typical four ideal ILD functions (Figure 8) can easily be perceived among these seven type of ILD functions here; what makes the another three "transitional" cluster findings is significantly important in this study. Type of ILD functions are derived from each clustered data by averaging their objects. All maximum numbers of mean spike counts is scaled up to 45 for a comparison reason. For example, The Cluster-4 shows peak type ILD functions by averaging its (25) objects where the Cluster-6 also shows arisen-peak ILD functions by averaging its (11) objects. These numbers of objects are also shown in Figure 9.
Figure 8
Figure 8
The four ILD functions. Typical four ideal ILD functions (A), namely they are, Sigmoidal (EI), Sigmoidal (IE), Peak, and Insensitive. These four ILD type representations are slightly perturbed to give more realistic aspect (B). Four ILD patterns are described in numbers of spike counts "#sp.c." (spikes/ stimulus) which varied between maxima of ‘m’ units (m ∈ א) and minima of ‘0’ zero unit, within -30dB to +30dB interaural level differences.
Figure 9
Figure 9
The Voronoi diagram for seven clustered data. Each cluster (from Cluster-1 to Cluster-7) holds number of objects (bracketed, ‘n’) and Seven ILD functions ensemble by the averaged for each clustered-data, these are; ILD-1= mean(Cluster-1), ILD-2= mean(Cluster-2) ... ILD-7= mean(Cluster-7), from Figure 7. These seven ILD functions are positioned around the Voronoi diagram of clustered data to show the relationship between clustered data and its representation of the ILD function. Seven clustered-data and contained ‘n’ number of objects (‘n’/208) are viewed; Cluster-1 (61/208), Cluster-2 (19/208), Cluster-3 (36/208), Cluster-4 (25/208), Cluster-5 (21/208), Cluster-6 (11/208), Cluster-7 (35/208).
Figure 10
Figure 10
The simulation of nine possible ILD functions. Nine possible ILD functions are generated from four typical ILD sensitive function variations (Figure 8B) These are; Sigmoidals with varying number of spike count (# sp.c.) spikes/ stimulus (A), position of the cut-off (B), the steepness of the slope (C), and four Peaked with varying number of spikes/stimulus (D), the cut-off (E), the cut-off & slope (F), and Peaked with unilateral transition to Sigmoidal(G), and Peaked with bilateral transition to Insensitive(H), and four Insensitive with varying the number of spike count spikes/ stimulus (I).
Figure 11
Figure 11
Cophenetic Correlation Coefficient (CCC) utilization for dendrogram. The CCC determines suitable algorithms for dendrogram the combination of 24 different (six pairwise-distance and four linkage) algorithms used to find most two suitable algorithms for the dendrogram plot. The optimization criteria using CCCs applied for 24 combined algorithms are also shown in Table 3. The algorithms used for pairwise-distance and linkage methods and CCC application operated from MATLAB version 6.5.
Figure 12
Figure 12
Dendrogram for cluster composition. Using CCC as a dissimilarity cluster measurement to select Cosine pairwise-distance and Ward linkage algorithms for cluster composition in the Dendrogram. The horizontal axis represents the all observation of subjects with 207 (208–1) numbers of clustered nodes; vertical axis represents the distance between the observed subjects in a logarithmic scale from 10-5 to 101.

References

    1. Buonomano DV. The biology of time across different scales. Nat Chem Biol. 2007;14(10):594–597. doi: 10.1038/nchembio1007-594. - DOI - PubMed
    1. Irvine DR. Interaural intensity differences in the cat: changes in sound pressure level at the two ears associated with azimuthal displacements in the frontal horizontal plane. Hear Res. 1987;14(3):267–286. doi: 10.1016/0378-5955(87)90063-3. - DOI - PubMed
    1. Hartmann WM, Rakerd B. Interaural level differences: Diffraction and localization by human listeners. J Acoust Soc Am. 2011;14(4):2622.
    1. Park TJ, Pollak GD. GABA shapes sensitivity to interaural intensity disparities in the mustache bat's inferior colliculus: implications for encoding sound location. J Neurosci. 1993;14(5):2050–2067. - PMC - PubMed
    1. Park TJ, Klug A, Holinstat M, Grothe B. Interaural level difference processing in the lateral superior olive and the inferior colliculus. Journal Of Neurophysiology. 2004;14(1):289–301. doi: 10.1152/jn.00961.2003. - DOI - PubMed

Publication types

LinkOut - more resources

Full text links
BioMed Central full text link BioMed Central Free PMC article
Cite

AltStyle によって変換されたページ (->オリジナル) /