Textural Features for Image Classification

Haralick, Robert M.; Shanmugam, Karthikeyan; Dinstein, I.

doi:10.1109/tsmc.1973.4309314

Cited by 19,404 publications

(12,011 citation statements)

References 8 publications

Supporting

Mentioning

11,557

Contrasting

Unclassified

427

Order By: Relevance

“…These are related to nuclear area, total optical density and chromatin distribution. [11][12][13] Table 1 gives a sample list of features used in the discriminant function analyses and in the unsupervised learning program P-index [11][12][13] (see below) (all features are given in relative units of measure) (the values in parenthesis refer to an arbitrary code number with which the feature is identified in the computer program). The P-index groupings are based on two composite features: the discriminant function I score, and the nuclear abnormality.…”

Section: Karyometric and Statistical Analysesmentioning

confidence: 99%

See 1 more Smart Citation

Karyometry detects subvisual differences in chromatin organization state between cribriform and flat high-grade prostatic intraepithelial neoplasia

et al. 2004

View full text Add to dashboard Cite

This digital texture analysis-based study evaluates the chromatin organization state in flat and cribriform highgrade prostatic intraepithelial neoplasia (PIN), in the adjacent normal looking secretory epithelium and in the co-occurring adenocarcinoma. Digital texture analysis (karyometry) was carried out on hematoxylin and eosinstained sections from 24 radical prostatectomy specimens with high-grade PIN (12 with flat and 12 with cribriform architectural pattern, respectively) and cancer. Quantification was also conducted on the normal looking secretory epithelium. Discriminant analysis and the nonsupervised learning algorithm P-index were used to identify suitable subsets of features useful for the discrimination and classification of pathological groups and to explore multivariate data structure in the pathological subgroups. The average nuclear abnormality increases monotonically from the histologically normal appearing secretory epithelium to highgrade PIN and to adenocarcinoma. The nuclei from the so-called perimeter compartment of the flat high-grade PIN lesions show a higher nuclear abnormality compared to the nuclei of the cribriform high-grade PINs. Discriminant analysis shows that flat and cribriform high-grade PINs fall into two populations. Processing by the nonsupervised learning algorithm P-index revealed the existence of three well-defined, distinct subpopulations of nuclei of different chromatin phenotype. In the flat high-grade PIN lesions the proportions of nuclei in the three subpopulations are 16.5% (low abnormality), 25.0% (mid abnormality) and 58.5% (high abnormality), respectively. In the cribriform high-grade PIN lesions, 100% of the nuclei are in the midabnormality subpopulation. These differences are also discernible in the co-occurring adenocarcinoma and the histologically normal appearing secretory epithelium. To conclude, karyometry and statistical analysis detect the existence of distinct cell subpopulations of different chromatin packaging and phenotype, with the nuclei from the flat high-grade PIN lesions, adjacent normal looking epithelium and co-occurring adenocarcinoma expressing a greater nuclear abnormality than in the specimens with cribriform high-grade PIN.

show abstract

Section: Karyometric and Statistical Analysesmentioning

confidence: 99%

“…[11][12][13] These values express the relative deviation of each feature from 'normal', as assessed from the set of 'normal' reference nuclei. These features have the advantage that they are based on a relative, 'internal' standard.…”

Section: Nuclear Abnormality Nuclear Signature and Lesion Signaturementioning

confidence: 99%

Karyometry detects subvisual differences in chromatin organization state between cribriform and flat high-grade prostatic intraepithelial neoplasia

et al. 2004

View full text Add to dashboard Cite

show abstract

“…Inspiratory and expiratory image datasets were then analyzed through a computational software developed in our laboratory which is capable of generating several hundred distinct metrics encompassing various aspects of lung physiology (e.g. pulmonary volumetric and gross tissue indices, attenuation histogram statistics, deformation indices, co-occurrence [12] and run-length [13] matrix texture indices, and attenuation mask indices), gleaned from the relevant literature. For this study, we only focused on a portion of these metrics .…”

Section: Methodsmentioning

confidence: 99%

“…It also includes more sophisticated second-order statistical quantities related to the texture of lung parenchyma, i.e. the co-occurrence [12] and run-length [13] matrix texture indices.…”

Section: Methodsmentioning

confidence: 99%

A Comparative Study of HRCT Image Metrics and PFT Values for Characterization of ILD and COPD

et al. 2012

View full text Add to dashboard Cite

Abstract. Several image metrics have been proposed for pulmonary assessment via thoracic high-resolution computed tomography (HRCT) for various pathologies. This paper describes a systematic analysis of the utility of such metrics for characterizing interstitial lung disease (ILD) and chronic obstructive pulmonary disease (COPD), in comparison to data from pulmonary function testing (PFT). HRCT inspiratory and expiratory images for 14 patients with ILD and 11 patients with COPD were acquired retrospectively. PFT values were also acquired retrospectively for each patient. Using a statistical feature selection scheme, our study demonstrates that the quantitative image features perform quite well in comparison with the clinically-used PFT values. In the first 25 selected features out of the total 114 mixed image metrics and PFT values, 21 are from the image metrics. The classification using mixed selected image features and PFT values outperforms using PFT values alone. Our study also shows that these image metrics are not redundant with respect to the PFT values for characterization of ILD and COPD.

show abstract

“…For describing autocorrelation texture information, a co-occurrence matrix is utilized. Each element of the cooccurrence C i; j d 0 ; θ 0 j ð Þ is defined as the joint probability of gray levels i and j separated by distance d′ and along direction θ′ [29]. The autocorrelation texture, ACOR, is defined as:…”

Section: Texture Information Comparison Functionmentioning

confidence: 99%

An Effective Approach of Lesion Segmentation Within the Breast Ultrasound Image Based on the Cellular Automata Principle

et al. 2012

View full text Add to dashboard Cite

In this paper, a novel lesion segmentation within breast ultrasound (BUS) image based on the cellular automata principle is proposed. Its energy transition function is formulated based on global image information difference and local image information difference using different energy transfer strategies. First, an energy decrease strategy is used for modeling the spatial relation information of pixels. For modeling global image information difference, a seed information comparison function is developed using an energy preserve strategy. Then, a texture information comparison function is proposed for considering local image difference in different regions, which is helpful for handling blurry boundaries. Moreover, two neighborhood systems (von Neumann and Moore neighborhood systems) are integrated as the evolution environment, and a similarity-based criterion is used for suppressing noise and reducing computation complexity. The proposed method was applied to 205 clinical BUS images for studying its characteristic and functionality, and several overlapping area error metrics and statistical evaluation methods are utilized for evaluating its performance. The experimental results demonstrate that the proposed method can handle BUS images with blurry boundaries and low contrast well and can segment breast lesions accurately and effectively.

show abstract

Textural Features for Image Classification

Cited by 19,404 publications

References 8 publications

Karyometry detects subvisual differences in chromatin organization state between cribriform and flat high-grade prostatic intraepithelial neoplasia

Karyometry detects subvisual differences in chromatin organization state between cribriform and flat high-grade prostatic intraepithelial neoplasia

A Comparative Study of HRCT Image Metrics and PFT Values for Characterization of ILD and COPD

An Effective Approach of Lesion Segmentation Within the Breast Ultrasound Image Based on the Cellular Automata Principle

Contact Info

Product

Resources

About