Texture analysis is a non-invasive, mathematical method assessing the spatial heterogeneity of regions of interest in medical imaging, its primary application is in the assessment of tumors. Although not a new topic of research, the past decade has seen a significant resurgence of texture analysis in the field of radiomics 1,2.
Traditionally, the interpretation of tumor bodies in medical imaging whether that be CT, MRI or x-ray will report on the size and parameter metrics 3. It is now well known that intratumor heterogeneity is a marker of malignancy; texture analysis attempts to provide a comprehensive quantitative analysis of heterogeneity via the assessment of pixels and voxels within a tumor image 2,3.
Statistical based modelling
Texture analysis employs a plethora of models to achieve an accurate assessment of tumor heterogeneity, including model-based, transform-based, and statistical-based 2,4. The utilization of statistical based modelling is the most common in texture analysis, involving three orders of measure parameters; first-order statistics, second-order statistics and higher-order statistics.
Explores the frequency distribution in the region of interest via a histogram, it does not consider pixels around the region of interest, the first order statistics measures parameters such as intensity, standard deviation, skewness, and kurtosis 2,4.
Explores via a run-length matrix, co-occurrence measurements assessing a length of pixels consecutively that have equal grey-level intensities. This will provide information regarding the texture of the region of interest. Fine texture will have shorter run lengths and a more consistent range of intensities and less fine, coarse regions having an opposite read 2,4.
Second order statistics via a grey-level co-occurrence matrix will explore similar parameters, however, will present how often pairs of pre-determined pixel values arise within a spatial range in the image.
Explores the overall differences between pixels or voxels within the context of the entire region of interest, often via the utilization of neighborhood grey-tone-difference matrix. Using higher order statistics one can obtain metrics such as variations within the image and the spatial rate of grey-level change. Higher order statistics provides a broader overall report of the region of interest texture metrics 2,4.
Clinical applications studies are substantial, and beyond the realm of this article, however as with all emerging technology/research, clinical validation is a priority 3.
The utilization of texture analysis presents a non-invasive method to identify and characterize tumors using conventional cross-sectional imaging such as CT and MRI. It enhances the characterization of tumor bodies using complex algorithms and has the potential to overcome the challenges of biopsy 4.
Texture analysis has also been explored to better characterize hepatic fibrosis, emphysema, and liver cirrhosis, although this research is also still in the experimental phase 5-7.
- 1. Radiomics and Deep Learning in Clinical Imaging: What Should We Do?. (2018) Nuclear Medicine and Molecular Imaging. 52 (2): 89. doi:10.1007/s13139-018-0514-0 - Pubmed
- 2. Davnall F, Yip CS, Ljungqvist G, Selmi M, Ng F, Sanghera B, Ganeshan B, Miles KA, Cook GJ, Goh V. Assessment of tumor heterogeneity: an emerging imaging tool for clinical practice?. (2012) Insights into imaging. 3 (6): 573-89. doi:10.1007/s13244-012-0196-6 - Pubmed
- 3. O'Connor JP, Rose CJ, Waterton JC, Carano RA, Parker GJ, Jackson A. Imaging intratumor heterogeneity: role in therapy response, resistance, and clinical outcome. (2015) Clinical cancer research : an official journal of the American Association for Cancer Research. 21 (2): 249-57. doi:10.1158/1078-0432.CCR-14-0990 - Pubmed
- 4. Lubner MG, Smith AD, Sandrasegaran K, Sahani DV, Pickhardt PJ. CT Texture Analysis: Definitions, Applications, Biologic Correlates, and Challenges. (2017) Radiographics : a review publication of the Radiological Society of North America, Inc. 37 (5): 1483-1503. doi:10.1148/rg.2017170056 - Pubmed
- 5. Liu H, Shao Y, Guo D, Zheng Y, Zhao Z, Qiu T. Cirrhosis classification based on texture classification of random features. (2014) Computational and mathematical methods in medicine. 2014: 536308. doi:10.1155/2014/536308 - Pubmed
- 6. Park HJ, Lee SM, Song JW, Lee SM, Oh SY, Kim N, Seo JB. Texture-Based Automated Quantitative Assessment of Regional Patterns on Initial CT in Patients With Idiopathic Pulmonary Fibrosis: Relationship to Decline in Forced Vital Capacity. (2016) AJR. American journal of roentgenology. 207 (5): 976-983. doi:10.2214/AJR.16.16054 - Pubmed
- 7. Lubner MG, Malecki K, Kloke J, Ganeshan B, Pickhardt PJ. Texture analysis of the liver at MDCT for assessing hepatic fibrosis. (2017) Abdominal radiology (New York). 42 (8): 2069-2078. doi:10.1007/s00261-017-1096-5 - Pubmed