Texture Analysis in Magnetic Resonance Imaging: Review and Considerations for Future Applications

Abstract

Texture analysis is a technique used for the quantification of image texture. It has been successfully used in many fields, and in the past years it has been applied in magnetic resonance imaging (MRI) as a computer-aided diagnostic tool. Quantification of the intrinsic heterogeneity of different tissues and lesions is necessary as they are usually imperceptible to the human eye. In the present chapter, we describe texture analysis as a process consisting of six steps: MRI acquisition, region of interest (ROI) definition, ROI preprocessing, feature extraction, feature selection, and classification. There is a great variety of methods and techniques to be chosen at each step and all of them can somehow affect the outcome of the texture analysis application. We reviewed the literature regarding texture analysis in clinical MRI focusing on the important considerations to be taken at each step of the process in order to obtain maximum benefits and to avoid misleading results.

Assessment of prostate cancer prognostic Gleason grade group using zonal‐specific features extracted from biparametric MRI using a KNN classifier

“…As the lesion size varies, it is likely that some noncancerous tissue is included in the ROI and for some lesions not included the whole lesion. Previous studies have shown that delineation of the entire lesion improves accuracy compared to bounding box approach . A lesion delineation might improve the models; however, such delineation requires experienced personnel and is very time‐consuming.…”

Initial assessment of 3D magnetic resonance fingerprinting (MRF) towards quantitative brain imaging for radiation therapy

“…For example, quantitative analysis of intensity‐normalized longitudinal T1‐/T2‐weighted MRIs could distinguish radiation necrosis from tumor regression of brain metastasis and identify hypoxia pattern, an important radio‐resistant mechanism, in glioblastoma . However, any normalization approach will introduce bias to the normalized MRI data, which potentially leads to unwanted alternations of statistical properties of the MRI data and possible degradation of statistical power …”

Radiomics for diagnosis and assessing brain diseases: an approach based on texture analysis on magnetic resonance imaging

“…Radiomics can be applied to different/multiple imaging modalities, and the selection of the appropriate technique to investigate each disease or lesion depends on several factors, including the resolution of the images and the tissue under analysis. However, in the last years, magnetic resonance imaging (MRI) has become popular in radiomics studies due to its growing availability in the clinical routine and the resulting high-quality images that offer excellent anatomic details thanks to new advances in technology [13]. Radiomics practice overview showing the major steps: image acquisition, segmentation and pre-processing of the images, feature extraction, and data mining.…”

“…It is reported that the effect of the ROI size in the texture parameters becomes insignificant when using large areas in terms of pixels [62]. However, the region size depends directly on the imaging acquisition parameters and, although certain texture features may obey this rule, there is a wide range of texture features and not all of them have been proven to be unaffected by the ROI/VOI size [13]. Therefore, when performing texture analysis, it is important to assure that the regions are large enough to capture texture heterogeneity and that the difference between ROI/VOI sizes among groups is not statistically significant.…”

“…To avoid possible effects of the region size, equalsized geometric regions can be used, but, as mentioned before, the complete delineation of the ROI might offer better results. In conclusion, the use of a segmentation approach is recommended when the range of ROI/VOI sizes among samples is not significantly different or when the selected texture features are not affected by this difference; otherwise, geometric regions of the same size might be a better approach [13].…”