Introduction
Process
Image acquisition
Image reconstruction and pre-processing
Pre-processing technique | Effect on image |
---|---|
Resampling | Changing the number of pixels in the image using interpolation (linear, polynomial, spline, etc.) |
Normalization or intensity standardization | Changing the range of pixel intensity values, in order to remove bias, scaling factors and outliers from the image |
Quantization of gray levels | Reduction of gray levels used to represent the image |
Motion correction | Reduction of motion confounds |
Filtering to remove noise and/or improve image characteristics | Laplacian: bringing out area of rapid intensity change and usually used for edge detection Gaussian: smoothing the image and reducing noise Edge filters: resulting in edge enhancement by calculating an approximation of the derivatives in horizontal and vertical directions Laws’ filters: emphasizing image textures of edge, spot, ripple, wave, undulation and oscillation Wavelet filtering or transform methods: decomposing the original image and offering some advantages, such as variation of the spatial resolution (to represent textures at the most appropriate scale), enhancement of the texture appearance and a very wide range of choices for the wavelet function that can be adjusted for specific applications |
Inhomogeneity correction | performed on MR images, where the residual effect of the variation of intensity, mainly caused by static magnetic field inhomogeneity and imperfections of the radiofrequency coils, is not eliminated by the previous normalization |
Segmentation
Features extraction
2D | Measure |
---|---|
Mesh Surface | The sum of all areas defined for each triangle in the mesh |
Pixel Surface | The surface area of a single pixel multiplicated by the number of pixels in the region of interest |
Perimeter | The sum of all perimeters of each line in the mesh circumference |
Perimeter to Surface ratio | The ratio of the Perimeter to the Mesh Surface |
Sphericity | The ratio of the perimeter of the tumor region to the perimeter of a circle with the same surface area as the tumor region and therefore a measure of the roundness of The shape of the tumor region relative to a circle |
Spherical Disproportion | The ratio of the perimeter of the tumor region to the perimeter of a circle with the same surface area as the tumor region, and by definition, the inverse of Sphericity |
Maximum 2D diameter | The largest pairwise Euclidean distance between tumor surface mesh vertices |
Major and Minor Axis Length | The largest and the second-largest axis length of the region of interest-enclosing ellipsoid and is calculated using the largest principal component |
The inverse ratio of the major and minor principal axis lengths that could be viewed as the extent to the section is circle-like (not elongated) than it is 1 dimensional line (maximally elongated) | |
3D | Measure |
Compactness 1 and 2 | A measure of how compact the shape of the tumor is relative to a sphere (most compact) |
Spherical disproportion | The ratio of the surface area of the tumor region to the surface area of a sphere with the same volume as the tumor region, and by definition, the inverse of Sphericity |
Sphericity | A measure of the roundness of the shape of the tumor region relative to a sphere |
Mesh Volume | the sum of all volumes defined for each face in the triangle mesh of the region of interest |
Voxel Volume | The volume of a single voxel multiplicated by the number of voxels in the region of interest |
Surface Area | the sum of all areas of each triangle in the mesh |
Surface Area to Volume Ratio | The ratio of the Surface Area to the Mesh Volume |
Maximum 3D diameter | The largest pairwise Euclidean distance between tumor surface mesh vertices |
Major, Minor and Least Axis Length | Respectively, the largest, the second-largest and the smallest axis length of the region of interest-enclosing ellipsoid, calculated using the largest principal component |
Elongation | The inverse ratio of the major and minor principal axis lengths that could be viewed as the extent to which a volume is longer than it is wide, i.e., is eccentric |
Flatness | The inverse ratio of the major and least axis lengths that could be viewed as the extent to which a volume is flat relative to its length |
Type of feature | Measure |
---|---|
Energy | A measure of the magnitude of voxel values in an image |
Total Energy | The value of Energy feature scaled by the volume of the voxel in cubic mm |
Entropy | a measure of the inherent randomness in the gray level intensities of the image |
Minimum | The lowest intensity present |
Maximum | The maximum gray level intensity within the region of interest |
10th and 90th percentile | The 10th and 90th percentile of the gray level intensity within the region of interest |
Mean | The average gray level intensity within the region of interest |
Median | The median gray level intensity within the region of interest |
Range | The range of gray values in the region of interest |
Interquartile Range | The range between the 25th and 75th percentile of the image array |
Mean Absolute Deviation (MAD) | The mean distance of all intensity values from the Mean Value of the image array |
Robust Mean Absolute Deviation (rMAD) | The mean distance of all intensity values from the Mean Value calculated on the subset of image array with gray levels in between, or equal to the 10th and 90th percentile |
Root Mean Squared (RMS) | The square-root of the mean of all the squared intensity values |
Standard Deviation | The amount of variation or dispersion from the Mean Value |
Skewness | The asymmetry of the distribution of values about the Mean value |
Kurtosis | a measure of the ‘peakedness’ of the distribution of values in the image region of interest |
Variance | The mean of the squared distances of each intensity value from the Mean value and a measure of the spread of the distribution about the mean |
Uniformity | The sum of the squares of each intensity value and a measure of the homogeneity of the image array |
Group of features | Type of feature | Measure |
---|---|---|
Gray Level Co-occurrence Matrix (GLCM) | Correlation | Linear dependency of gray level values to their respective voxels in the GLCM |
Difference Entropy | a measure of the randomness/variability in neighborhood intensity value differences | |
Difference Average | A measure of the relationship between occurrences of pairs with similar intensity values and occurrences of pairs with differing intensity values | |
Difference Variance | A measure of heterogeneity that places higher weights on differing intensity level pairs that deviate more from the mean | |
Joint Energy | A measure of homogeneous patterns in the image | |
Joint Entropy | A measure of the randomness/variability in neighborhood intensity values | |
Inverse Difference Moment (IDM), Inverse Difference Moment Normalized (IDMN) and Inverse Difference (ID) | A measure of the local homogeneity of an image | |
Autocorrelation | A measure of the magnitude of the fineness and coarseness of texture | |
Joint Average | The mean gray level intensity of the i distribution | |
Cluster Prominence | A measure of the skewness and asymmetry of the GLCM | |
Cluster Shade | A measure of the skewness and uniformity of the GLCM | |
Cluster Tendency | A measure of groupings of voxels with similar gray level values | |
Contrast | A measure of the local intensity variation, favoring values away from the diagonal | |
Informational Measure of Correlation (IMC) 1 and 2 and Maximal Correlation Coefficient (MCC) | A quantification of the complexity of the texture | |
Maximum Probability | Occurrences of the most predominant pair of neighboring intensity values | |
Sum Average | A measure of the relationship between occurrences of pairs with lower intensity values and occurrences of pairs with higher intensity values | |
Sum Entropy | A sum of neighborhood intensity value differences | |
Sum of Squares | A measure in the distribution of neighboring intensity level pairs about the mean intensity level in the GLCM | |
Gray Level Run Length Matrix (GLRLM) | Short Run Emphasis (SRE) | A measure of the distribution of short run lengths, with a greater value indicative of shorter run lengths and more fine textural textures |
Long Run Emphasis (LRE) | A measure of the distribution of long run lengths, with a greater value indicative of longer run lengths and more coarse structural textures | |
Gray Level Non-Uniformity (GLN) and Gray Level Non-Uniformity Normalized (GLNN) | A measure of the similarity of gray level intensity values in the image | |
Run Length Non-Uniformity (RLN) and Run Length Non-Uniformity Normalized (RLNN) | A measure of the similarity of run lengths throughout the image | |
Run Percentage (RP) | A measure of the coarseness of the texture by taking the ratio of number of runs and number of voxels in the ROI | |
Gray Level Variance (GLV) | The variance in gray level intensity for the runs | |
Run Variance (RV) | The variance in runs for the run lengths | |
Run Entropy (RE) | a measure of the uncertainty/randomness in the distribution of run lengths and gray levels | |
Low and High Gray Level Run Emphasis (LGLRE and HGLRE) | A measure of the distribution of low and higher gray level values | |
Short and Long Run Low Gray Level Emphasis (SRLGLE and LRLGLE) | A measure of the joint distribution of shorter and long run lengths with lower gray level values | |
Short and Long Run High Gray Level Run Emphasis (SRHGLE and LRHGLE) | A measure of the joint distribution of shorter and long run lengths with higher gray level values | |
Gray Level Size Zone Matrix (GLSZM) | Small and Large Area Emphasis (SAE and LAE) | A measure of the distribution of small and large size zones |
Gray Level Non-Uniformity (GLN) and Gray Level Non-Uniformity Normalized (GLNN) | A measure of the variability of gray level intensity values in the image | |
Size Zone Non-Uniformity (SZN) and Size Zone Non-Uniformity Normalized (SZNN) | A measure of the variability of size zone volumes in the image | |
Zone Percentage (ZP) | A measure of the coarseness of the texture by taking the ratio of number of zones and number of voxels in the region of interest | |
Gray Level Variance (GLV) | The variance in gray level intensities for the zones | |
Zone Variance (ZV) | The variance in zone size volumes for the zones | |
Zone Entropy (ZE) | A measure of the uncertainty/randomness in the distribution of zone sizes and gray levels | |
Low and High Gray Level Zone Emphasis (LGLZE and HGLZE) | A measure of the distribution of lower and higher gray level size zones | |
Small Area Low and High Gray Level Emphasis (SALGLE and SAHGLE) | A measure of the proportion in the image of the joint distribution of smaller size zones with lower and higher gray level values | |
Large Area Low and High Gray Level Emphasis (LALGLE and LAHGLE) | A measure of the proportion in the image of the joint distribution of larger size zones with lower gray level values | |
Neighboring Gray Tone Difference Matrix (NGTDM) | Coarseness | An average difference between the center voxel and its neighborhood and an indication of the spatial rate of change |
Contrast | A measure of the spatial intensity change, also dependent on the overall gray level dynamic range | |
Busyness | A measure of the change from a pixel to its neighbor | |
Complexity | A measure of non-uniformity and rapid changes in gray levels | |
Strength | A measure of the primitives in an image | |
Gray Level Dependence Matrix (GLDM) | Small and Large Dependence Emphasis (SDE and LDE) | A measure of the distribution of small and large dependencies |
Gray Level Non-Uniformity (GLN) | The similarity of gray level intensity values in the image | |
Dependence Non-Uniformity (DN) and Dependence Non-Uniformity Normalized (DNN) | The similarity of dependence throughout the image | |
Gray Level Variance | The variance in gray level in the image | |
Dependence Variance | the variance in dependence size in the image | |
Low and High Gray Level Emphasis (LGLE and HGLE) | The distribution of low and high gray level values | |
Small and Large Dependence Low Gray Level Emphasis (SDLGLE and LDLGLE) | The joint distribution of small and large dependence with lower gray level values | |
Small and Large Dependence High Gray Level Emphasis (SDHGLE and LDHGLE) | The joint distribution of small and large dependence with higher gray level values |
Imaging biobanks to collect and validate radiomics
Project | Description |
---|---|
PRIMAGE [37] | PRedictive In-silico Multiscale Analytics to support cancer personalized diaGnosis and prognosis, Empowered by imaging biomarkers) project, mainly focused on childhood cancer |
CHAIMELEON [38] | Accelerating the lab to market transition of AI tools for cancer management |
Procancer-I [39] | An AI Platform integrating imaging data and models, supporting precision care through prostate cancer’s continuum |
EuCanImage [40] | A European Cancer Image Platform Linked to Biological and Health Data for Next-Generation Artificial Intelligence and Precision Medicine in Oncology |
INCISIVE [41] | A multimodal AI-based toolbox and an interoperable health imaging repository for the empowerment of imaging analysis related to the diagnosis, prediction and follow-up of cancer |
EuCanShare [42] | An EU-Canada joint infrastructure for next-generation multi-Study Heart research |
Features analysis
Classifier model
Deep learning models
Radiomics in clinical practice
Challenges and potential solutions
Radiogenomics
Conclusion
In the present study, a thorough review of the typical radiomics analysis process is offered. A detailed explanation of all the steps used in the extraction of quantitative data from medical images, and their subsequent analysis is proposed. The aim was to summarize the several methods currently used in the various steps of the workflow, providing a deep technical overview of the analysis conducted in each step. This description stresses how all the different processes that can be used deeply affect the results and how this could be a problem for the repeatability and reliability of the analysis. Furthermore, we highlighted the reasons why radiomics analysis is of unique importance, encouraging the scientific community in establishing benchmarks and fostering the effective use of these promising research tools also in clinical settings.A brief overview of the latest and most relevant clinical applications of radiomics in oncology is presented, sorting through the possibilities of advancement in prediction, diagnosis and treatment evaluation mainly in the studies of brain, prostate, breast and lung cancers. Although it is almost impossible to explore all the different applications in a single review, our aim was to provide the reader a general idea of the extent of different possible application domains, which could make radiomics such a powerful quantitative analysis tool. The variety of the radiomics studies carried out in research show another key point, i.e., how radiomics could offer the physician a non-invasive tool for a personalized medicine approach to the patient, in particular with the development of radiogenomics. An important stressed issue is that of limitations and challenges related to reproducibility, data sharing and lack of standardization, for which potential solutions that would help, such as standardization strategies and data sharing development, have been addressed. Therefore, the take home message is that an enormous effort should be encouraged to overcome these limitations and move the field of radiomics toward clinical implementation, by using it as an effective support in clinical decisions.To summarize, a deep look into radiomics has been proposed, from the detailed description of current methods and different types of features that can be analyzed, to a wide and overall view of applications and future research directions, with a particular emphasis on the evolving branches of imaging biobanking and radiogenomics. Only an in-depth and comprehensive description of current methods and applications can reveal the potential power of radiomics and the need to translate the successful outcomes in research into an effective tool suitable in clinical practice.