Texture-based classification of focal liver lesions on MRI at 3.0 Tesla: a feasibility study in cysts and hemangiomas

Multi-institutional validation of a novel textural analysis tool for preoperative stratification of suspected thyroid tumors on diffusion-weighted MRI

PURPOSE

Ultrasound-guided fine needle aspirate cytology fails to diagnose many malignant thyroid nodules; consequently, patients may undergo diagnostic lobectomy. This study assessed whether textural analysis (TA) could noninvasively stratify thyroid nodules accurately using diffusion-weighted MRI (DW-MRI).

METHODS

This multi-institutional study examined 3T DW-MRI images obtained with spin echo echo planar imaging sequences. The training data set included 26 patients from Cambridge, United Kingdom, and the test data set included 18 thyroid cancer patients from Memorial Sloan Kettering Cancer Center (New York, New York, USA). Apparent diffusion coefficients (ADCs) were compared over regions of interest (ROIs) defined on thyroid nodules. TA, linear discriminant analysis (LDA), and feature reduction were performed using the 21 MaZda-generated texture parameters that best distinguished benign and malignant ROIs.

RESULTS

Training data set mean ADC values were significantly different for benign and malignant nodules (P = 0.02) with a sensitivity and specificity of 70% and 63%, respectively, and a receiver operator characteristic (ROC) area under the curve (AUC) of 0.73. The LDA model of the top 21 textural features correctly classified 89/94 DW-MRI ROIs with 92% sensitivity, 96% specificity, and an AUC of 0.97. This algorithm correctly classified 16/18 (89%) patients in the independently obtained test set of thyroid DW-MRI scans.

CONCLUSION

TA classifies thyroid nodules with high sensitivity and specificity on multi-institutional DW-MRI data sets. This method requires further validation in a larger prospective study. Magnetic Resonance in Medicine published by Wiley Periodicals, Inc. on behalf of International Society for Magnetic Resonance.

Radiomics: a new application from established techniques

The increasing use of biomarkers in cancer have led to the concept of personalized medicine for patients. Personalized medicine provides better diagnosis and treatment options available to clinicians. Radiological imaging techniques provide an opportunity to deliver unique data on different types of tissue. However, obtaining useful information from all radiological data is challenging in the era of "big data". Recent advances in computational power and the use of genomics have generated a new area of research termed Radiomics. Radiomics is defined as the high throughput extraction of quantitative imaging features or texture (radiomics) from imaging to decode tissue pathology and creating a high dimensional data set for feature extraction. Radiomic features provide information about the gray-scale patterns, inter-pixel relationships. In addition, shape and spectral properties can be extracted within the same regions of interest on radiological images. Moreover, these features can be further used to develop computational models using advanced machine learning algorithms that may serve as a tool for personalized diagnosis and treatment guidance.

Functional imaging of hepatocellular carcinoma

Imaging plays a key role in the clinical management of hepatocellular carcinoma (HCC), but conventional imaging techniques have limited sensitivity in visualizing small tumors and assessing response to locoregional treatments and sorafenib. Functional imaging techniques allow visualization of organ and tumor physiology. Assessment of functional characteristics of tissue, such as metabolism, proliferation and stiffness, may overcome some of the limitations of structural imaging. In particular, novel molecular imaging agents offer a potential tool for early diagnosis of HCC, and radiomics may aid in response assessment and generate prognostic models. Further prospective research is warranted to evaluate emerging techniques and their cost-effectiveness in the context of HCC in order to improve detection and response assessment.

Texture analysis in gel electrophoresis images using an integrative kernel-based approach

Texture information could be used in proteomics to improve the quality of the image analysis of proteins separated on a gel. In order to evaluate the best technique to identify relevant textures, we use several different kernel-based machine learning techniques to classify proteins in 2-DE images into spot and noise. We evaluate the classification accuracy of each of these techniques with proteins extracted from ten 2-DE images of different types of tissues and different experimental conditions. We found that the best classification model was FSMKL, a data integration method using multiple kernel learning, which achieved AUROC values above 95% while using a reduced number of features. This technique allows us to increment the interpretability of the complex combinations of textures and to weight the importance of each particular feature in the final model. In particular the Inverse Difference Moment exhibited the highest discriminating power. A higher value can be associated with an homogeneous structure as this feature describes the homogeneity; the larger the value, the more symmetric. The final model is performed by the combination of different groups of textural features. Here we demonstrated the feasibility of combining different groups of textures in 2-DE image analysis for spot detection.

Three-dimensional texture analysis of contrast enhanced CT images for treatment response assessment in Hodgkin lymphoma: comparison with F-18-FDG PET

PURPOSE

To determine the diagnostic performance of three-dimensional (3D) texture analysis (TA) of contrast-enhanced computed tomography (CE-CT) images for treatment response assessment in patients with Hodgkin lymphoma (HL), compared with F-18-fludeoxyglucose (FDG) positron emission tomography/CT.

METHODS

3D TA of 48 lymph nodes in 29 patients was performed on venous-phase CE-CT images before and after chemotherapy. All lymph nodes showed pathologically elevated FDG uptake at baseline. A stepwise logistic regression with forward selection was performed to identify classic CT parameters and texture features (TF) that enable the separation of complete response (CR) and persistent disease.

RESULTS

The TF fraction of image in runs, calculated for the 45° direction, was able to correctly identify CR with an accuracy of 75%, a sensitivity of 79.3%, and a specificity of 68.4%. Classical CT features achieved an accuracy of 75%, a sensitivity of 86.2%, and a specificity of 57.9%, whereas the combination of TF and CT imaging achieved an accuracy of 83.3%, a sensitivity of 86.2%, and a specificity of 78.9%.

CONCLUSIONS

3D TA of CE-CT images is potentially useful to identify nodal residual disease in HL, with a performance comparable to that of classical CT parameters. Best results are achieved when TA and classical CT features are combined.

Angiomyolipoma with minimal fat: differentiation from clear cell renal cell carcinoma and papillary renal cell carcinoma by texture analysis on CT images

RATIONALE AND OBJECTIVES

To retrospectively evaluate the diagnostic performance of texture analysis (TA) for the discrimination of angiomyolipoma (AML) with minimal fat, clear cell renal cell cancer (ccRCC), and papillary renal cell cancer (pRCC) on computed tomography (CT) images and to determine the scanning phase, which contains the strongest discriminative power.

MATERIALS AND METHODS

Patients with pathologically proved AMLs (n = 18) lacking visible macroscopic fat at CT and patients with pathologically proved ccRCCs (n = 18) and pRCCs (n = 14) were included. All patients underwent CT scan with three phases (precontrast phase [PCP], corticomedullary phase [CMP], and nephrographic phase [NP]). The selected images were analyzed and classified with TA software (MaZda). Texture classification was performed for 1) minimal fat AML versus ccRCC, 2) minimal fat AML versus pRCC, and 3) ccRCC versus pRCC. The classification results were arbitrarily divided into several levels according to the misclassification rates: excellent (misclassification rates ≤10%), good (10%< misclassification rates ≤20%), moderate (20%< misclassification rates ≤30%), fair (30%< misclassification rates ≤40%), and poor (misclassification rates ≥40%).

RESULTS

Excellent classification results (error of 0.00%-9.30%) were obtained with nonlinear discriminant analysis for all the three groups, no matter which phase was used. On comparison of the three scanning phases, we observed a trend toward better lesion classification with PCP for minimal fat AML versus ccRCC, CMP, and NP images for ccRCC versus pRCC and found similar discriminative power for minimal fat AML versus pRCC.

CONCLUSIONS

TA might be a reliable quantitative method for the discrimination of minimal fat AML, ccRCC, and pRCC.

Hepatitis C related chronic liver cirrhosis: feasibility of texture analysis of MR images for classification of fibrosis stage and necroinflammatory activity grade

PURPOSE

To assess the feasibility of texture analysis for classifying fibrosis stage and necroinflammatory activity grade in patients with chronic hepatitis C on T2-weighted (T2W), T1-weighted (T1W) and Gd-EOB-DTPA-enhanced hepatocyte-phase (EOB-HP) imaging.

MATERIALS AND METHODS

From April 2008 to June 2012, MR images from 123 patients with pathologically proven chronic hepatitis C were retrospectively analyzed. Texture parameters derived from histogram, gradient, run-length matrix, co-occurrence matrix, autoregressive model and wavelet transform methods were estimated with imaging software. Fisher, probability of classification error and average correlation, and mutual information coefficients were used to extract subsets of optimized texture features. Linear discriminant analysis in combination with 1-nearest neighbor classifier (LDA/1-NN) was used for lesion classification. In compliance with the software requirement, classification was performed based on datasets from all patients, the patient group with necroinflammatory activity grade 1, and that with fibrosis stage 4, respectively.

RESULTS

Based on all patient dataset, LDA/1-NN produced misclassification rates of 28.46%, 35.77% and 20.33% for fibrosis staging and 34.15%, 25.20% and 28.46% for necroinflammatory activity grading in T2W, T1W and EOB-HP images. In the patient group with necroinflammatory activity grade 1, LDA/1-NN yielded misclassification rates of 5.00%, 0% and 12.50% for fibrosis staging in T2W, T1W and EOB-HP images respectively. In the patient group with fibrosis stage 4, LDA/1-NN yielded misclassification rates of 5.88%, 12.94% and 11.76% for necroinflammatory activity grading in T2W, T1W and EOB-HP images respectively.

CONCLUSION

Texture quantitative parameters of MR images facilitate classification of the fibrosis stage as well as necroinflammatory activity grade in chronic hepatitis C, especially after categorizing the input dataset according to the activity or fibrosis degree in order to remove the interference between the fibrosis stage and necroinflammatory activity grade on texture features.

Textural differences in apparent diffusion coefficient between low- and high-stage clear cell renal cell carcinoma

OBJECTIVE

The purpose of this article is to evaluate differences in texture measures on apparent diffusion coefficient (ADC) maps between low- and high-stage clear cell renal cell carcinomas (RCCs).

MATERIALS AND METHODS

In this retrospective study, 61 patients with clear cell RCC at pathologic examination and who underwent preoperative MRI with diffusion-weighted imaging were included. Clear cell RCCs were clinically staged on review of preoperative MRI by a board-certified radiologist blinded to the pathologic findings. Whole lesions were segmented on ADC maps by two readers independently, from which first-order texture features (i.e., mean and skewness) and second-order texture features (i.e., cooccurrence matrix measures) were calculated. Texture metrics were compared between low- and high-stage clear cell RCC.

RESULTS

In 61 patients, there were 62 clear cell RCCs (33 low stage [stages I and II] and 29 high stage [stages III and IV]) at pathologic examination. Staging accuracy of qualitative interpretation was 100% for low-stage lesions and 37.9% (11/29) for high-stage lesions. There was no statistically significant difference in mean ADC between high- and low-stage clear cell RCCs (1.77×10(-3) vs 1.80×10(-3) mm2/s; p=0.7). However, high-stage clear cell RCCs were larger (6.96±2.93 vs 3.49±1.57 cm; p<0.0001) and had statistically significantly (p≤0.0001) higher ADC skewness (0.02±0.33 vs -0.52±0.65) and cooccurrence matrix correlation (0.64±0.11 vs 0.49±0.13). Multivariate logistic regression identified size, skewness, and cooccurrence matrix correlation as significant independent predictors of high stage (AUC=0.92). Interreader correlation in texture metrics ranged from 0.82 to 0.89.

CONCLUSION

First- and second-order ADC texture metrics differ between low- and high-stage clear cell RCCs. A model that includes size and ADC texture measures may help to stage clear cell RCCs noninvasively.

Quantification of heterogeneity as a biomarker in tumor imaging: a systematic review

BACKGROUND

Many techniques are proposed for the quantification of tumor heterogeneity as an imaging biomarker for differentiation between tumor types, tumor grading, response monitoring and outcome prediction. However, in clinical practice these methods are barely used. This study evaluates the reported performance of the described methods and identifies barriers to their implementation in clinical practice.

METHODOLOGY

The Ovid, Embase, and Cochrane Central databases were searched up to 20 September 2013. Heterogeneity analysis methods were classified into four categories, i.e., non-spatial methods (NSM), spatial grey level methods (SGLM), fractal analysis (FA) methods, and filters and transforms (F&T). The performance of the different methods was compared.

PRINCIPAL FINDINGS

Of the 7351 potentially relevant publications, 209 were included. Of these studies, 58% reported the use of NSM, 49% SGLM, 10% FA, and 28% F&T. Differentiation between tumor types, tumor grading and/or outcome prediction was the goal in 87% of the studies. Overall, the reported area under the curve (AUC) ranged from 0.5 to 1 (median 0.87). No relation was found between the performance and the quantification methods used, or between the performance and the imaging modality. A negative correlation was found between the tumor-feature ratio and the AUC, which is presumably caused by overfitting in small datasets. Cross-validation was reported in 63% of the classification studies. Retrospective analyses were conducted in 57% of the studies without a clear description.

CONCLUSIONS

In a research setting, heterogeneity quantification methods can differentiate between tumor types, grade tumors, and predict outcome and monitor treatment effects. To translate these methods to clinical practice, more prospective studies are required that use external datasets for validation: these datasets should be made available to the community to facilitate the development of new and improved methods.

MRI texture analysis predicts p53 status in head and neck squamous cell carcinoma

BACKGROUND AND PURPOSE

Head and neck cancer is common, and understanding the prognosis is an important part of patient management. In addition to the Tumor, Node, Metastasis staging system, tumor biomarkers are becoming more useful in understanding prognosis and directing treatment. We assessed whether MR imaging texture analysis would correctly classify oropharyngeal squamous cell carcinoma according to p53 status.

MATERIALS AND METHODS

A cohort of 16 patients with oropharyngeal squamous cell carcinoma was prospectively evaluated by using standard clinical, histopathologic, and imaging techniques. Tumors were stained for p53 and scored by an anatomic pathologist. Regions of interest on MR imaging were selected by a neuroradiologist and then analyzed by using our 2D fast time-frequency transform tool. The quantified textures were assessed by using the subset-size forward-selection algorithm in the Waikato Environment for Knowledge Analysis. Features found to be significant were used to create a statistical model to predict p53 status. The model was tested by using a Bayesian network classifier with 10-fold stratified cross-validation.

RESULTS

Feature selection identified 7 significant texture variables that were used in a predictive model. The resulting model predicted p53 status with 81.3% accuracy (P < .05). Cross-validation showed a moderate level of agreement (κ = 0.625).

CONCLUSIONS

This study shows that MR imaging texture analysis correctly predicts p53 status in oropharyngeal squamous cell carcinoma with ∼80% accuracy. As our knowledge of and dependence on tumor biomarkers expand, MR imaging texture analysis warrants further study in oropharyngeal squamous cell carcinoma and other head and neck tumors.

Texture-based and diffusion-weighted discrimination of parotid gland lesions on MR images at 3.0 Tesla

The purpose of this study was to evaluate whether texture-based analysis of standard MRI sequences and diffusion-weighted imaging can help in the discrimination of parotid gland masses. The MR images of 38 patients with a biopsy- or surgery-proven parotid gland mass were retrospectively analyzed. All patients were examined on the same 3.0 Tesla MR unit, with one standard protocol. The ADC (apparent diffusion coefficient) values of the tumors were measured with three regions of interest (ROIs) covering the entire tumor. Texture-based analysis was performed with the texture analysis software MaZda (version 4.7), with ROI measurements covering the entire tumor in three slices. COC (co-occurrence matrix), RUN (run-length matrix), GRA (gradient), ARM (auto-regressive model), and WAV (wavelet transform) features were calculated for all ROIs. Three subsets of 10 texture features each were used for a linear discriminant analysis (LDA) in combination with k nearest neighbor classification (k-NN). Using histology as a standard of reference, benign tumors, including subtypes, and malignant tumors were compared with regard to ADC and texture-based values, with a one-way analysis of variance with post-hoc t-tests. Significant differences were found in the mean ADC values between Warthin tumors and pleomorphic adenomas, as well as between Warthin tumors and benign lesions. Contrast-enhanced T1-weighted images contained the most relevant textural information for the discrimination between benign and malignant parotid masses, and also for the discrimination between pleomorphic adenomas and Warthin tumors. STIR images contained the least relevant texture features, particularly for the discrimination between pleomorphic adenomas and Warthin tumors. Texture analysis proved to differentiate benign from malignant lesions, as well as pleomorphic adenomas from Warthin tumors, based on standard T(1w) sequences (without and with contrast). Of all benign parotid masses, Warthin tumors had significantly lower ADC values than the other entities.

Texture-based classification of different gastric tumors at contrast-enhanced CT

PURPOSE

To determine the feasibility of texture analysis for the classification of gastric adenocarcinoma, lymphoma, and gastrointestinal stromal tumors on contrast-enhanced hydrodynamic-MDCT images.

MATERIALS AND METHODS

The arterial phase scans of 47 patients with adenocarcinoma (AC) and a histologic tumor grade of [AC-G1, n=4, G1, n=4; AC-G2, n=7; AC-G3, n=16]; GIST, n=15; and lymphoma, n=5, and the venous phase scans of 48 patients with AC-G1, n=3; AC-G2, n=6; AC-G3, n=14; GIST, n=17; lymphoma, n=8, were retrospectively reviewed. Based on regions of interest, texture analysis was performed, and features derived from the gray-level histogram, run-length and co-occurrence matrix, absolute gradient, autoregressive model, and wavelet transform were calculated. Fisher coefficients, probability of classification error, average correlation coefficients, and mutual information coefficients were used to create combinations of texture features that were optimized for tumor differentiation. Linear discriminant analysis in combination with a k-nearest neighbor classifier was used for tumor classification.

RESULTS

On arterial-phase scans, texture-based lesion classification was highly successful in differentiating between AC and lymphoma, and GIST and lymphoma, with misclassification rates of 3.1% and 0%, respectively. On venous-phase scans, texture-based classification was slightly less successful for AC vs. lymphoma (9.7% misclassification) and GIST vs. lymphoma (8% misclassification), but enabled the differentiation between AC and GIST (10% misclassification), and between the different grades of AC (4.4% misclassification). No texture feature combination was able to adequately distinguish between all three tumor types.

CONCLUSION

Classification of different gastric tumors based on textural information may aid radiologists in establishing the correct diagnosis, at least in cases where the differential diagnosis can be narrowed down to two histological subtypes.

MALDI imaging MS reveals candidate lipid markers of polycystic kidney disease

Autosomal recessive polycystic kidney disease (ARPKD) is a severe, monogenetically inherited kidney and liver disease. PCK rats carrying the orthologous mutant gene serve as a model of human disease, and alterations in lipid profiles in PCK rats suggest that defined subsets of lipids may be useful as molecular disease markers. Whereas MALDI protein imaging mass spectrometry (IMS) has become a promising tool for disease classification, widely applicable workflows that link MALDI lipid imaging and identification as well as structural characterization of candidate disease-classifying marker lipids are lacking. Here, we combine selective MALDI imaging of sulfated kidney lipids and Fisher discriminant analysis (FDA) of imaging data sets for identification of candidate markers of progressive disease in PCK rats. Our study highlights strong increases in lower mass lipids as main classifiers of cystic disease. Structure determination by high-resolution mass spectrometry identifies these altered lipids as taurine-conjugated bile acids. These sulfated lipids are selectively elevated in the PCK rat model but not in models of related hepatorenal fibrocystic diseases, suggesting that they be molecular markers of the disease and that a combination of MALDI imaging with high-resolution MS methods and Fisher discriminant data analysis may be applicable for lipid marker discovery.

Customized first and second order statistics based operators to support advanced texture analysis of MRI images

Texture analysis is the process of highlighting key characteristics thus providing an exhaustive and unambiguous mathematical description of any object represented in a digital image. Each characteristic is connected to a specific property of the object. In some cases the mentioned properties represent aspects visually perceptible which can be detected by developing operators based on Computer Vision techniques. In other cases these properties are not visually perceptible and their computation is obtained by developing operators based on Image Understanding approaches. Pixels composing high quality medical images can be considered the result of a stochastic process since they represent morphological or physiological processes. Empirical observations have shown that these images have visually perceptible and hidden significant aspects. For these reasons, the operators can be developed by means of a statistical approach. In this paper we present a set of customized first and second order statistics based operators to perform advanced texture analysis of Magnetic Resonance Imaging (MRI) images. In particular, we specify the main rules defining the role of an operator and its relationship with other operators. Extensive experiments carried out on a wide dataset of MRI images of different body regions demonstrating usefulness and accuracy of the proposed approach are also reported.

A computerized MRI biomarker quantification scheme for a canine model of Duchenne muscular dystrophy

PURPOSE

Golden retriever muscular dystrophy (GRMD) is a widely used canine model of Duchenne muscular dystrophy (DMD). Recent studies have shown that magnetic resonance imaging (MRI) can be used to non-invasively detect consistent changes in both DMD and GRMD. In this paper, we propose a semiautomated system to quantify MRI biomarkers of GRMD.

METHODS

Our system was applied to a database of 45 MRI scans from 8 normal and 10 GRMD dogs in a longitudinal natural history study. We first segmented six proximal pelvic limb muscles using a semiautomated full muscle segmentation method. We then performed preprocessing, including intensity inhomogeneity correction, spatial registration of different image sequences, intensity calibration of T2-weighted and T2-weighted fat-suppressed images, and calculation of MRI biomarker maps. Finally, for each of the segmented muscles, we automatically measured MRI biomarkers of muscle volume, intensity statistics over MRI biomarker maps, and statistical image texture features.

RESULTS

The muscle volume and the mean intensities in T2 value, fat, and water maps showed group differences between normal and GRMD dogs. For the statistical texture biomarkers, both the histogram and run-length matrix features showed obvious group differences between normal and GRMD dogs. The full muscle segmentation showed significantly less error and variability in the proposed biomarkers when compared to the standard, limited muscle range segmentation.

CONCLUSION

The experimental results demonstrated that this quantification tool could reliably quantify MRI biomarkers in GRMD dogs, suggesting that it would also be useful for quantifying disease progression and measuring therapeutic effect in DMD patients.

Assessment of tumor heterogeneity: an emerging imaging tool for clinical practice?

Background

Tumor spatial heterogeneity is an important prognostic factor, which may be reflected in medical images

Methods

Image texture analysis is an approach of quantifying heterogeneity that may not be appreciated by the naked eye. Different methods can be applied including statistical-, model-, and transform-based methods.

Results

Early evidence suggests that texture analysis has the potential to augment diagnosis and characterization as well as improve tumor staging and therapy response assessment in oncological practice.

Conclusion

This review provides an overview of the application of texture analysis with different imaging modalities, CT, MRI, and PET, to date and describes the technical challenges that have limited its widespread clinical implementation so far. With further efforts to refine its application, image texture analysis has the potential to develop into a valuable clinical tool for oncologic imaging.

Teaching Points

• Tumor spatial heterogeneity is an important prognostic factor.

• Image texture analysis is an approach of quantifying heterogeneity.

• Different methods can be applied, including statistical-, model-, and transform-based methods.

• Texture analysis could improve the diagnosis, tumor staging, and therapy response assessment.

[New developments in MRI of the liver]

Radiology has gained an exceptional position in medicine because a correct diagnosis is the most crucial issue in determining an accurate and personalized therapeutic strategy. This has a direct influence not only on the individual patient but also on the socio-economic aspects of healthcare services in terms of shortening the time interval to establish a diagnosis and to avoid risk-associated invasive diagnostic methods or long-term, cost-intensive follow-up. Magnetic resonance imaging (MRI) is an excellent example of this which due to continuous technological developments and emerging techniques allows a non-invasive diagnosis of the different hepatic diseases. In this article, we illustrate the direct correlation between the recent technical advances in MRI, such as 3.0 T, diffusion-weighted imaging, perfusion imaging, spectroscopy, texture analysis and MR elastography and obtaining a confident non-invasive diagnosis of focal and diffuse liver diseases.