Haralick Texture Features Research Articles

SU-D-BRA-01: Therapeutic Response Assessment Using a Novel Gray Level Local Power Matrix (GLLPM) in DCE-MRI Texture Analysis: Feasibility Study

Purpose: Most current DCE-MRI texture analysis methods focus on the spatial information of chosen contrast-enhanced MR volumes or pharmacokinetic (PK) parameter maps, and the temporal information is not well included. This work proposed a novel texture matrix called Gray Level Local Power Matrix (GLLPM) for the accurate and efficient spatiotemporal DCE-MRI texture analysis in therapeutic response assessment. Methods: A retrospective study with two groups (n=8/group) of tumor implanted mice was conducted. The treatment/control groups received bevacizumab/saline treatment with pre- and post-treatment DCE-MRI exams. For each scan, the GLLPM was calculated and compared with classic 3D/4D Gray Level Co-Occurrence Matrices (GLCOM) using the CA concentration maps in the first 10-minute post-injection time. The calculation time of each matrix was recorded for efficiency evaluation. Using each matrix, a set of 22 Haralick texture features’ dynamic curves were calculated. The Mann-Whitney U-test was used to assess the differences of the Area Under Curve (AUC) of all derived texture feature curves between treatment/control groups. The post-treatment texture feature curves were fitted by cubic polynomial. Experiments using support vector machine in a leave-one-out approach were performed to validate the use of fitted polynomial coefficients of each texture feature curve in treatment/control group classification. Results: The computation efficiency of GLLPM had improved factors of 3 and 20 in comparison with 3D/4D GLCOM, respectively. 21 out of 22 GLLPM texture feature dynamic curves’ AUCs between treatment/control groups had significant differences in post-treatment scan but not in pre-treatment scan. N=19 dynamic curves from GLLPM can be fitted by cubic polynomial (R2>0.8), and N for 3D/4D GLCOM were 14 and 19, respectively. The averaged classification accuracies using the post-treatment texture features curves based on GLLPM, 3D/4D GLCOM were (84.5±12.1)%, (65.6±10.5)% and (73.3±12.8)%, respectively. Conclusion: The proposed GLLPM and its features can be used for the efficient DCE-MRI therapeutic response assessment.

Haralick texture analysis of prostate MRI: utility for differentiating non-cancerous prostate from prostate cancer and differentiating prostate cancers with different Gleason scores.

To investigate Haralick texture analysis of prostate MRI for cancer detection and differentiating Gleason scores (GS). One hundred and forty-seven patients underwent T2- weighted (T2WI) and diffusion-weighted prostate MRI. Cancers ≥0.5ml and non-cancerous peripheral (PZ) and transition (TZ) zone tissue were identified on T2WI and apparent diffusion coefficient (ADC) maps, using whole-mount pathology as reference. Texture features (Energy, Entropy, Correlation, Homogeneity, Inertia) were extracted and analysed using generalized estimating equations. PZ cancers (n = 143) showed higher Entropy and Inertia and lower Energy, Correlation and Homogeneity compared to non-cancerous tissue on T2WI and ADC maps (p-values: <.0001-0.008). In TZ cancers (n = 43) we observed significant differences for all five texture features on the ADC map (all p-values: <.0001) and for Correlation (p = 0.041) and Inertia (p = 0.001) on T2WI. On ADC maps, GS was associated with higher Entropy (GS 6 vs. 7: p = 0.0225; 6 vs. >7: p = 0.0069) and lower Energy (GS 6 vs. 7: p = 0.0116, 6 vs. >7: p = 0.0039). ADC map Energy (p = 0.0102) and Entropy (p = 0.0019) were significantly different in GS ≤3 + 4 versus ≥4 + 3 cancers; ADC map Entropy remained significant after controlling for the median ADC (p = 0.0291). Several Haralick-based texture features appear useful for prostate cancer detection and GS assessment. • Several Haralick texture features may differentiate non-cancerous and cancerous prostate tissue. • Tumour Energy and Entropy on ADC maps correlate with Gleason score. • T2w-image-derived texture features are not associated with the Gleason score.

Assessing the scale of tumor heterogeneity by complete hierarchical segmentation of MRI

In many cancers, intratumoral heterogeneity has been found in histology, genetic variation and vascular structure. We developed an algorithm to interrogate different scales of heterogeneity using clinical imaging. We hypothesize that heterogeneity of perfusion at coarse scale may correlate with treatment resistance and propensity for disease recurrence. The algorithm recursively segments the tumor image into increasingly smaller regions. Each dividing line is chosen so as to maximize signal intensity difference between the two regions. This process continues until the tumor has been divided into single voxels, resulting in segments at multiple scales. For each scale, heterogeneity is measured by comparing each segmented region to the adjacent region and calculating the difference in signal intensity histograms. Using digital phantom images, we showed that the algorithm is robust to image artifacts and various tumor shapes. We then measured the primary tumor scales of contrast enhancement heterogeneity in MRI of 18 rhabdomyosarcoma patients. Using Cox proportional hazards regression, we explored the influence of heterogeneity parameters on relapse-free survival. Coarser scale of maximum signal intensity heterogeneity was prognostic of shorter survival (p = 0.05). By contrast, two fractal parameters and three Haralick texture features were not prognostic. In summary, our algorithm produces a biologically motivated segmentation of tumor regions and reports the amount of heterogeneity at various distance scales. If validated on a larger dataset, this prognostic imaging biomarker could be useful to identify patients at higher risk for recurrence and candidates for alternative treatment.

Statistical Analysis of Haralick Texture Features to Discriminate Lung Abnormalities.

The Haralick texture features are a well-known mathematical method to detect the lung abnormalities and give the opportunity to the physician to localize the abnormality tissue type, either lung tumor or pulmonary edema. In this paper, statistical evaluation of the different features will represent the reported performance of the proposed method. Thirty-seven patients CT datasets with either lung tumor or pulmonary edema were included in this study. The CT images are first preprocessed for noise reduction and image enhancement, followed by segmentation techniques to segment the lungs, and finally Haralick texture features to detect the type of the abnormality within the lungs. In spite of the presence of low contrast and high noise in images, the proposed algorithms introduce promising results in detecting the abnormality of lungs in most of the patients in comparison with the normal and suggest that some of the features are significantly recommended than others.

3-D liver segmentation from CTA images with patient adaptive Bayesian model

Precise identification of liver region from abdominal Computed Tomography-Angiography (CTA) plays an important role in the evaluation of donor for liver transplantation surgery. Nevertheless, the issues like intensity similarity of liver with neighbouring tissues and inter-intra patient liver shape variability; left the task of liver segmentation challenging. Here, we focus on improving the accuracy and reliability of liver donor evaluation system by customising its crucial step - liver segmentation and volume measurement. For achieving this, a Bayesian classifier is iteratively trained with salient features of liver, namely Haralick texture features and spatial information computed from the individual patient dataset. The proposed method is a combination of two techniques namely, advanced region growing and Bayesian classification. The agreement between the proposed method with the manual segmentation was satisfactory with Relative Volume Difference (RVD), Dice Similarity Coefficient (DSC), False-Positive Ratio (FPR), False-Negative Ratio (FNR) with values 8.98, 94.8 ± 1.5, 3.1 ± 2.8 and 5.67 ± 1.8, respectively.

Abstract LB-160: Collagen I fiber signatures in breast cancer predict lymph node metastasis

Abstract The extracellular matrix (ECM) in breast cancers is altered and provides new biomarkers for cancer diagnosis [1] and prognosis [2,3]. Collagen I (Col1) fibers are the major ECM component in breast tumors, and increased Col1 has been found to facilitate breast tumor formation, invasion, and metastasis [2,3]. We tested the relationship between occurrence of lymph node (LN) and Col1 fibers in primary tumors on a tissue microarray by using second harmonic generation (SHG) imaging and image analyses. A breast tissue microarray that had 50 LN positive (LN+) and 50 LN negative (LN-) status patient's primary breast tumor biopsy samples was used in this study. Z-stacks of SHG tile images were acquired on a Zeiss 710 multiphoton microscope. We analyzed the SHG images from all breast biopsies for geometric features such as fiber volume, inter-fiber distance [4], and both 2D and 3D Haralick features such as energy, entropy, and correlation, among others. K-mean clustering of a case-controlled subset (19 LN+ and 16 LN- cases with grade 2, Stage II, T2) was performed. Principal component analysis (PCA) was performed to reduce data dimensions. Supervised nonlinear support vector machine (SVM) classification of 50 LN+ and 50 LN- samples with cross validation was performed on the 3D Haralick features. The detection errors of LN+ and LN- samples based on individual geometric, Haralick texture features and their combinations, which were obtained from the analysis of the Col 1 fiber patterns of 35 case-controlled samples, were calculated. The features that best differentiate LN+ and LN- samples were 3D entropy and 3D energy in the analysis of 35 case-controlled samples. A receiver operating characteristic (ROC) curve of SVM was generated to classify 3D Haralick features of 50 LN+ and 50 LN- samples, in which a Gaussian kernel with a sigma of 2.9 was the best model and gave a false negative rate of 0.13, a false positive rate of 0.31, and an area under the ROC curve of 0.81. In this larger statistical analysis from 100 breast cancer samples, we performed supervised SVM classification with cross validation to ensure the robustness and obtained a best false positive rate of 0.3. This could be because some LN- cases may have turned out as LN+ cases later on, and no follow-up data were available to us. A larger number of breast cancer samples with patient follow-up information is currently being imaged in our lab to train a more robust statistical SVM model. SHG imaging of Col1 fibers in primary breast tumors followed by image analyses can serve as a predictive biomarker to assess LN status and thereby predict outcome of breast cancer. [1]. A. Bergamaschi et al., J. Pathol. (2008). [2] P. P. Provenzano et al., BMC Med. (2008). [3] M. W. Conklin et al., Am. J. Pathol. (2011). [4] S. Kakkad et al., J. Biomed Optics (2012). This work was supported by NIH P50 CA103175. Citation Format: Jiang Lu, Samata Kakkad, Tiffany Greenwood, Meiyappan Solaiyappan, Zaver M. Bhujwalla, Xingde Li, Kristine Glunde. Collagen I fiber signatures in breast cancer predict lymph node metastasis. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr LB-160. doi:10.1158/1538-7445.AM2014-LB-160

WE‐E‐17A‐06: Assessing the Scale of Tumor Heterogeneity by Complete Hierarchical Segmentation On MRI

Purpose:In many cancers, intratumoral heterogeneity exists in vascular and genetic structure. We developed an algorithm which uses clinical imaging to interrogate different scales of heterogeneity. We hypothesize that heterogeneity of perfusion at large distance scales may correlate with propensity for disease recurrence. We applied the algorithm to initial diagnosis MRI of rhabdomyosarcoma patients to predict recurrence.Methods:The Spatial Heterogeneity Analysis by Recursive Partitioning (SHARP) algorithm recursively segments the tumor image. The tumor is repeatedly subdivided, with each dividing line chosen to maximize signal intensity difference between the two subregions. This process continues to the voxel level, producing segments at multiple scales. Heterogeneity is measured by comparing signal intensity histograms between each segmented region and the adjacent region. We measured the scales of contrast enhancement heterogeneity of the primary tumor in 18 rhabdomyosarcoma patients. Using Cox proportional hazards regression, we explored the influence of heterogeneity parameters on relapse‐free survival (RFS). To compare with existing methods, fractal and Haralick texture features were also calculated.Results:The complete segmentation produced by SHARP allows extraction of diverse features, including the amount of heterogeneity at various distance scales, the area of the tumor with the most heterogeneity at each scale, and for a given point in the tumor, the heterogeneity at different scales. 10/18 rhabdomyosarcoma patients suffered disease recurrence. On contrast‐enhanced MRI, larger scale of maximum signal intensity heterogeneity, relative to tumor diameter, predicted for shorter RFS (p=0.05). Fractal dimension, fractal fit, and three Haralick features did not predict RFS (p=0.09‐0.90).Conclusion:SHARP produces an automatic segmentation of tumor regions and reports the amount of heterogeneity at various distance scales. In rhabdomyosarcoma, RFS was shorter when the primary tumor exhibited larger scale of heterogeneity on contrast‐enhanced MRI. If validated on a larger dataset, this imaging biomarker could be useful to help personalize treatment.

Automated Focal Liver Lesion Staging Classification based on Haralick Texture Features and Multi-SVM

paper proposes automated identification and classification of various stages of focal liver lesions based on the Multi-Support Vector Machine (Multi-SVM). The proposed system can be used to discriminate focal liver diseases such as Cyst, Hemangioma, and Hepatocellular carcinoma along with normal liver. The multi-class scenario is a composition of a series of two-class problems, using one- against-all which is the earliest and one of the most widely used implementations. We formulate the discrimination between cysts, cavernous hemangioma, hepatocellular carcinoma, and normal tissue as a supervised learning problem, and apply Multi-SVM to classify the diseases using Haralick local texture descriptors and histogram based features calculated from Regions Of Interest (ROIs), as input. Selection of ROI significantly impact the classification performances, thus we proposes an automatic ROI selection using Fuzzy c-means initialized by level set technique. For multi-class classification, we adopt the One-Against-All (OAA) method. The proposed Multi-SVM based CAD system using 10-fold cross validation yielded classification accuracy of 96.11% with the individual class accuracy of 97.78%, 95.56%, 93.33% and 97.78% for NOR, Cyst, HEM and HCC cases respectively. The proposed Multi-SVM based system is compared with the K-Nearest Neighbor (KNN) based approaches. Experimental results have demonstrated that the Multi-SVM based system greatly outperforms KNN-based approaches and other methods in the literature. The good performance of the proposed method shows a reliable indicator that can improve the information in the staging of focal liver lesion diseases.

SVM Classification of High Resolution Urban Satellites Images using Composite Kernels and Haralick Features

The classification of remotely sensed images knows a large progress taking in consideration the availability of images with different resolutions as well as the abundance of classification’s algorithms. A number of works have shown promising results by the fusion of spatial and spectral information using Support vector machines (SVM) which are a group of supervised classification algorithms that have been recently used in the remote sensing field. For this purpose we propose a methodology exploiting the properties of Mercer’s kernels to construct a family of composite kernels that easily combine multi-spectral features and Haralick texture features as data source. The proposed approach was tested on common scenes of urban imagery. The three different kernels tested allow a significant improvement of the classification performances and a flexibility to balance between the spatial and spectral information in the classifier. The experimental results indicate an accuracy value of 92.55% which is very promising.

Seismic Profile Texture Feature Extraction Algorithm Merged Directional Derivative into GLCM Method

In view of the texture direction suppression problem of seismic profile, this paper proposes a new texture feature extraction algorithm merged directional derivative into GLCM method. The algorithm introduces the weight factor by utilizing the directional derivative of the seismic profile. Moreover, the Haralick texture features such as ASM, contrast, entropy and IDM are extracted based on the GLCM method. Then, the ultimate feature set is extracted by merging the Haralick texture features and weight factor together. The experimental results show that the proposed algorithm can extract more accurate seismic profile feature set and avoid successfully the texture direction suppression problem of seismic profile. The ability of profile identification and segmentation has a certain improved compared with the GLCM method.

Texture‐based classification of liver fibrosis using MRI

To investigate the ability of texture analysis of MRI images to stage liver fibrosis. Current noninvasive approaches for detecting liver fibrosis have limitations and cannot yet routinely replace biopsy for diagnosing significant fibrosis. Forty-nine patients with a range of liver diseases and biopsy-confirmed fibrosis were enrolled in the study. For texture analysis all patients were scanned with a T2 -weighted, high-resolution, spin echo sequence and Haralick texture features applied. The area under the receiver operating characteristics curve (AUROC) was used to assess the diagnostic performance of the texture analysis. The best mean AUROC achieved for separating mild from severe fibrosis was 0.81. The inclusion of age, liver fat and liver R2 variables into the generalized linear model improved AUROC values for all comparisons, with the F0 versus F1-4 comparison the highest (0.91). Our results suggest that a combination of MRI measures, that include selected texture features from T2 -weighted images, may be a useful tool for excluding fibrosis in patients with liver disease. However, texture analysis of MRI performs only modestly when applied to the classification of patients in the mild and intermediate fibrosis stages.

Parametrization of textural patterns in123I-ioflupane imaging for the automatic detection of Parkinsonism

A novel approach to a computer aided diagnosis system for the Parkinson's disease is proposed. This tool is intended as a supporting tool for physicians, based on fully automated methods that lead to the classification of (123)I-ioflupane SPECT images. (123)I-ioflupane images from three different databases are used to train the system. The images are intensity and spatially normalized, then subimages are extracted and a 3D gray-level co-occurrence matrix is computed over these subimages, allowing the characterization of the texture using Haralick texture features. Finally, different discrimination estimation methods are used to select a feature vector that can be used to train and test the classifier. Using the leave-one-out cross-validation technique over these three databases, the system achieves results up to a 97.4% of accuracy, and 99.1% of sensitivity, with positive likelihood ratios over 27. The system presents a robust feature extraction method that helps physicians in the diagnosis task by providing objective, operator-independent textural information about (123)I-ioflupane images, commonly used in the diagnosis of the Parkinson's disease. Textural features computation has been optimized by using a subimage selection algorithm, and the discrimination estimation methods used here makes the system feature-independent, allowing us to extend it to other databases and diseases.

Predicting pathologic tumor response to chemoradiotherapy with histogram distances characterizing longitudinal changes in 18F‐FDG uptake patterns

A family of fluorine-18 ((18)F)-fluorodeoxyglucose ((18)F-FDG) positron-emission tomography (PET) features based on histogram distances is proposed for predicting pathologic tumor response to neoadjuvant chemoradiotherapy (CRT). These features describe the longitudinal change of FDG uptake distribution within a tumor. Twenty patients with esophageal cancer treated with CRT plus surgery were included in this study. All patients underwent PET/CT scans before (pre-) and after (post-) CRT. The two scans were first rigidly registered, and the original tumor sites were then manually delineated on the pre-PET/CT by an experienced nuclear medicine physician. Two histograms representing the FDG uptake distribution were extracted from the pre- and the registered post-PET images, respectively, both within the delineated tumor. Distances between the two histograms quantify longitudinal changes in FDG uptake distribution resulting from CRT, and thus are potential predictors of tumor response. A total of 19 histogram distances were examined and compared to both traditional PET response measures and Haralick texture features. Receiver operating characteristic analyses and Mann-Whitney U test were performed to assess their predictive ability. Among all tested histogram distances, seven bin-to-bin and seven crossbin distances outperformed traditional PET response measures using maximum standardized uptake value (AUC = 0.70) or total lesion glycolysis (AUC = 0.80). The seven bin-to-bin distances were: L(2) distance (AUC = 0.84), χ(2) distance (AUC = 0.83), intersection distance (AUC = 0.82), cosine distance (AUC = 0.83), squared Euclidean distance (AUC = 0.83), L(1) distance (AUC = 0.82), and Jeffrey distance (AUC = 0.82). The seven crossbin distances were: quadratic-chi distance (AUC = 0.89), earth mover distance (AUC = 0.86), fast earth mover distance (AUC = 0.86), diffusion distance (AUC = 0.88), Kolmogorov-Smirnov distance (AUC = 0.88), quadratic form distance (AUC = 0.87), and match distance (AUC = 0.84). These crossbin histogram distance features showed slightly higher prediction accuracy than texture features on post-PET images. The results suggest that longitudinal patterns in (18)F-FDG uptake characterized using histogram distances provide useful information for predicting the pathologic response of esophageal cancer to CRT.

AN OPTIMIZED FEATURE EXTRACTION TECHNIQUE FOR CONTENT BASED IMAGE RETRIEVAL

Content-based image retrieval (CBIR) is an active research area with the development of multimedia technologies and has become a source of exact and fast retrieval. The aim of CBIR is to search and retrieve images from a large database and find out the best match for the given query. Accuracy and efficiency for high dimensional datasets with enormous number of samples is a challenging arena. In this paper, Content Based Image Retrieval using various features such as color, shape, texture is made and a comparison is made among them. The performance of the retrieval system is evaluated depending upon the features extracted from an image. The performance was evaluated using precision and recall rates. Haralick texture features were analyzed at 0 o, 45 o, 90 o, 180 o using gray level co-occurrence matrix. Color feature extraction was done using color moments. Structured features and multiple feature fusion are two main technologies to ensure the retrieval accuracy in the system. GIST is considered as one of the main structured features. It was experimentally observed that combination of these techniques yielded superior performance than individual features. The results for the most efficient combination of techniques have also been presented and optimized for each class of query.

Potential of texture-based classification in urban landscapes using multispectral aerial photos

Multispectral remote sensing application in thematic urban land-use or land-cover (LULC) classification has gained popularity in the recent past. However, as a result of the complexity of urban landscapes and spectral limitations in commonly used imagery, accurate urban LULC classification has often been impeded by confusion of spectra among multiple urban LULC types. The emergence of multispectral aerial photographs, characterised by high spatial resolution and multispectral information, offers great potential for LULC classification. In this study, we hypothesised that textural information using optimum Haralick textural features inherent in multispectral aerial photographs can be used to generate reliable land-cover maps in heterogeneous urban landscapes. Haralick textural feature optimisation and object-based classification were used to discriminate diverse urban LULC types. Grey-level co-occurrence matrix (GLCM) Entropy, GLCM Mean and GLCM Angular Second Moment texture features were used to discriminate different LULC types while the Jeffreys–Matisuta separability analysis was used to identify optimum thresholds for the development of object-based classification rules. Results from object-based classification were also compared to classification output using the aerial photograph’s spectral information. Results show that use of both object-based Haralick textural features and the spectral characteristics on multispectral aerial photographs can be used to generate reliable LULC classes. Classification based on object-based Haralick textural features produced higher accuracy than that based on spectral information. Multispectral aerial photographs using both object-based Haralick textural features and spectral information offer great potential in mapping urban landscapes often characterised by heterogeneous cover types.

Detection of Breast Cancer Cells by Using Texture Analysis

In Brief Breast cancer is a serious public health problem worldwide. This study presents a texture analysis approach to detect cancer on mammograms. Fifty-five regions of interest (ROIs) containing abnormal breast tissue and 192 ROIs containing normal breast tissue were extracted from digital mammograms obtained from the Mammographic Image Analysis Society database. Haralick texture features derived from spatial grey-level dependence matrix were calculated for each ROI. The importance of each feature in distinguishing abnormal from normal tissues was determined by stepwise linear discriminant analysis. The methodology obtained 92.7% of accuracy for the detection of abnormal tissue (cancer) on digitalized mammograms. Breast cancer is a serious public health problem worldwide. This study presents a texture analysis approach to detect cancer on mammograms.

SVM Classification of Urban High-Resolution Imagery Using Composite Kernels and Contour Information

The classification of remote sensing images has done great forward taking into account the image’s availability with different resolutions, as well as an abundance of very efficient classification algorithms. A number of works have shown promising results by the fusion of spatial and spectral information using Support Vector Machines (SVM) which are a group of supervised classification algorithms that have been recently used in the remote sensing field, however the addition of contour information to both spectral and spatial information still less explored. For this purpose we propose a methodology exploiting the properties of Mercer’s kernels to construct a family of composite kernels that easily combine multi-spectral features and Haralick texture features as data source. The composite kernel that gives the best results will be used to introduce contour information in the classification process. The proposed approach was tested on common scenes of urban imagery. The three different kernels tested allow a significant improvement of the classification performances and a flexibility to balance between the spatial and spectral information in the classifier. The experimental results indicate a global accuracy value of 93.52%, the addition of contour information, described by the Fourier descriptors, Hough transform and Zernike moments, allows increasing the obtained global accuracy by 1.61% which is very promising.

Classification of Small Lesions in Breast MRI: Evaluating The Role of Dynamically Extracted Texture Features Through Feature Selection.

Dynamic texture quantification, i.e., extracting texture features from the lesion enhancement pattern in all available post-contrast images, has not been evaluated in terms of its ability to classify small lesions. This study investigates the classification performance achieved with texture features extracted from all five post-contrast images of lesions (mean lesion diameter of 1.1 cm) annotated in dynamic breast magnetic resonance imaging exams. Sixty lesions are characterized dynamically using Haralick texture features. The texture features are then used in a classification task with support vector regression and a fuzzy k-nearest neighbor classifier; free parameters of these classifiers are optimized using random sub-sampling cross-validation. Classifier performance is determined through receiver-operator characteristic (ROC) analysis, specifically through computation of the area under the ROC curve (AUC). Mutual information is used to evaluate the contribution of texture features extracted from different post-contrast stages to classifier performance. Significant improvements (p < 0.05) are observed for six of the thirteen texture features when the lesion enhancement pattern is quantified using the proposed approach of dynamic texture quantification. The highest AUC value observed (0.82) is achieved with texture features responsible for capturing aspects of lesion heterogeneity. Mutual information analysis reveals that texture features extracted from the third and fourth post-contrast images contributed most to the observed improvement in classifier performance. These results show that the performance of automated character classification with small lesions can be significantly improved through dynamic texture quantification of the lesion enhancement pattern.

Classification of small lesions in dynamic breast MRI: eliminating the need for precise lesion segmentation through spatio-temporal analysis of contrast enhancement

Characterizing the dignity of breast lesions as benign or malignant is specifically difficult for small lesions; they don't exhibit typical characteristics of malignancy and are harder to segment since margins are harder to visualize. Previous attempts at using dynamic or morphologic criteria to classify small lesions (mean lesion diameter of about 1 cm) have not yielded satisfactory results. The goal of this work was to improve the classification performance in such small diagnostically challenging lesions while concurrently eliminating the need for precise lesion segmentation. To this end, we introduce a method for topological characterization of lesion enhancement patterns over time. Three Minkowski Functionals were extracted from all five post-contrast images of sixty annotated lesions on dynamic breast MRI exams. For each Minkowski Functional, topological features extracted from each post-contrast image of the lesions were combined into a high-dimensional texture feature vector. These feature vectors were classified in a machine learning task with support vector regression. For comparison, conventional Haralick texture features derived from gray-level co-occurrence matrices (GLCM) were also used. A new method for extracting thresholded GLCM features was also introduced and investigated here. The best classification performance was observed with Minkowski Functionals area and perimeter, thresholded GLCM features f8 and f9, and conventional GLCM features f4 and f6. However, both Minkowski Functionals and thresholded GLCM achieved such results without lesion segmentation while the performance of GLCM features significantly deteriorated when lesions were not segmented (p < 0.05). This suggests that such advanced spatio-temporal characterization can improve the classification performance achieved in such small lesions, while simultaneously eliminating the need for precise segmentation.

Automated morphological classification of lung cancer subtypes using H&amp;E tissue images

Patient-targeted therapies have recently been highlighted as important. An important development in the treatment of metastatic non-small cell lung cancer (NSCLC) has been the tailoring of therapy on the basis of histology. A pathology diagnosis of "non-specified NSCLC" is no longer routinely acceptable; an effective approach for classification of adenocarcinoma (AC) and squamous carcinoma (SC) histotypes is needed for optimizing therapy. In this study, we present a robust and objective automatic computer vision system for real-time classification of AC and SC based on the morphological tissue patterns of hematoxylin and eosin (H&E) staining images to assist medical experts in the diagnosis of lung cancer. Various original and extended densitometric and Haralick's texture features are used to extract image features, and a boosting algorithm is utilized to train the classifier, together with alternative decision tree as the base learner. For evaluation, two types of data with 653 tissue samples were tested, including 369 samples from tissue microarray data set and 284 samples from full-face tissue sections. Regarding the data distribution, 45 % are AC samples (288) and 55 % are SC samples (365), which is considerably well balanced for each class. Using tenfold cross-validation, the technique achieved high accuracy of $$92.41~\%$$ on tissue microarray cores and $$95.42~\%$$ on full tissue sections. We also found that the two boosting algorithms (cw-Boost and AdaBoost.M1) perform consistently well in comparison with other popularly adopted machine learning methods, including support vector machine, neural network and decision tree. This approach offers a robust, objective and rapid procedure for optimized patient-targeted therapy.