A general and simple automated impervious surface mapping approach based on three-dimensional texture features (3DTF) using fine spatial resolution remotely sensed imagery

The mapping of impervious surfaces using remote sensing techniques offer essential technical support for sustainable development objectives and safeguard the environment. In this study, we developed an automated method without training samples for mapping impervious surfaces using texture features. The different aggregated impervious surface patterns and distributions in study areas of Site A-C in China (Beijing, Huainan, Jinhua) were considered. The Site D-E in Dubai and Tehran, surrounded with deserts in arid areas. They were selected to develop and evaluate the performance of the proposed automated method. The texture features of the Contrast, Gabor wavelets, and secondary texture extraction (Con_Gabor) derived from Sentinel-2 images at each site were used to construct the three-dimensional texture features (3DTF) of impervious surfaces. The 3DTF-combined K-means classifier was used to automatically map the impervious surfaces. The results showed that the overall accuracies of mapping impervious surface were 91.15 %, 89.75 %, and 91.90 % in Site A-C. The overall accuracies of mapping impervious surface were 90.95 %, 91.45 % and 88.23 % in rural areas. The distributions of impervious surface on automated method, GHS-BUILT-S and ESA WorldCover were similar in study areas. The automated method for mapping impervious surfaces performed as well as the artificial neural network (ANN) and Random Forest (RF), and the advantage of not requiring training samples. The automated method was tested in the in Dubai and Tehran. The overall accuracies of the automatic method for mapping impervious surfaces >89 % at Site D-E, and >88 % at rural area. In addition, the 3DTF was proved as the simplest and most effective feature combination to map impervious surfaces. The impervious surface mapped using the automated method was robust across bands, seasons and sensors. However, further evaluation is necessary to assess the effectiveness of automated methods using high spatial resolution images for mapping impervious surface in complex areas.

Multiclassification of Hepatic Cystic Echinococcosis by Using Multiple Kernel Learning Framework and Ultrasound Images

To properly treat and care for hepatic cystic echinococcosis (HCE), it is essential to make an accurate diagnosis before treatment.

OBJECTIVE

The objective of this study was to assess the diagnostic accuracy of computer-aided diagnosis techniques in classifying HCE ultrasound images into five subtypes.

METHODS

A total of 1820 HCE ultrasound images collected from 967 patients were included in the study. A multi-kernel learning method was developed to learn the texture and depth features of the ultrasound images. Combined kernel functions were built-in Support Vector Machine (MK-SVM) for the classification work. The experimental results were evaluated using five-fold cross-validation. Finally, our approach was compared with three other machine learning algorithms: the decision tree classifier, random forest, and gradient boosting decision tree.

RESULTS

Among all the methods used in the study, the MK-SVM achieved the highest accuracy of 96.6% on the fused feature set.

CONCLUSION

The multi-kernel learning method effectively learns different image features from ultrasound images by utilizing various kernels. The MK-SVM method, which combines the learning of texture features and depth features separately, has significant application value in HCE classification tasks.

A novel approach to texture recognition combining deep learning orthogonal convolution with regional input features

Textures provide a powerful segmentation and object detection cue. Recent research has shown that deep convolutional nets like Visual Geometry Group (VGG) and ResNet perform well in non-stationary texture datasets. Non-stationary textures have local structures that change from one region of the image to the other. This is consistent with the view that deep convolutional networks are good at detecting local microstructures disguised as textures. However, stationary textures are textures that have statistical properties that are constant or slow varying over the entire region are not well detected by deep convolutional networks. This research demonstrates that simple seven-layer convolutional networks can obtain better results than deep networks using a novel convolutional technique called orthogonal convolution with pre-calculated regional features using grey level co-occurrence matrix. We obtained an average of 8.5% improvement in accuracy in texture recognition on the Outex dataset over GoogleNet, ResNet, VGG and AlexNet.

A machine learning-based approach to identify peripheral artery disease using texture features from contrast-enhanced magnetic resonance imaging

Diagnosing and assessing the risk of peripheral artery disease (PAD) has long been a focal point for medical practitioners. The impaired blood circulation in PAD patients results in altered microvascular perfusion patterns in the calf muscles which is the primary location of intermittent claudication pain. Consequently, we hypothesized that changes in perfusion and increase in connective tissue could lead to alterations in the appearance or texture patterns of the skeletal calf muscles, as visualized with non-invasive imaging techniques. We designed an automatic pipeline for textural feature extraction from contrast-enhanced magnetic resonance imaging (CE-MRI) scans and used the texture features to train machine learning models to detect the heterogeneity in the muscle pattern among PAD patients and matched controls. CE-MRIs from 36 PAD patients and 20 matched controls were used for preparing training and testing data at a 7:3 ratio with cross-validation (CV) techniques. We employed feature arrangement and selection methods to optimize the number of features. The proposed method achieved a peak accuracy of 94.11% and a mean testing accuracy of 84.85% in a 2-class classification approach (controls vs. PAD). A three-class classification approach was performed to identify a high-risk PAD sub-group which yielded an average test accuracy of 83.23% (matched controls vs. PAD without diabetes vs. PAD with diabetes). Similarly, we obtained 78.60% average accuracy among matched controls, PAD treadmill exercise completers, and PAD exercise treadmill non-completers. Machine learning and imaging-based texture features may be of interest in the study of lower extremity ischemia.

Identification of high-risk population of pneumoconiosis using deep learning segmentation of lung 3D images and radiomics texture analysis

OBJECTION

The aim of this study is to develop an early-warning model for identifying high-risk populations of pneumoconiosis by combining lung 3D images and radiomics lung texture features.

METHODS

A retrospective study was conducted, including 600 dust-exposed workers and 300 confirmed pneumoconiosis patients. Chest computed tomography (CT) images were divided into a training set and a test set in a 2:1 ratio. Whole-lung segmentation was performed using deep learning models for feature extraction of radiomics. Two feature selection algorithms and five classification models were used. The optimal model was selected using a 10-fold cross-validation strategy, and the calibration curve and decision curve were evaluated. To verify the applicability of the model, the diagnostic efficiency and accuracy between the model and human interpretation were compared. Additionally, the risk probabilities for different risk groups defined by the model were compared at different time intervals.

RESULTS

Four radiomics features were ultimately used to construct the predictive model. The logistic regression model was the most stable in both the training set and testing set, with an area under curve (AUC) of 0.964 (95 % confidence interval [CI], 0.950-0.976) and 0.947 (95 %CI, 0.925-0.964). In the training and testing sets, the Brier scores were 0.092 and 0.14, respectively, with threshold probability ranges of 2 %-99 % and 2 %-85 %. These results indicate that the model exhibits good calibration and clinical benefit. The comparison between the model and human interpretation showed that the model was not inferior in terms of diagnostic efficiency and accuracy. Additionally, the high-risk population identified by the model was diagnosed as pneumoconiosis two years later.

CONCLUSION

This study provides a meticulous and quantifiable method for detecting and assessing the risk of pneumoconiosis, building upon accurate diagnosis. Employing risk scoring and probability estimation, not only enhances the efficiency of diagnostic physicians but also provides a valuable reference for controlling the occurrence of pneumoconiosis.

Myofascial Trigger Point Identification in B-Mode Ultrasound: Texture Analysis Versus a Convolutional Neural Network Approach

OBJECTIVE

Myofascial pain syndrome (MPS) is one of the most common causes of chronic pain and affects a large portion of patients seen in specialty pain centers as well as primary care clinics. Diagnosis of MPS relies heavily on a clinician's ability to identify the presence of a myofascial trigger point (MTrP). Ultrasound can help, but requires the user to be experienced in ultrasound. Thus, this study investigates the use of texture features and deep learning strategies for the automatic identification of muscle with MTrPs (i.e., active and latent MTrPs) from normal (i.e., no MTrP) muscle.

METHODS

Participants (n = 201) were recruited from Toronto Rehabilitation Institute, and ultrasound videos of their trapezius muscles were acquired. This new data set consists of 1344 images (248 active, 120 latent, 976 normal) collected from these videos. For texture analysis, several features were investigated with varying parameters (i.e., region of interest size, feature type and pixel pair relationships). Convolutional neural networks (CNN) were also applied to observe the performance of deep learning approaches. Performance was evaluated based on the classification accuracy, micro F1-score, sensitivity, specificity, positive predictive value and negative predictive value.

RESULTS

The best CNN approach was able to differentiate between muscles with and without MTrPs better than the best texture feature approach, with F1-scores of 0.7299 and 0.7135, respectively.

CONCLUSION

The results of this study reveal the challenges associated with MTrP identification and the potential and shortcomings of CNN and radiomics approaches in detail.

Classification of stages in cervical cancer MRI by customized CNN and transfer learning

Cervical cancer is the common cancer among women, where early-stage diagnoses of cervical cancer lead to recovery from the deadly cervical cancer. Correct cervical cancer staging is predominant to decide the treatment. Hence, cervical cancer staging is an important problem in designing automatic detection and diagnosing applications of the medical field. Convolutional Neural Networks (CNNs) often plays a greater role in object identification and classification. The performance of CNN in medical image classification can already compete with radiologists. In this paper, we planned to build a deep Capsule Network (CapsNet) for medical image classification that can achieve high accuracy using cervical cancer Magnetic Resonance (MR) images. In this study, a customized deep CNN model is developed using CapsNet to automatically predict the cervical cancer from MR images. In CapsNet, each layer receives input from all preceding layers, which helps to classify the features. The hyper parameters are estimated and it controls the backpropagation gradient at the initial learning. To improve the CapsNet performance, residual blocks are included between dense layers. Training and testing are performed with around 12,771 T2-weighted MR images of the TCGA-CESC dataset publicly available for research work. The results show that the accuracy of Customized CNN using CapsNetis higher and behaves well in classifying the cervical cancer. Thus, it is evident that CNN models can be used in developing automatic image analysis tools in the medical field.

Flavor characteristics of peeled walnut kernels under two-steps roasting processes

Currently, the effects of roasting methods on the flavor profile of peeled walnut kernels (PWKs) remain unknown. The effects of hot air binding (HAHA), radio frequency (HARF), and microwave irradiation (HAMW) on PWK were evaluated using olfactory, sensory, and textural techniques. Solvent Assisted Flavor Evaporation-Gas Chromatography-Olfactometry (SAFE-GC-O) identified 21 odor-active compounds with total concentrations of 229 μg/kg, 273 μg/kg and 499 μg/kg due to HAHA, HARF, and HAMW, respectively. HAMW exhibited the most prominent nutty taste, with the highest response among roasted milky sensors with the typical aroma of 2-ethyl-5-methylpyrazine. HARF had the highest values for chewiness (5.83 N·mm) and brittleness (0.68 mm); however, these attributes did not contribute to the flavor profile. The partial least squares regression (PLSR) model and VIP values showed 13 odor-active compounds were responsible for the sensory differences from different processes. The two-step treatment with HAMW improved the flavor quality of PWK.

Multiclass tumor identification using combined texture and statistical features

Early detection and diagnosis of brain tumors are essential for early intervention and eventually successful treatment plans leading to either a full recovery or an increase in the patient lifespan. However, diagnosis of brain tumors is not an easy task since it requires highly skilled professionals, making this procedure both costly and time-consuming. The diagnosis process relying on MR images gets even harder in the presence of similar objects in terms of their density, size, and shape. No matter how skilled professionals are, their task is still prone to human error. The main aim of this work is to propose a system that can automatically classify and diagnose glioma brain tumors into one of the four tumor types: (1) necrosis, (2) edema, (3) enhancing, and (4) non-enhancing. In this paper, we propose a combined texture discrete wavelet transform (DWT) and statistical features based on the first- and second-order features for the accurate classification and diagnosis of multiclass glioma tumors. Four well-known classifiers, namely, support vector machines (SVM), random forest (RF), multilayer perceptron (MLP), and naïve Bayes (NB), are used for classification. The BraTS 2018 dataset is used for the experiments, and with the combined DWT and statistical features, the RF classifier achieved the highest average accuracy whether for separated modalities or combined modalities. The highest average accuracy of 89.59% and 90.28% for HGG and LGG, respectively, was reported in this paper. It has also been observed that the proposed method outperforms similar existing methods reported in the extant literature.

Whole Tumor Radiomics Analysis for Risk Factors Associated With Rapid Growth of Vestibular Schwannoma in Contrast-Enhanced T1-Weighted Images

OBJECTIVE

To investigate the features associated with rapid growth of vestibular schwannoma using radiomics analysis on magnetic resonance imaging (MRI) together with clinical factors.

METHODS

From August 2005 to February 2019, 67 patients with vestibular schwannoma underwent contrast-enhanced T1-weighted MRI at least twice as part of their diagnosis. After excluding 3 cases with an extremely short follow-up period of 15 days or less, 64 patients were finally enrolled in this study. Ninety-three texture features were extracted from the tumor image data using 3D Slicer software (http://www.slicer.org/). We determined the texture features that significantly affected maximal tumor diameter growth of more than 2 mm/year using Random Forest and Bounty. We also analyzed age and tumor size as clinical factors. We calculated the areas under the curve (AUCs) using receiver operating characteristic analysis for prediction models using texture, clinical, and mixed factors by Random Forest and 5-fold cross-validation.

RESULTS

Two texture features, low minimum signal and high inverse difference moment normalized (Idmn), were significantly associated with rapid growth of vestibular schwannoma. The mixed model of texture features and clinical factors offered the highest AUC (0.69), followed by the pure texture (0.67), and pure clinical (0.63) models. The minimum signal was the most important variable followed by tumor size, Idmn, and age.

CONCLUSIONS

Our radiomics analysis found that texture features were significantly associated with the rapid growth of vestibular schwannoma in contrast-enhanced T1-weighted images. The mixed model offered a higher diagnostic performance than the pure texture or clinical models.

Study on Identification Method of Pulmonary Nodules: Improved Random Walk Pulmonary Parenchyma Segmentation and Fusion Multi-Feature VGG16 Nodule Classification

Purpose

The purpose of this study was to realize automatic segmentation of lung parenchyma based on random walk algorithm to ensure the accuracy of lung parenchyma segmentation. The explicable features of pulmonary nodules were added into VGG16 neural network to improve the classification accuracy of pulmonary nodules.

Materials and Methods

LIDC-IDRI, a public dataset containing lung Computed Tomography images/pulmonary nodules, was used as experimental data. In lung parenchyma segmentation, the maximum Between-Class Variance method (OTSU), corrosion and expansion methods were used to automatically obtain the foreground and background seed points of random walk algorithm in lung parenchyma region. The shortest distance between point sets was added as one of the criteria of prospect probability in the calculation of random walk weight function to achieve accurate segmentation of pulmonary parenchyma. According to the location of the nodules marked by the doctor, the nodules were extracted. The texture features and grayscale features were extracted by Volume Local Direction Ternary Pattern (VLDTP) method and gray histogram. The explicable features were input into VGG16 network in series mode and fused with depth features to achieve accurate classification of nodules. Intersection of Union (IOU) and false positive rate (FPR) were used to measure the segmentation results. Accuracy, Sensitivity, Specificity, Accuracy and F1 score were used to evaluate the results of nodule classification.

Results

The automatic random walk algorithm is effective in lung parenchyma segmentation, and its segmentation efficiency is improved obviously. In VGG16 network, the accuracy of nodular classification is 0.045 higher than that of single depth feature classification.

Conclusion

The method proposed in this paper can effectively and accurately achieve automatic segmentation of lung parenchyma. In addition, the fusion of multi-feature VGG16 network is effective in the classification of pulmonary nodules, which can improve the accuracy of nodular classification.

Machine learning principles applied to CT radiomics to predict mucinous pancreatic cysts

PURPOSE

Current diagnostic and treatment modalities for pancreatic cysts (PCs) are invasive and are associated with patient morbidity. The purpose of this study is to develop and evaluate machine learning algorithms to delineate mucinous from non-mucinous PCs using non-invasive CT-based radiomics.

METHODS

A retrospective, single-institution analysis of patients with non-pseudocystic PCs, contrast-enhanced computed tomography scans within 1 year of resection, and available surgical pathology were included. A quantitative imaging software platform was used to extract radiomics. An extreme gradient boosting (XGBoost) machine learning algorithm was used to create mucinous classifiers using texture features only, or radiomic/radiologic and clinical combined models. Classifiers were compared using performance scoring metrics. Shapely additive explanation (SHAP) analyses were conducted to identify variables most important in model construction.

RESULTS

Overall, 99 patients and 103 PCs were included in the analyses. Eighty (78%) patients had mucinous PCs on surgical pathology. Using multiple fivefold cross validations, the texture features only and combined XGBoost mucinous classifiers demonstrated an area under the curve of 0.72 ± 0.14 and 0.73 ± 0.14, respectively. By SHAP analysis, root mean square, mean attenuation, and kurtosis were the most predictive features in the texture features only model. Root mean square, cyst location, and mean attenuation were the most predictive features in the combined model.

CONCLUSION

Machine learning principles can be applied to PC texture features to create a mucinous phenotype classifier. Model performance did not improve with the combined model. However, specific radiomic, radiologic, and clinical features most predictive in our models can be identified using SHAP analysis.

Role of correlated noise in textural features extraction

Predictive models of tumor response based on heterogeneity metrics in medical images, such as textural features, are highly suggestive. However, the demonstrated sensitivity of these features to noise does affect the model being developed. An in-depth analysis of the noise influence on the extraction of texture features was performed based on the assumption that an improvement in information quality can also enhance the predictive model. A heuristic approach was used that recognizes from the beginning that the noise has its own texture and it was analysed how it affects the quantitative signal data. A simple procedure to obtain noise image estimation is shown; one which makes it possible to extract the noise-texture features at each observation. The distance measured between the textural features in signal and estimated noise images allows us to determine the features affected in each observation by the noise and, for example, to exclude some of them from the model. A demonstration was carried out using synthetic images applying realistic noise models found in medical images. Drawn conclusions were applied to a public cohort of clinical images obtained using FDG-PET to show how the predictive model could be improved. A gain in the area under the receiver operating characteristic curve between 10 and 20% when noise texture information is used was shown. An improvement between 20 and 30% can be appreciated in the estimated model quality.

A neural network for glomerulus classification based on histological images of kidney biopsy

Background

Computer-aided diagnosis (CAD) systems based on medical images could support physicians in the decision-making process. During the last decades, researchers have proposed CAD systems in several medical domains achieving promising results.

CAD systems play an important role in digital pathology supporting pathologists in analyzing biopsy slides by means of standardized and objective workflows. In the proposed work, we designed and tested a novel CAD system module based on image processing techniques and machine learning, whose objective was to classify the condition affecting renal corpuscles (glomeruli) between sclerotic and non-sclerotic. Such discrimination is useful for the biopsy slides evaluation performed by pathologists.

Results

We collected 26 digital slides taken from the kidneys of 19 donors with Periodic Acid-Schiff staining. Expert pathologists have conducted the slides preparation, digital acquisition and glomeruli annotations. Before setting the classifiers, we evaluated several feature extraction techniques from the annotated regions. Then, a feature reduction procedure followed by a shallow artificial neural network allowed discriminating between the glomeruli classes.

We evaluated the workflow considering an independent dataset (i.e., processing images not used in the training procedure). Ten independent runs of the training algorithm, and evaluation, allowed achieving MCC and Accuracy of 0.95 (± 0.01) and 0.99 (standard deviation < 0.00), respectively. We also obtained good precision (0.9844 ± 0.0111) and recall (0.9310 ± 0.0153).

Conclusions

Results on the test set confirm that the proposed workflow is consistent and reliable for the investigated domain, and it can support the clinical practice of discriminating the two classes of glomeruli. Analyses on misclassifications show that the involved images are usually affected by staining artefacts or present partial sections due to slice preparation and staining processes. In clinical practice, however, pathologists discard images showing such artefacts.

Exploring the diagnostic effectiveness for myocardial ischaemia based on CCTA myocardial texture features

BACKGROUND

To explore the characteristics of myocardial textures on coronary computed tomography angiography (CCTA) images in patients with coronary atherosclerotic heart disease, a classification model was established, and the diagnostic effectiveness of CCTA for myocardial ischaemia patients was explored.

METHODS

This was a retrospective analysis of the CCTA images of 155 patients with clinically diagnosed coronary heart disease from September 2019 to January 2020, 79 of whom were considered positive (myocardial ischaemia) and 76 negative (normal myocardial blood supply) according to their clinical diagnoses. By using the deep learning model-based CQK software, the myocardium was automatically segmented from the CCTA images and used to extract texture features. All patients were randomly divided into a training cohort and a test cohort at a 7:3 ratio. The Spearman correlation and least absolute shrinkage and selection operator (LASSO) method were used for feature selection. Based on the selected features of the training cohort, a multivariable logistic regression model was established. Finally, the test cohort was used to verify the regression model.

RESULTS

A total of 387 features were extracted from the CCTA images of the 155 coronary heart disease patients. After performing dimensionality reduction with the Spearman correlation and LASSO, three texture features were selected. The accuracy, area under the curve, specificity, sensitivity, positive predictive value and negative predictive value of the constructed multivariable logistic regression model with the test cohort were 0.783, 0.875, 0.733, 0.875, 0.650 and 0.769, respectively.

CONCLUSION

CCTA imaging texture features of the myocardium have potential as biomarkers for diagnosing myocardial ischaemia.

Classification of oral salivary gland tumors based on texture features in optical coherence tomography images

Currently, the diagnoses of oral diseases primarily depend on the visual recognition of experienced clinicians. It has been proven that automatic recognition based on images can support clinical decision-making by extracting and analyzing objective hidden information. In recent years, optical coherence tomography (OCT) has become a powerful optical imaging technique with the advantages of high resolution and non-invasion. In our study, a dataset composed of four kinds of oral salivary gland tumors (SGTs) was obtained from a homemade swept-source OCT, including two benign and two malignant tumors. Seventy-six texture features were extracted from OCT images to create computational models of diseases. It was demonstrated that the artificial neural network (ANN) based on principal component analysis (PCA) can obtain high diagnostic sensitivity and specificity (higher than 99%) for these four kinds of tumors. The classification accuracy of each tumor is larger than 99%. In addition, the performances of two classifiers (ANN and support vector machine) were quantitatively evaluated based on SGTs. It was proven that the texture features in OCT images provided objective information to classify oral tumors.

Machine learning classification of texture features of MRI breast tumor and peri-tumor of combined pre- and early treatment predicts pathologic complete response

Purpose

This study used machine learning classification of texture features from MRI of breast tumor and peri-tumor at multiple treatment time points in conjunction with molecular subtypes to predict eventual pathological complete response (PCR) to neoadjuvant chemotherapy.

Materials and method

This study employed a subset of patients (N = 166) with PCR data from the I-SPY-1 TRIAL (2002–2006). This cohort consisted of patients with stage 2 or 3 breast cancer that underwent anthracycline–cyclophosphamide and taxane treatment. Magnetic resonance imaging (MRI) was acquired pre-neoadjuvant chemotherapy, early, and mid-treatment. Texture features were extracted from post-contrast-enhanced MRI, pre- and post-contrast subtraction images, and with morphological dilation to include peri-tumoral tissue. Molecular subtypes and Ki67 were also included in the prediction model. Performance of classification models used the receiver operating characteristics curve analysis including area under the curve (AUC). Statistical analysis was done using unpaired two-tailed t-tests.

Results

Molecular subtypes alone yielded moderate prediction performance of PCR (AUC = 0.82, p = 0.07). Pre-, early, and mid-treatment data alone yielded moderate performance (AUC = 0.88, 0.72, and 0.78, p = 0.03, 0.13, 0.44, respectively). The combined pre- and early treatment data markedly improved performance (AUC = 0.96, p = 0.0003). Addition of molecular subtypes improved performance slightly for individual time points but substantially for the combined pre- and early treatment (AUC = 0.98, p = 0.0003). The optimal morphological dilation was 3–5 pixels. Subtraction of post- and pre-contrast MRI further improved performance (AUC = 0.98, p = 0.00003). Finally, among the machine-learning algorithms evaluated, the RUSBoosted Tree machine-learning method yielded the highest performance.

Conclusion

AI-classification of texture features from MRI of breast tumor at multiple treatment time points accurately predicts eventual PCR. Longitudinal changes in texture features and peri-tumoral features further improve PCR prediction performance. Accurate assessment of treatment efficacy early on could minimize unnecessary toxic chemotherapy and enable mid-treatment modification for patients to achieve better clinical outcomes.

A multiparametric MRI-based CAD system for accurate diagnosis of bladder cancer staging

Appropriate treatment of bladder cancer (BC) is widely based on accurate and early BC staging. In this paper, a multiparametric computer-aided diagnostic (MP-CAD) system is developed to differentiate between BC staging, especially T1 and T2 stages, using T2-weighted (T2W) magnetic resonance imaging (MRI) and diffusion-weighted (DW) MRI. Our framework starts with the segmentation of the bladder wall (BW) and localization of the whole BC volume (V_t) and its extent inside the wall (V_w). Our segmentation framework is based on a fully connected convolution neural network (CNN) and utilized an adaptive shape model followed by estimating a set of functional, texture, and morphological features. The functional features are derived from the cumulative distribution function (CDF) of the apparent diffusion coefficient. Texture features are radiomic features estimated from T2W-MRI, and morphological features are used to describe the tumors' geometric. Due to the significant texture difference between the wall and bladder lumen cells, V_t is parcelled into a set of nested equidistance surfaces (i.e., iso-surfaces). Finally, features are estimated for individual iso-surfaces, which are then augmented and used to train and test machine learning (ML) classifier based on neural networks. The system has been evaluated using 42 data sets, and a leave-one-subject-out approach is employed. The overall accuracy, sensitivity, specificity, and area under the receiver operating characteristics (ROC) curve (AUC) are 95.24%, 95.24%, 95.24%, and 0.9864, respectively. The advantage of fusion multiparametric iso-features is highlighted by comparing the diagnostic accuracy of individual MRI modality, which is confirmed by the ROC analysis. Moreover, the accuracy of our pipeline is compared against other statistical ML classifiers (i.e., random forest (RF) and support vector machine (SVM)). Our CAD system is also compared with other techniques (e.g., end-to-end convolution neural networks (i.e., ResNet50).

Prognostic Value of Computed Tomography and/or 18F-Fluorodeoxyglucose Positron Emission Tomography Radiomics Features in Locally Advanced Non-small Cell Lung Cancer

INTRODUCTION

We investigated whether adding computed tomography (CT) and/or ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) PET radiomics features to conventional prognostic factors (CPFs) improves prognostic value in locally advanced non-small cell lung cancer (NSCLC).

MATERIALS AND METHODS

We retrospectively identified 39 cases with stage III NSCLC who received chemoradiotherapy and underwent planning CT and staging ¹⁸F-FDG PET scans. Seven CPFs were recorded. Feature selection was performed on 48 CT and 49 PET extracted radiomics features. A penalized multivariate Cox proportional hazards model was used to generate models for overall survival based on CPFs alone, CPFs with CT features, CPFs with PET features, and CPFs with CT and PET features. Linear predictors generated and categorized into 2 risk groups for which Kaplan-Meier survival curves were calculated. A log-rank test was performed to quantify the discrimination between the groups and calculated the Harrell's C-index to quantify the discriminatory power. A likelihood ratio test was performed to determine whether adding CT and/or PET features to CPFs improved model performance.

RESULTS

All 4 models significantly discriminated between the 2 risk groups. The discriminatory power was significantly increased when CPFs were combined with PET features (C-index 0.82; likelihood ratio test P < .01) or with both CT and PET features (0.83; P < .01) compared with CPFs alone (0.68). There was no significant improvement when CPFs were combined with CT features (0.68).

CONCLUSION

Adding PET radiomics features to CPFs yielded a significant improvement in the prognostic value in locally advanced NSCLC; adding CT features did not.

A Screening CAD Tool for the Detection of Microcalcification Clusters in Mammograms

Breast cancer is the most common cancer diagnosed in women worldwide. Up to 50% of non-palpable breast cancers are detected solely through microcalcification clusters in mammograms. This article presents a novel and completely automated algorithm for the detection of microcalcification clusters in a mammogram. A multiscale 2D non-linear energy operator is proposed for enhancing the contrast between the microcalcifications and the background. Several texture, shape, intensity, and histogram of oriented gradients (HOG)-based features are used to distinguish microcalcifications from other brighter mammogram regions. A new majority class data reduction technique based on data distribution is proposed to counter data imbalance problem. The algorithm is able to achieve 100% sensitivity with 2.59, 1.78, and 0.68 average false positives per image on Digital Database for Screening Mammography (scanned film), INbreast (direct radiography) database, and PGIMER-IITKGP mammogram (direct radiography) database, respectively. Thus, it might be used as a second reader as well as a screening tool to reduce the burden on radiologists.

Automated detection of diabetic retinopathy in fundus images using fused features

Diabetic retinopathy (DR) is one of the severe eye conditions due to diabetes complication which can lead to vision loss if left untreated. In this paper, a computationally simple, yet very effective, DR detection method is proposed. First, a segmentation independent two-stage preprocessing based technique is proposed which can effectively extract DR pathognomonic signs; both bright and red lesions, and blood vessels from the eye fundus image. Then, the performance of Local Binary Patterns (LBP), Local Ternary Patterns (LTP), Dense Scale-Invariant Feature Transform (DSIFT) and Histogram of Oriented Gradients (HOG) as a feature descriptor for fundus images, is thoroughly analyzed. SVM kernel-based classifiers are trained and tested, using a 5-fold cross-validation scheme, on both newly acquired fundus image database from the local hospital and combined database created from the open-sourced available databases. The classification accuracy of 96.6% with 0.964 sensitivity and 0.969 specificity is achieved using a Cubic SVM classifier with LBP and LTP fused features for the local database. More importantly, in out-of-sample testing on the combined database, the model gives an accuracy of 95.21% with a sensitivity of 0.970 and specificity of 0.932. This indicates the proposed model is very well-fitted and generalized which is further corroborated by the presented train-test curves.

Development of an automated system for the detection of genotype in polypoidal choroidal vasculopathy using retinal image phenotype

BACKGROUND AND OBJECTIVES

Polypoidal choroidal vasculopathy (PCV) is a retinal disorder characterized by the presence of aneurismal polypoidal lesions in the choroidal vasculature. A single nucleotide polymorphism (SNP) is a common genetic variant which may be associated with the disease. This study is to investigate the association of HERPUD1 (rs2217332) gene with PCV in the Indian population and develop an automated system for genotype and phenotype correlation using fundus images and machine learning methods.

METHODS

A cohort of 54 PCV patients and 120 control subjects were recruited for the study. Genotyping of SNP (HERPUD1, rs2217332) was performed by following polymerase chain reaction and direct sequencing method. Statistical association of SNP to PCV was determined using chi-square analysis. The acquired GG and AG images were preprocessed using an adaptive histogram. 19 and 18 texture features were extracted from the images in the PCV naïve cases and PCV patients on treatment, respectively. Student's independent t-test was then employed for the selection of significant features, which were input to the ensemble tree for automated classification. Leave-one-out validation was used to evaluate the system.

RESULTS

HERPUD1 rs2217332 SNP is significantly associated in PCV patients compared to control (P = 0.0296, odds ratio [OD] = 2.297, 95% confidence interval [CI] = 1.087-4.856) in the Indian population. High F1 and precision values of 85.71%, 86.84% and 85.71%, 93.75% were achieved in the pre and post- treatment phases, respectively.

CONCLUSION

Our results suggest that the HERPUD1 polymorphism is associated in PCV patients. Based on our analysis, it may be possible to predict the genotype and disease status of PCV patients using fundus images in assistance with a machine learning algorithm.

Predicting ventricular tachyarrhythmia in patients with systolic heart failure based on texture features of the gray zone from contrast-enhanced magnetic resonance imaging

BACKGROUND

Previous research showed that gray zone detected by late gadolinium enhancement cardiovascular magnetic resonance (LGE-CMR) imaging could help identify high-risk patients. In this study, we investigated whether LGE-CMR gray zone heterogeneity measured by image texture features could predict cardiovascular events in patients with heart failure (HF).

METHOD

This is a retrospective cohort study. Patients with systolic HF undergoing CMR imaging were enrolled. Cine and LGE images were analyzed to derive left ventricular (LV) function and scar characteristics. Entropy and uniformity of gray zones were derived by texture analysis.

RESULTS

A total of 82 systolic HF patients were enrolled. After a median 1021 (25%-75% quartiles, 205-2066) days of follow-up, the entropy (0.60 ± 0.260 vs. 0.87 ± 0.28, p = 0.013) was significantly increased while the uniformity (0.68 ± 0.14 vs. 0.53±0.15, p = 0.016) was significantly decreased in patients with ventricular tachycardia or ventricular fibrillation (VT/VF). The percentage of core scar (21.9 ± 10.6 vs. 30.6 ± 10.4, p = 0.029) was higher in cardiac mortality group than survival group while the uniformity (0.55 ± 0.17 vs. 0.67 ± 0.14, p = 0.018) was lower in cardiac mortality group than survival group. A multivariate Cox regression model showed that higher percentage of gray zone area (HR = 8.805, 1.620-47.84, p = 0.045), higher entropy (>0.85) (HR = 1.391, 1.092-1.772, p = 0.024) and lower uniformity (≦0.54) (HR = 0.535, 0.340-0.842, p = 0.022) were associated with VT/VF attacks. Also, higher percentage of gray zone area (HR = 5.716, 1.379-23.68, p = 0.017), core scar zone (HR = 1.939, 1.056-3.561, p = 0.025), entropy (>0.85) (HR = 1.434, 1.076-1.911, p = 0.008) and lower uniformity (≦0.54) (HR = 0.513, 0.296-0.888, p = 0.009) were associated with cardiac mortality during follow-up.

CONCLUSIONS

Gray zone heterogeneity by texture analysis method could provide additional prognostic value to traditional LGE-CMR substrate analysis method.

Early detection of ankylosing spondylitis using texture features and statistical machine learning, and deep learning, with some patient age analysis

Ankylosing spondylitis (AS) is an arthritis with symptoms visible in medical imagery. This paper proposes, to the authors' best knowledge, the first use of statistical machine learning- and deep learning-based classifiers to detect erosion, an early AS symptom, via analysis of computed tomography (CT) imagery, giving some consideration to patient age in so doing. We used gray-level co-occurrence matrices and local binary patterns to generate input features to machine learning algorithms, specifically k-nearest neighbors (k-NN) and random forest. Deep learning solutions based on a modified InceptionV3 architecture were designed and tested, with one classifier produced by training with a cross-entropy loss function and another produced by additionally seeking to minimize validation loss. We found that the random forest classifiers outperform the k-NN classifiers and achieve an eightfold cross-validation average accuracy, recall, and area under receiver operator characteristic curve (ROC AUC) of 96.0%, 92.9%, and 0.97, respectively, for erosion vs. young control patients, and 82.4%, 80.6%, and 0.91, respectively, for erosion vs. old control patients. We found that the deep learning classifier trained without minimizing validation loss was best and achieves an eightfold cross-validation accuracy, recall, and ROC AUC of 99.0%, 97.5%, and 0.97, respectively, for erosion vs. all (combined young and old) control patients; this classifier outperforms a musculoskeletal radiologist with 9 years of experience in raw sensitivity and specificity by 8.4% and 9.5%, respectively. Despite the relatively small dataset on which we trained and cross-validated, our results indicate the potential of machine and deep learning to aid AS diagnosis, and further research using larger datasets should be conducted.

Does image normalization and intensity resolution impact texture classification?

Image texture is a very important component in many types of images, including medical images. Medical images are often corrupted by noise and affected by artifacts. Some of the texture-based features that should describe the structure of the tissue under examination may also reflect, for example, the uneven sensitivity of the scanner within the tissue region. This in turn may lead to an inappropriate description of the tissue or incorrect classification. To limit these phenomena, the analyzed regions of interest are normalized. In texture analysis methods, image intensity normalization is usually followed by a reduction in the number of levels coding the intensity. The aim of this work was to analyze the impact of different image normalization methods and the number of intensity levels on texture classification, taking into account noise and artifacts related to uneven background brightness distribution. Analyses were performed on four sets of images: modified Brodatz textures, kidney images obtained by means of dynamic contrast-enhanced magnetic resonance imaging, shoulder images acquired as T2-weighted magnetic resonance images and CT heart and thorax images. The results will be of use for choosing a particular method of image normalization, based on the types of noise and distortion present in the images.

Heterogeneity in tumours: Validating the use of radiomic features on 18F-FDG PET/CT scans of lung cancer patients as a prognostic tool

AIM

The aim was to validate promising radiomic features (RFs)1 on 18F-flourodeoxyglucose positron emission tomography/computed tomography-scans (18F-FDG PET/CT) of non-small cell lung cancer (NSCLC) patients undergoing definitive chemo-radiotherapy.

METHODS

18F-FDG PET/CT scans performed for radiotherapy (RT) planning were retrieved. Auto-segmentation with visual adaption was used to define the primary tumour on PET images. Six pre-selected prognostic and reproducible PET texture -and shape-features were calculated using texture respectively shape analysis. The correlation between these RFs and metabolic active tumour volume (MTV)3, gross tumour volume (GTV)4 and maximum and mean of standardized uptake value (SUV)5 was tested with a Spearman's Rank test. The prognostic value of RFs was tested in a univariate cox regression analysis and a multivariate cox regression analysis with GTV, clinical stage and histology. P-value ≤ 0.05 were considered significant.

RESULTS

Image analysis was performed for 233 patients: 145 males and 88 females, mean age of 65.7 and clinical stage II-IV. Mean GTV was 129.87 cm3 (SD 130.30 cm3). Texture and shape-features correlated more strongly to MTV and GTV compared to SUV-measurements. Four RFs predicted PFS in the univariate analysis. No RFs predicted PFS in the multivariate analysis, whereas GTV and clinical stage predicted PFS (p = 0.001 and p = 0.008 respectively).

CONCLUSION

The pre-selected RFs were insignificant in predicting PFS in combination with GTV, clinical stage and histology. These results might be due to variations in technical parameters. However, it is relevant to question whether RFs are stable enough to provide clinically useful information.

Longitudinal brain tumor segmentation prediction in MRI using feature and label fusion

This work proposes a novel framework for brain tumor segmentation prediction in longitudinal multi-modal MRI scans, comprising two methods; feature fusion and joint label fusion (JLF). The first method fuses stochastic multi-resolution texture features with tumor cell density feature to obtain tumor segmentation predictions in follow-up timepoints using data from baseline pre-operative timepoint. The cell density feature is obtained by solving the 3D reaction-diffusion equation for biophysical tumor growth modelling using the Lattice-Boltzmann method. The second method utilizes JLF to combine segmentation labels obtained from (i) the stochastic texture feature-based and Random Forest (RF)-based tumor segmentation method; and (ii) another state-of-the-art tumor growth and segmentation method, known as boosted Glioma Image Segmentation and Registration (GLISTRboost, or GB). We quantitatively evaluate both proposed methods using the Dice Similarity Coefficient (DSC) in longitudinal scans of 9 patients from the public BraTS 2015 multi-institutional dataset. The evaluation results for the feature-based fusion method show improved tumor segmentation prediction for the whole tumor(DSC _WT = 0.314, p = 0.1502), tumor core (DSC _TC = 0.332, p = 0.0002), and enhancing tumor (DSC _ET = 0.448, p = 0.0002) regions. The feature-based fusion shows some improvement on tumor prediction of longitudinal brain tumor tracking, whereas the JLF offers statistically significant improvement on the actual segmentation of WT and ET (DSC _WT = 0.85 ± 0.055, DSC _ET = 0.837 ± 0.074), and also improves the results of GB. The novelty of this work is two-fold: (a) exploit tumor cell density as a feature to predict brain tumor segmentation, using a stochastic multi-resolution RF-based method, and (b) improve the performance of another successful tumor segmentation method, GB, by fusing with the RF-based segmentation labels.

Computational texture features of dermoscopic images and their link to the descriptive terminology: A survey

Computer-extracted texture features are relevant to diagnose cutaneous lesions such as melanomas. Our goal is to set a relationship between a well-established descriptive terminology, which describes the attributes of dermoscopic structures based on their aspect rather than their underlying causes, and the computational methods to extract texture-based features. By tackling this problem, we can ascertain what indicators used by dermatologists are reflected in the extracted texture features. We first review the state-of-the-art models for texture extraction in dermoscopic images. By comparing the methods' performance and goals, we conclude that (I) a single color space does not seem to give performances as good as using several ones, thus the latter is reasonable (II) the optimal number of extracted features seems to vary depending on the method's goal, and extracting a large number of features can lead to a loss of models robustness (III) methods such as GLCM, Sobel or Law energy filters are mainly used to capture local properties to detect specific dermoscopic structures (IV) methods that extract local and global features, like Gabor wavelets or SPT, tend to be used to analyze the presence of certain patterns of dermoscopic structures, e.g. globular, reticular, etc.

Prediction of progression in idiopathic pulmonary fibrosis using CT scans at baseline: A quantum particle swarm optimization - Random forest approach

Idiopathic pulmonary fibrosis (IPF) is a fatal lung disease characterized by an unpredictable progressive decline in lung function. Natural history of IPF is unknown and the prediction of disease progression at the time of diagnosis is notoriously difficult. High resolution computed tomography (HRCT) has been used for the diagnosis of IPF, but not generally for monitoring purpose. The objective of this work is to develop a novel predictive model for the radiological progression pattern at voxel-wise level using only baseline HRCT scans. Mainly, there are two challenges: (a) obtaining a data set of features for region of interest (ROI) on baseline HRCT scans and their follow-up status; and (b) simultaneously selecting important features from high-dimensional space, and optimizing the prediction performance. We resolved the first challenge by implementing a study design and having an expert radiologist contour ROIs at baseline scans, depending on its progression status in follow-up visits. For the second challenge, we integrated the feature selection with prediction by developing an algorithm using a wrapper method that combines quantum particle swarm optimization to select a small number of features with random forest to classify early patterns of progression. We applied our proposed algorithm to analyze anonymized HRCT images from 50 IPF subjects from a multi-center clinical trial. We showed that it yields a parsimonious model with 81.8% sensitivity, 82.2% specificity and an overall accuracy rate of 82.1% at the ROI level. These results are superior to other popular feature selections and classification methods, in that our method produces higher accuracy in prediction of progression and more balanced sensitivity and specificity with a smaller number of selected features. Our work is the first approach to show that it is possible to use only baseline HRCT scans to predict progressive ROIs at 6 months to 1year follow-ups using artificial intelligence.

Global parenchymal texture features based on histograms of oriented gradients improve cancer development risk estimation from healthy breasts

BACKGROUND

The breast dense tissue percentage on digital mammograms is one of the most commonly used markers for breast cancer risk estimation. Geometric features of dense tissue over the breast and the presence of texture structures contained in sliding windows that scan the mammograms may improve the predictive ability when combined with the breast dense tissue percentage.

METHODS

A case/control study nested within a screening program covering 1563 women with craniocaudal and mediolateral-oblique mammograms (755 controls and the contralateral breast mammograms at the closest screening visit before cancer diagnostic for 808 cases) aging 45 to 70 from Comunitat Valenciana (Spain) was used to extract geometric and texture features. The dense tissue segmentation was performed using DMScan and validated by two experienced radiologists. A model based on Random Forests was trained several times varying the set of variables. A training dataset of 1172 patients was evaluated with a 10-stratified-fold cross-validation scheme. The area under the Receiver Operating Characteristic curve (AUC) was the metric for the predictive ability. The results were assessed by only considering the output after applying the model to the test set, which was composed of the remaining 391 patients.

RESULTS

The AUC score obtained by the dense tissue percentage (0.55) was compared to a machine learning-based classifier results. The classifier, apart from the percentage of dense tissue of both views, firstly included global geometric features such as the distance of dense tissue to the pectoral muscle, dense tissue eccentricity or the dense tissue perimeter, obtaining an accuracy of 0.56. By the inclusion of a global feature based on local histograms of oriented gradients, the accuracy of the classifier was significantly improved (0.61). The number of well-classified patients was improved up to 236 when it was 208.

CONCLUSION

Relative geometric features of dense tissue over the breast and histograms of standardized local texture features based on sliding windows scanning the whole breast improve risk prediction beyond the dense tissue percentage adjusted by geometrical variables. Other classifiers could improve the results obtained by the conventional Random Forests used in this study.

Optimal respiratory-gated [18F]FDG PET/CT significantly impacts the quantification of metabolic parameters and their correlation with overall survival in patients with pancreatic ductal adenocarcinoma

PURPOSE

Metabolic parameters are increasingly being used to characterize tumors. Motion artifacts due to patient respiration introduce uncertainties in quantification of metabolic parameters during positron emission tomography (PET) image acquisition. The present study investigates the impact of amplitude-based optimal respiratory gating (ORG) on quantification of PET-derived image features in patients with pancreatic ductal adenocarcinoma (PDAC), in correlation with overall survival (OS).

METHODS

Sixty-nine patients with histologically proven primary PDAC underwent 2'-deoxy-2'-[¹⁸F]fluoroglucose ([¹⁸F]FDG) PET/CT imaging during diagnostic work-up. Standard image acquisition and reconstruction was performed in accordance with the EANM guidelines and ORG images were reconstructed with a duty cycle of 35%. PET-derived image features, including standard parameters, first- and second-order texture features, were calculated from the standard and corresponding ORG images, and correlation with OS was assessed.

RESULTS

ORG significantly impacts the quantification of nearly all features; values of single-voxel parameters (e.g., SUV_max) showed a wider range, volume-based parameters (e.g., SUV_mean) were reduced, and texture features were significantly changed. After correction for motion artifacts using ORG, some features that describe intra-tumoral heterogeneity were more strongly correlated to OS.

CONCLUSIONS

Correction for respiratory motion artifacts using ORG impacts the quantification of metabolic parameters in PDAC lesions. The correlation of metabolic parameters with OS was significantly affected, in particular parameters that describe intra-tumor heterogeneity. Therefore, interpretation of single-voxel or average metabolic parameters in relation to clinical outcome should be done cautiously. Furthermore, ORG is a valuable tool to improve quantification of intra-tumoral heterogeneity in PDAC.

Predicting locally advanced rectal cancer response to neoadjuvant therapy with 18F-FDG PET and MRI radiomics features

PURPOSE

Pathological complete response (pCR) following neoadjuvant chemoradiotherapy or radiotherapy in locally advanced rectal cancer (LARC) is reached in approximately 15-30% of cases, therefore it would be useful to assess if pretreatment of ¹⁸F-FDG PET/CT and/or MRI texture features can reliably predict response to neoadjuvant therapy in LARC.

METHODS

Fifty-two patients were dichotomized as responder (pR+) or non-responder (pR-) according to their pathological tumor regression grade (TRG) as follows: 22 as pR+ (nine with TRG = 1, 13 with TRG = 2) and 30 as pR- (16 with TRG = 3, 13 with TRG = 4 and 1 with TRG = 5). First-order parameters and 21 second-order texture parameters derived from the Gray-Level Co-Occurrence matrix were extracted from semi-automatically segmented tumors on T2w MRI, ADC maps, and PET/CT acquisitions. The role of each texture feature in predicting pR+ was assessed with monoparametric and multiparametric models.

RESULTS

In the mono-parametric approach, PET homogeneity reached the maximum AUC (0.77; sensitivity = 72.7% and specificity = 76.7%), while PET glycolytic volume and ADC dissimilarity reached the highest sensitivity (both 90.9%). In the multiparametric analysis, a logistic regression model containing six second-order texture features (five from PET and one from T2w MRI) yields the highest predictivity in distinguish between pR+ and pR- patients (AUC = 0.86; sensitivity = 86%, and specificity = 83% at the Youden index).

CONCLUSIONS

If preliminary results of this study are confirmed, pretreatment PET and MRI could be useful to personalize patient treatment, e.g., avoiding toxicity of neoadjuvant therapy in patients predicted pR-.

Breast tumor classification using different features of quantitative ultrasound parametric images

RATIONALE AND OBJECTIVES

The ultrasound B-mode-based morphological and texture analysis and Nakagami parametric imaging have been proposed to characterize breast tumors. Since these three feature categories of ultrasonic tissue characterization supply information on different physical characteristics of breast tumors, by combining the above methods is expected to provide more clues for classifying breast tumors.

MATERIALS AND METHODS

To verify the validity of the concept, raw data were obtained from 160 clinical cases. Six different types of morphological-feature parameters, four texture features, and the Nakagami parameter of benignancy and malignancy were extracted for evaluation. The Pearson's correlation matrix was used to calculate the correlation between different feature parameters. The fuzzy c-means clustering and stepwise regression techniques were utilized to determine the optimal feature set, respectively. The logistic regression, receiver operating characteristic curve, and support vector machine were used to estimate the diagnostic ability.

RESULTS

The best performance was obtained by combining morphological-feature parameter (e.g., standard deviation of the shortest distance), texture feature (e.g., variance), and the Nakagami parameter, with an accuracy of 89.4%, a specificity of 86.3%, a sensitivity of 92.5%, and an area under receiver operating characteristic curve of 0.96. There was no significant difference between using fuzzy c-means clustering, logistic regression, and support vector machine based on the optimal feature set for breast tumors classification.

CONCLUSION

Therefore, we verified that different physical ultrasonic features are functionally complementary and thus improve the performance in diagnosing breast tumors. Moreover, the optimal feature set had the maximum discriminating performance should be irrelative to the power of classifiers.

Evaluation of texture features at staging liver fibrosis based on phase contrast X-ray imaging

BACKGROUND

The purpose of this study is to explore the potential of phase contrast imaging to detect fibrotic progress in its early stage; to investigate the feasibility of texture features for quantified diagnosis of liver fibrosis; and to evaluate the performance of back propagation (BP) neural net classifier for characterization and classification of liver fibrosis.

METHODS

Fibrous mouse liver samples were imaged by X-ray phase contrast imaging, nine texture measures based on gray-level co-occurrence matrix were calculated and the feasibility of texture features in the characterization and discrimination of liver fibrosis at early stages was investigated. Furthermore, 36 or 18 features were applied to the input of BP classifier; the classification performance was evaluated using receiver operating characteristic curve.

RESULTS

The phase contrast images displayed a vary degree of texture pattern from normal to severe fibrosis stages. The BP classifier could distinguish liver fibrosis among normal, mild, moderate and severe stages; the average accuracy was 95.1% for 36 features, and 91.1% for 18 features.

CONCLUSION

The study shows that early stages of liver fibrosis can be discriminated by the morphological features on the phase contrast images. BP network model based on combination of texture features is demonstrated effective for staging liver fibrosis.

Automated detection and classification of liver fibrosis stages using contourlet transform and nonlinear features

BACKGROUND AND OBJECTIVE

Liver fibrosis is a type of chronic liver injury that is characterized by an excessive deposition of extracellular matrix protein. Early detection of liver fibrosis may prevent further growth toward liver cirrhosis and hepatocellular carcinoma. In the past, the only method to assess liver fibrosis was through biopsy, but this examination is invasive, expensive, prone to sampling errors, and may cause complications such as bleeding. Ultrasound-based elastography is a promising tool to measure tissue elasticity in real time; however, this technology requires an upgrade of the ultrasound system and software. In this study, a novel computer-aided diagnosis tool is proposed to automatically detect and classify the various stages of liver fibrosis based upon conventional B-mode ultrasound images.

METHODS

The proposed method uses a 2D contourlet transform and a set of texture features that are efficiently extracted from the transformed image. Then, the combination of a kernel discriminant analysis (KDA)-based feature reduction technique and analysis of variance (ANOVA)-based feature ranking technique was used, and the images were then classified into various stages of liver fibrosis.

RESULTS

Our 2D contourlet transform and texture feature analysis approach achieved a 91.46% accuracy using only four features input to the probabilistic neural network classifier, to classify the five stages of liver fibrosis. It also achieved a 92.16% sensitivity and 88.92% specificity for the same model. The evaluation was done on a database of 762 ultrasound images belonging to five different stages of liver fibrosis.

CONCLUSIONS

The findings suggest that the proposed method can be useful to automatically detect and classify liver fibrosis, which would greatly assist clinicians in making an accurate diagnosis.

Computer-aided classification of prostate cancer grade groups from MRI images using texture features and stacked sparse autoencoder

A novel method to determine the Grade Group (GG) in prostate cancer (PCa) using multi-parametric magnetic resonance imaging (mpMRI) biomarkers is investigated in this paper. In this method, high-level features are extracted from hand-crafted texture features using a deep network of stacked sparse autoencoders (SSAE) and classified them using a softmax classifier (SMC). Transaxial T2 Weighted (T2W), Apparent Diffusion Coefficient (ADC) and high B-Value Diffusion-Weighted (BVAL) images obtained from PROSTATEx-2 2017 challenge dataset are used in this technique. The method was evaluated on the challenge dataset composed of a training set of 112 lesions and a test set of 70 lesions. It achieved a quadratic-weighted Kappa score of 0.2772 on evaluation using test dataset of the challenge. It also reached a Positive Predictive Value (PPV) of 80% in predicting PCa with GG > 1. The method achieved first place in the challenge, winning over 43 methods submitted by 21 groups. A 3-fold cross-validation using training data of the challenge was further performed and the method achieved a quadratic-weighted kappa score of 0.2326 and Positive Predictive Value (PPV) of 80.26% in predicting PCa with GG > 1. Even though the training dataset is a highly imbalanced one, the method was able to achieve a fair kappa score. Being one of the pioneer methods which attempted to classify prostate cancer into 5 grade groups from MRI images, it could serve as a base method for further investigations and improvements.

LGE-CMR-derived texture features reflect poor prognosis in hypertrophic cardiomyopathy patients with systolic dysfunction: preliminary results

OBJECTIVES

To evaluate the prognostic value of texture features based on late gadolinium enhancement cardiac magnetic resonance (LGE-CMR) images in hypertrophic cardiomyopathy (HCM) patients with systolic dysfunction.

METHODS

67 HCM patients with systolic dysfunction (41 male and 26 female, mean age ± standard deviation, 46.20 years ± 13.38) were enrolled. All patients underwent 1.5 T CMR cine and LGE imaging. Texture features were extracted from LGE images. Cox proportional hazard analysis and Kaplan-Meier analysis were used to determine the association of texture features and traditional parameters with event free survival.

RESULTS

Family history (hazard ratio [HR]=2.558, 95 % confidence interval [CI]=1.060-6.180), NYHA III-IV (HR=5.627, CI=1.652-19.173), left ventricular ejection fraction (HR=0.945, CI=0.902-0.991), left ventricular end-diastolic volume index (HR=1.006, CI=1.000-1.012), LGE extent (HR=1.911, CI=1.348-2.709) and three texture parameters [X0_H_skewness (HR=0.783, CI=0.691-0.889), X0_GLCM_cluster_tendency (HR=0.735, CI=0.616-0.877) and X0_GLRLM_energy (HR=1.344, CI=1.173-1.540)] were significantly associated with event free survival in univariate analysis (p<0.05). The HR of LGE extent (HR=1.548 [CI=1.046-2.293], 1.650 [CI=1.122-2.428] and 1.586 [CI=1.044-2.409] per 10 % increase, p<0.05) remained significant when adjusted by one of the three texture features.

CONCLUSION

Increased LGE heterogeneity (higher X0_GLRLM_energy, lower X0_H_skewness and lower X0_GLCM_cluster_tendency) was associated with adverse events in HCM patients with systolic dysfunction.

KEY POINTS

• Textural analysis from CMR can be applied in HCM. • Texture features derived from LGE images can capture fibrosis heterogeneity. • CMR texture analysis provides prognostic information in HCM patients.

A new near-term breast cancer risk prediction scheme based on the quantitative analysis of ipsilateral view mammograms

PURPOSE

To help improve efficacy of screening mammography and eventually establish an optimal personalized screening paradigm, this study aimed to develop and test a new near-term breast cancer risk prediction scheme based on the quantitative analysis of ipsilateral view of the negative screening mammograms.

METHODS

The dataset includes digital mammograms acquired from 392 women with two sequential full-field digital mammography examinations. All the first ("prior") sets of mammograms were interpreted as negative during the original reading. In the sequential ("current") screening, 202 were proved positive and 190 remained negative/benign. For each pair of the "prior" ipsilateral mammograms, we adaptively fused the image features computed from two views. Using four different types of image features, we built four elastic net support vector machine (EnSVM) based classifiers. Then, the initial prediction scores form the 4 EnSVMs were combined to build a final artificial neural network (ANN) classifier that produces the final risk prediction score. The performance of the new scheme was evaluated by using a 10-fold cross-validation method and an assessment index of the area under the receiver operating characteristic curve (AUC).

RESULTS

A total number of 466 features were initially extracted from each pair of ipsilateral mammograms. Among them, 51 were selected to build the EnSVM based prediction scheme. The AUC = 0.737 ± 0.052 was yielded using the new scheme. Applying an optimal operating threshold, the prediction sensitivity was 60.4% (122 of 202) and the specificity was 79.0% (150 of 190).

CONCLUSION

The study results showed moderately high positive association between computed risk scores using the "prior" negative mammograms and the actual outcome of the image-detectable breast cancers in the next subsequent screening examinations. The study also demonstrated that quantitative analysis of the ipsilateral views of the mammograms enabled to provide useful information in predicting near-term breast cancer risk.

Critical texture pattern feature assessment for characterizing colonies of induced pluripotent stem cells through machine learning techniques

The objectives of this study are to assess various automated texture features obtained from the segmented colony regions of induced pluripotent stem cells (iPSCs) and confirm their potential for characterizing the colonies using different machine learning techniques. One hundred and fifty-one features quantified using shape-based, moment-based, statistical and spectral texture feature groups are extracted from phase-contrast microscopic colony images of iPSCs. The forward stepwise regression model is implemented to select the most appropriate features required for categorizing the colonies. Support vector machine (SVM), random forest (RF), multilayer perceptron (MLP), decision tree (DT), and adaptive boosting (Adaboost) classifiers are used with ten-fold cross-validation to evaluate the texture features within each texture feature group and fused-features group to characterize healthy and unhealthy colonies of iPSCs. Overall, based on the classification performances of the four texture feature groups using the five classifier models, statistical features always exhibit a high predictive capacity (>87.5%). However, the classification performance using fused texture patterns with statistical, shape-based, and moment-based features was found to be robust and reliable with fewer false positive and false negative values compared to the features when either one is used for the classification of colonies of iPSCs. Furthermore, the results showcase that the SVM, RF and Adaboost classifiers deliver better classification performances than DT and MLP. Our findings suggest that the proposed automated fused statistical, shape-based, and moment-based texture pattern features trained with machine learning techniques are potentially more appropriate and helpful to biologists for characterizing colonies of stem cells.

The effect of post-injection 18F-FDG PET scanning time on texture analysis of peripheral nerve sheath tumours in neurofibromatosis-1

BACKGROUND

Texture features are being increasingly evaluated in 18F-fluorodeoxyglucose positron emission tomography (18F-FDG PET) as adjunctive imaging biomarkers in a number of different cancers. Whilst studies have reported repeatability between scans, there have been no studies that have specifically investigated the effect that the time of acquisition post-injection of 18F-FDG has on texture features. The aim of this study was to investigate if texture features change between scans performed at different time points post-injection.

RESULTS

Fifty-four patients (30 male, 24 female, mean age 35.1 years) with neurofibromatosis-1 and suspected malignant transformation of a neurofibroma underwent 18F-FDG PET/computed tomography (CT) scans at 101.5 ± 15.0 and 251.7 ± 18.4 min post-injection of 350 MBq 18F-FDG to a standard clinical protocol. Following tumour segmentation on both early and late scans, first- (n = 37), second- (n = 25) and high-order (n = 31) statistical features, as well as fractal texture features (n = 6), were calculated and a comparison was made between the early and late scans for each feature. Of the 54 tumours, 30 were benign and 24 malignant on histological analysis or on clinical follow-up for at least 5 years. Overall, 25/37 first-order, 9/25 second-order, 13/31 high-order and 3/6 fractal features changed significantly (p < 0.05) between early and late scans. The corresponding proportions for the 30 benign tumours alone were 22/37, 7/25, 8/31 and 2/6 and for the 24 malignant tumours, 11/37, 6/25, 8/31 and 0/6, respectively.

CONCLUSIONS

Several texture features change with time post-injection of 18F-FDG. Thus, when comparing texture features in intra- and inter-patient studies, it is essential that scans are obtained at a consistent time post-injection of 18F-FDG.

A novel and robust Bayesian approach for segmentation of psoriasis lesions and its risk stratification

BACKGROUND AND OBJECTIVE

The need for characterization of psoriasis lesion severity is clinically valuable and vital for dermatologists since it provides a reliable and precise decision on risk assessment. The automated delineation of lesion is a prerequisite prior to characterization, which is challenging itself. Thus, this paper has two major objectives: (a) design of a segmentation system which can model by learning the lesion characteristics and this is posed as a Bayesian model; (b) develop a psoriasis risk assessment system (pRAS) by crisscrossing the blocks which drives the fundamental machine learning paradigm.

METHODS

The segmentation system uses the knowledge derived by the experts along with the features reflected by the lesions to build a Bayesian framework that helps to classify each pixel of the image into lesion vs.

BACKGROUND

Since this lesion has several stages and grades, hence the system undergoes the risk assessment to classify into five levels of severity: healthy, mild, moderate, severe and very severe. We build nine kinds of pRAS utilizing different combinations of the key blocks. These nine pRAS systems use three classifiers (Support Vector Machine (SVM), Decision Tree (DT) and Neural Network (NN)) and three feature selection techniques (Principal Component Analysis (PCA), Fisher Discriminant Ratio (FDR) and Mutual Information (MI)). The two major experiments conducted using these nine systems were: (i) selection of best system combination based on classification accuracy and (ii) understanding the reliability of the system. This leads us to computation of key system performance parameters such as: feature retaining power, aggregated feature effect and reliability index besides conventional attributes like accuracy, sensitivity, specificity.

RESULTS

Using the database used in this study consisted of 670 psoriasis images, the combination of SVM and FDR was revealed as the optimal pRAS system and yielded a classification accuracy of 99.84% using cross-validation protocol. Further, SVM-FDR system provides the reliability of 99.99% using cross-validation protocol.

CONCLUSIONS

The study demonstrates a fully novel model of segmentation embedded with risk assessment. Among all nine systems, SVM-FDR produced best results. Further, we validated our pRAS system with automatic segmented lesions against manually segmented lesions showing comparable performance.

Integrated local binary pattern texture features for classification of breast tissue imaged by optical coherence microscopy

This paper proposes a texture analysis technique that can effectively classify different types of human breast tissue imaged by Optical Coherence Microscopy (OCM). OCM is an emerging imaging modality for rapid tissue screening and has the potential to provide high resolution microscopic images that approach those of histology. OCM images, acquired without tissue staining, however, pose unique challenges to image analysis and pattern classification. We examined multiple types of texture features and found Local Binary Pattern (LBP) features to perform better in classifying tissues imaged by OCM. In order to improve classification accuracy, we propose novel variants of LBP features, namely average LBP (ALBP) and block based LBP (BLBP). Compared with the classic LBP feature, ALBP and BLBP features provide an enhanced encoding of the texture structure in a local neighborhood by looking at intensity differences among neighboring pixels and among certain blocks of pixels in the neighborhood. Fourty-six freshly excised human breast tissue samples, including 27 benign (e.g. fibroadenoma, fibrocystic disease and usual ductal hyperplasia) and 19 breast carcinoma (e.g. invasive ductal carcinoma, ductal carcinoma in situ and lobular carcinoma in situ) were imaged with large field OCM with an imaging area of 10 × 10 mm² (10, 000 × 10, 000 pixels) for each sample. Corresponding H&E histology was obtained for each sample and used to provide ground truth diagnosis. 4310 small OCM image blocks (500 × 500 pixels) each paired with corresponding H&E histology was extracted from large-field OCM images and labeled with one of the five different classes: adipose tissue (n = 347), fibrous stroma (n = 2,065), breast lobules (n = 199), carcinomas (pooled from all sub-types, n = 1,127), and background (regions outside of the specimens, n = 572). Our experiments show that by integrating a selected set of LBP and the two new variant (ALBP and BLBP) features at multiple scales, the classification accuracy increased from 81.7% (using LBP features alone) to 93.8% using a neural network classifier. The integrated feature was also used to classify large-field OCM images for tumor detection. A receiver operating characteristic (ROC) curve was obtained with an area under the curve value of 0.959. A sensitivity level of 100% and specificity level of 85.2% was achieved to differentiate benign from malignant samples. Several other experiments also demonstrate the complementary nature of LBP and the two variants (ALBP and BLBP features) and the significance of integrating these texture features for classification. Using features from multiple scales and performing feature selection are also effective mechanisms to improve accuracy while maintaining computational efficiency.

Correlation between tumour characteristics, SUV measurements, metabolic tumour volume, TLG and textural features assessed with 18F-FDG PET in a large cohort of oestrogen receptor-positive breast cancer patients

PURPOSE

The study was designed to evaluate 1) the relationship between PET image textural features (TFs) and SUVs, metabolic tumour volume (MTV), total lesion glycolysis (TLG) and tumour characteristics in a large prospective and homogenous cohort of oestrogen receptor-positive (ER+) breast cancer (BC) patients, and 2) the capability of those parameters to predict response to neoadjuvant chemotherapy (NAC).

METHODS

171 consecutive patients with large or locally advanced ER+ BC without distant metastases underwent an ¹⁸F-FDG PET examination before NAC. The primary tumour was delineated with an adaptive threshold segmentation method. Parameters of volume, intensity and texture (entropy, homogeneity, contrast and energy) were measured and compared with tumour characteristics determined on pre-treatment breast biopsy (Wilcoxon rank-sum test). The correlation between PET-derived parameters was determined using Spearman's coefficient. The relationship between PET features and pathological findings was determined using the Wilcoxon rank-sum test.

RESULTS

Spearman's coefficients between SUV_max and TFs were 0.43, 0.24, -0.43 and -0.15 respectively for entropy, homogeneity, energy and contrast; they were higher between MTV and TFs: 0.99, 0.86, -0.99 and -0.87. All TFs showed a significant association with the histological type (IDC vs. ILC; 0.02 < P < 0.03) but didn't with immunohistochemical characteristics. SUV_max and TLG predicted the pathological response (P = 0.0021 and P = 0.02 respectively); TFs didn't (P: 0.27, 0.19, 0.94, 0.19 respectively for entropy, homogeneity, energy and contrast).

CONCLUSIONS

The correlation of TFs was poor with SUV parameters and high with MTV. TFs showed a significant association with the histological type. Finally, while SUV_max and TLG were able to predict response to NAC, TFs failed.

Characterization of Architectural Distortion in Mammograms Based on Texture Analysis Using Support Vector Machine Classifier with Clinical Evaluation

Architecture distortion (AD) is an important and early sign of breast cancer, but due to its subtlety, it is often missed on the screening mammograms. The objective of this study is to create a quantitative approach for texture classification of AD based on various texture models, using support vector machine (SVM) classifier. The texture analysis has been done on the region of interest (ROI) selected from the original mammogram. A comprehensive analysis has been done on samples from three databases; out of which, two data sets are from the public domain, and the third data set is for clinical evaluation. The public domain databases are IRMA version of digital database for screening mammogram (DDSM) and Mammographic Image Analysis Society (MIAS). For clinical evaluation, the actual patient's database has been obtained from ACE Healthways, Diagnostic Centre Ludhiana, India. The significant finding of proposed study lies in appropriate selection of the size of ROIs. The experiments have been done on fixed size of ROIs as well as on the ground truth (variable size) ROIs. Best results pertain to an accuracy of 92.94 % obtained in case of DDSM database for fixed-size ROIs. In case of MIAS database, an accuracy of 95.34 % is achieved in AD versus non-AD (normal) cases for ground truth ROIs. Clinically, an accuracy of 88 % was achieved for ACE dataset. The results obtained in the present study are encouraging, as optimal result has been achieved for the proposed study in comparison with other related work in the same area.

New Fully Automated Method for Segmentation of Breast Lesions on Ultrasound Based on Texture Analysis

The study described here explored a fully automatic segmentation approach based on texture analysis for breast lesions on ultrasound images. The proposed method involves two main stages: (i) In lesion region detection, the original gray-scale image is transformed into a texture domain based on log-Gabor filters. Local texture patterns are then extracted from overlapping lattices that are further classified by a linear discriminant analysis classifier to distinguish between the "normal tissue" and "breast lesion" classes. Next, an incremental method based on the average radial derivative function reveals the region with the highest probability of being a lesion. (ii) In lesion delineation, using the detected region and the pre-processed ultrasound image, an iterative thresholding procedure based on the average radial derivative function is performed to determine the final lesion contour. The experiments are carried out on a data set of 544 breast ultrasound images (including cysts, benign solid masses and malignant lesions) acquired with three distinct ultrasound machines. In terms of the area under the receiver operating characteristic curve, the one-way analysis of variance test (α=0.05) indicates that the proposed approach significantly outperforms two published fully automatic methods (p<0.001), for which the areas under the curve are 0.91, 0.82 and 0.63, respectively. Hence, these results suggest that the log-Gabor domain improves the discrimination power of texture features to accurately segment breast lesions. In addition, the proposed approach can potentially be used for automated computer diagnosis purposes to assist physicians in detection and classification of breast masses.

Computer-aided diagnosis of psoriasis skin images with HOS, texture and color features: A first comparative study of its kind

Psoriasis is an autoimmune skin disease with red and scaly plaques on skin and affecting about 125 million people worldwide. Currently, dermatologist use visual and haptic methods for diagnosis the disease severity. This does not help them in stratification and risk assessment of the lesion stage and grade. Further, current methods add complexity during monitoring and follow-up phase. The current diagnostic tools lead to subjectivity in decision making and are unreliable and laborious. This paper presents a first comparative performance study of its kind using principal component analysis (PCA) based CADx system for psoriasis risk stratification and image classification utilizing: (i) 11 higher order spectra (HOS) features, (ii) 60 texture features, and (iii) 86 color feature sets and their seven combinations. Aggregate 540 image samples (270 healthy and 270 diseased) from 30 psoriasis patients of Indian ethnic origin are used in our database. Machine learning using PCA is used for dominant feature selection which is then fed to support vector machine classifier (SVM) to obtain optimized performance. Three different protocols are implemented using three kinds of feature sets. Reliability index of the CADx is computed. Among all feature combinations, the CADx system shows optimal performance of 100% accuracy, 100% sensitivity and specificity, when all three sets of feature are combined. Further, our experimental result with increasing data size shows that all feature combinations yield high reliability index throughout the PCA-cutoffs except color feature set and combination of color and texture feature sets. HOS features are powerful in psoriasis disease classification and stratification. Even though, independently, all three set of features HOS, texture, and color perform competitively, but when combined, the machine learning system performs the best. The system is fully automated, reliable and accurate.

Fast and accurate identification of fat droplets in histological images

BACKGROUND AND OBJECTIVE

The accurate identification of fat droplets is a prerequisite for the automatic quantification of steatosis in histological images. A major challenge in this regard is the distinction between clustered fat droplets and vessels or tissue cracks.

METHODS

We present a new method for the identification of fat droplets that utilizes adjacency statistics as shape features. Adjacency statistics are simple statistics on neighbor pixels.

RESULTS

The method accurately identified fat droplets with sensitivity and specificity values above 90%. Compared with commonly-used shape features, adjacency statistics greatly improved the sensitivity toward clustered fat droplets by 29% and the specificity by 17%. On a standard personal computer, megapixel images were processed in less than 0.05s.

CONCLUSIONS

The presented method is simple to implement and can provide the basis for the fast and accurate quantification of steatosis.

Breast ultrasound despeckling using anisotropic diffusion guided by texture descriptors

Breast ultrasound (BUS) is considered the most important adjunct method to mammography for diagnosing cancer. However, this image modality suffers from an intrinsic artifact called speckle noise, which degrades spatial and contrast resolution and obscures the screened anatomy. Hence, it is necessary to reduce speckle artifacts before performing image analysis by means of computer-aided diagnosis systems, for example. In addition, the trade-off between smoothing level and preservation of lesion contour details should be addressed by speckle reduction schemes. In this scenario, we propose a BUS despeckling method based on anisotropic diffusion guided by Log-Gabor filters (ADLG). Because we assume that different breast tissues have distinct textures, in our approach we perform a multichannel decomposition of the BUS image using Log-Gabor filters. Next, the conduction coefficient of anisotropic diffusion filtering is computed using texture responses instead of intensity values as stated originally. The proposed algorithm is validated using both synthetic and real breast data sets, with 900 and 50 images, respectively. The performance measures are compared with four existing speckle reduction schemes based on anisotropic diffusion: conventional anisotropic diffusion filtering (CADF), speckle-reducing anisotropic diffusion (SRAD), texture-oriented anisotropic diffusion (TOAD), and interference-based speckle filtering followed by anisotropic diffusion (ISFAD). The validity metrics are the Pratt's figure of merit, for synthetic images, and the mean radial distance (in pixels), for real sonographies. Figure of merit and mean radial distance indices should tend toward '1' and '0', respectively, to indicate adequate edge preservation. The results suggest that ADLG outperforms the four speckle removal filters compared with respect to simulated and real BUS images. For each method--ADLG, CADF, SRAD, TOAD and ISFAD--the figure of merit median values are 0.83, 0.40, 0.39, 0.51 and 0.59, and the mean radial distance median results are 4.19, 6.29, 6.39, 6.43 and 5.88.

Computer-supported diagnosis for endotension cases in endovascular aortic aneurysm repair evolution

An abdominal aortic aneurysm (AAA) is a localized abnormal enlargement of the abdominal aorta with fatal consequences if not treated on time. The endovascular aneurysm repair (EVAR) is a minimal invasive therapy that reduces recovery times and improves survival rates in AAA cases. Nevertheless, post-operation difficulties can appear influencing the evolution of treatment. The objective of this work is to develop a pilot computer-supported diagnosis system for an automated characterization of EVAR progression from CTA images. The system is based on the extraction of texture features from post-EVAR thrombus aneurysm samples and on posterior classification. Three conventional texture-analysis methods, namely the gray level co-occurrence matrix (GLCM), the gray level run length matrix (GLRLM), the gray level difference method (GLDM), and a new method proposed by the authors, the run length matrix of local co-occurrence matrices (RLMLCM), were applied to each sample. Several classification schemes were experimentally evaluated. The ensembles of a k-nearest neighbor (k-NN), a multilayer perceptron neural network (MLP-NN), and a support vector machine (SVM) classifier fed with a reduced version of texture features resulted in a better performance (Az=94.35±0.30), as compared to the classification performance of the other alternatives.

Breast Tissue Classification in Digital Tomosynthesis Images Based on Global Gradient Minimization and Texture Features

Digital breast tomosynthesis (DBT) is a pseudo-three-dimensional x-ray imaging modality proposed to decrease the effect of tissue superposition present in mammography, potentially resulting in an increase in clinical performance for the detection and diagnosis of breast cancer. Tissue classification in DBT images can be useful in risk assessment, computer-aided detection and radiation dosimetry, among other aspects. However, classifying breast tissue in DBT is a challenging problem because DBT images include complicated structures, image noise, and out-of-plane artifacts due to limited angular tomographic sampling. In this project, we propose an automatic method to classify fatty and glandular tissue in DBT images. First, the DBT images are pre-processed to enhance the tissue structures and to decrease image noise and artifacts. Second, a global smooth filter based on L0 gradient minimization is applied to eliminate detailed structures and enhance large-scale ones. Third, the similar structure regions are extracted and labeled by fuzzy C-means (FCM) classification. At the same time, the texture features are also calculated. Finally, each region is classified into different tissue types based on both intensity and texture features. The proposed method is validated using five patient DBT images using manual segmentation as the gold standard. The Dice scores and the confusion matrix are utilized to evaluate the classified results. The evaluation results demonstrated the feasibility of the proposed method for classifying breast glandular and fat tissue on DBT images.