Breast Cancer Diagnosis Using Texture and Shape Features in MRI.

Contrast-enhanced magnetic resonance (MR) breast imaging represents a tool with great potential for the detection, evaluation and diagnosis of breast cancer (BC). Due to its high sensitivity and in combination with medical imaging biomarkers, it can overcome setbacks and limitations manifested in other diagnostic modalities such as mammography or ultrasound. In order to aid and assist clinicians in the diagnosis of BC, a methodology based on the extraction of 2D texture and 3D shape features in MR images is proposed. To categorize breast tumor malignancy, we considered its location in the coronal plane, divided into 4 quadrants (UOQ, UIQ, LOQ and LOQ), and the tumor type according to its genetic information (positive HER2 and Luminal B with negative HER2). In this regard, six different studies were conducted: one per feature type (texture and shape), as well as the combination of both features (texture + shape) for each of the two covariables (tumor type and location in the coronal plane). A dataset of 43 BC patients were considered. A radiomics approach was implemented extracting 43 texture and 17 shape features and using to train 5 different predictive models (Linear SVM, Gaussian SVM, Bagged Tree, KNN and Naïve Bayes). The highest precision result for the tumor type study (74.04% in terms of AUC) was obtained with 43 texture features. Whereas for the quadrant localization study, the highest precision result (67.99% AUC) was obtained as a combination of 3 textures and shape features. Both results were achieved with the SVM with Linear Kernel classification model.Clinical Relevance- This work emphasizes the use of quantitative biomarkers as texture and shape features in combination with machine learning techniques to aid in breast tumor malignancy diagnosis on MR imaging. Moreover, considering the location of the tumor in the coronal plane and its type according to its genetic information may improve the selection of appropriate treatments, survival rate, and quality of life for breast cancer patients.

Assessment and risk prediction of frailty using texture-based muscle ultrasound image analysis and machine learning techniques.

The purpose of this study was to evaluate texture-based muscle ultrasound image analysis for the assessment and risk prediction of frailty phenotype. This retrospective study of prospectively acquired data included 101 participants who underwent ultrasound scanning of the anterior thigh. Participants were subdivided according to frailty phenotype and were followed up for two years. Primary and secondary outcome measures were death and comorbidity, respectively. Forty-three texture features were computed from the rectus femoris and the vastus intermedius muscles using statistical methods. Model performance was evaluated by computing the area under the receiver operating characteristic curve (AUC) while outcome prediction was evaluated using regression analysis. Models developed achieved a moderate to good AUC (0.67 ≤ AUC ≤ 0.79) for categorizing frailty. The stepwise multiple logistic regression analysis demonstrated that they correctly classified 70-87% of the cases. The models were associated with increased comorbidity (0.01 ≤ p ≤ 0.18) and were predictive of death for pre-frail and frail participants (0.001 ≤ p ≤ 0.016). In conclusion, texture analysis can be useful to identify frailty and assess risk prediction (i.e. mortality) using texture features extracted from muscle ultrasound images in combination with a machine learning approach.

MRI texture-based radiomics analysis for the identification of altered functional networks in alcoholic patients and animal models.

Alcohol use disorder (AUD) is a complex condition representing a leading risk factor for death, disease and disability. Its high prevalence and severe health consequences make necessary a better understanding of the brain network alterations to improve diagnosis and treatment. The purpose of this study was to evaluate the potential of resting-state fMRI 3D texture features as a novel source of biomarkers to identify AUD brain network alterations following a radiomics approach. A longitudinal study was conducted in Marchigian Sardinian alcohol-preferring msP rats (N = 36) who underwent resting-state functional and structural MRI before and after 30 days of alcohol or water consumption. A cross-sectional human study was also conducted among 33 healthy controls and 35 AUD patients. The preprocessed functional data corresponding to control and alcohol conditions were used to perform a probabilistic independent component analysis, identifying seven independent components as resting-state networks. Forty-three radiomic features extracted from each network were compared using a Wilcoxon signed-rank test with Holm correction to identify the network most affected by alcohol consumption. Features extracted from this network were then used in the machine learning process, evaluating two feature selection methods and six predictive models within a nested cross-validation structure. The classification was evaluated by computing the area under the ROC curve. Images were quantized using different numbers of gray-levels to test their influence on the results. The influence of ageing, data preprocessing, and brain iron accumulation were also analyzed. The methodology was validated using structural scans. The striatal network in alcohol-exposed msP rats presented the most significant number of altered features. The radiomics approach supported this result achieving good classification performance in animals (AUC = 0.915 ± 0.100, with 12 features) and humans (AUC = 0.724 ± 0.117, with 9 features) using a random forest model. Using the structural scans, high accuracy was achieved with a multilayer perceptron in both species (animals: AUC > 0.95 with 2 features, humans: AUC > 0.82 with 18 features). The best results were obtained using a feature selection method based on the p-value. The proposed radiomics approach is able to identify AUD patients and alcohol-exposed rats with good accuracy, employing a subset of 3D features extracted from fMRI. Furthermore, it can help identify relevant networks in drug addiction.

Detection and Classification of Myocardial Infarction Transmurality Using Cardiac MR Image Analysis and Machine Learning Algorithms.

The presence of abnormalities when the left ventricle is deformed is related to the patients' prognosis after a first myocardial infarction. These deformations can be detected by performing a cardiac magnetic resonance (CMR) study. Currently, late gadolinium enhancement (LGE) is considered to be the gold standard when performing CMR imaging. However, CMR with LGE overestimates infarct size and underestimates recovery of dysfunctional segments after myocardial infarction. Based on this statement, the objective is to detect, characterize, and quantify the extent of myocardial infarction in patients with cardiac pathologies, using parameters derived from CMR, in order to obtain greater precision in patients' recovery predictions than when only studying LGE images. For this purpose, we studied the infarct presence and extension from a total of 105 images from 35 patients, and calculated myocardium strain and torsion to characterize and quantify the affected tissue. A total of twenty-one parameters were selected to create predictive models. Moreover, we compared two feature extraction methods, and the performance of five machine learning algorithms. Results show that both temporal and strain parameters are the most relevant to detect and characterize the extent of myocardial infarction. The use of imaging techniques and machine learning algorithms have great potential and show promising results when it comes to detecting the presence and extent of myocardial infarction. The current study proposes a novel approach to detect, quantify, and characterize cardiac infarction by using strain and torsion parameters from different CMR images and different Machine Learning algorithms. This would potentially overcome LGE, the current state of the art technique, in estimating the extension of damaged tissue and enable an objective diagnosis and clinical decision.

Analysis of New Biomarkers for the Study of Schizophrenia Following a Radiomics Approach on MR and PET Imaging.

Traditionally, the diagnosis of schizophrenia was based on the psychiatrist's introspective diagnosis through clinical stratification factors and score-scales, which led to heterogeneity and discrepancy in the symptoms and results. However, there are many studies trying to improve and assist in how its diagnosis could be performed. To objectively classify schizophrenia patients it is required to determine quantitative biomarkers of the disease. In this contribution we propose a method based on feature extraction both in magnetic resonance (MR) and Positron Emission Tomography (PET) imaging. A dataset of 34 participants (17 patients and 17 control subjects) were analyzed and 5 different brain regions were studied (frontal cortex, posterior cingulate cortex, temporal cortex, primary auditory cortex and thalamus). Following a radiomics approach, 43 texture features were extracted using five different statistical methods. These features were used for the training of the five different predictive models (Linear SVM, Gaussian SVM, Bagged Tree, KNN and Naive Bayes). The precision results were obtained classifying schizophrenia both in MR images (89% Area Under the Curve (AUC) in the posterior cingulate cortex) and with PET images (82% AUC in the frontal cortex), being Linear SVM and Naive Bayes the classification models with the highest predictive power. Clinical Relevance- The current study establishes a methodology to classify schizophrenia disease based on quantitative biomarkers using MR and PET images. This tool could assist the psychiatrist as an additional criterion for the diagnosis evaluation.

Multiclass Classification of Prostate Tumors Following an MR Image Analysis-Based Radiomics Approach.

Prostate cancer is one of the most common cancers in men, with symptoms that may be confused with those caused by benign prostatic hyperplasia. One of the key aspects of treating prostate cancer is its early detection, increasing life expectancy and improving the quality of life of those patients. However, the tests performed are often invasive, resulting in a biopsy. A non-invasive alternative is the magnetic resonance imaging (MRI)-based PI-RADS v2 classification. The aim of this work was to find objective biomarkers that allow the PI-RADS classification of prostate lesions using a radiomics approach on Multiparametric MRI. A total of 90 subjects were analyzed. From each segmented lesion, 609 different texture features were extracted using five different statistical methods. Two feature selection methods and eight multiclass predictive models were evaluated. This was a multiclass study in which the best AUC result was 0.7442 ± 0.0880, achieved with the Naïve Bayes model using a subset of 120 features. Valuable results were also obtained using the Random Forests model, obtaining an AUC of 0.7394 ± 0.0965 with a lower number of features (52). Clinical Relevance- The current study establishes a methodology for classifying prostate cancer and supporting clinical decision-making in a fast and efficient manner and avoiding additional invasive procedures using MRI.

Texture Analysis in MRI of the Knee for an Early Diagnosis of Osteoarthritis.

Osteoarthritis is one of the most disabling diseases in developed countries. Its etiology is not firmly established, and the diagnosis is made by observing radiographs, assigning a degree of severity based on the information displayed. For this reason, the diagnosis is usually late and determined by the subjectivity of the doctor, which implies a restriction of the treatment. Magnetic resonance imaging (MRI) has allowed us to see in greater detail the alterations produced in soft joint structures. In this work, biomarkers for an early diagnosis of knee osteoarthritis have been developed by means of textures analysis on MRI. For this purpose, 50 subjects underwent T1-weighted MR image acquisitions: 25 controls and 25 diagnosed with knee osteoarthritis between grades I and III. Six regions were segmented on these images, corresponding to the femorotibial cartilage, femoral condyles, and tibial plateau. 43 textures were extracted for each region of interest (ROI) employing 5 statistical methods and 5 different predictive models were trained and compared. In addition, a study of the thickness of the cartilage was carried out to make a comparison with the texture analysis. The best result has been obtained using a K-nearest neighbor model with the combination of 33 textures (maximum value of AUC = 0.7684). Furthermore, in the analysis of the cartilage thickness, no statistically significant differences were found. Finally, it is concluded that the texture analysis has great potential for the diagnosis of knee osteoarthritis. Clinical Relevance - The current study establishes a methodology for an early diagnosis of knee osteoarthritis by means of MRI-based texture analysis, in a fast and objective manner.

Glioblastomas and brain metastases differentiation following an MRI texture analysis-based radiomics approach.

PURPOSE: To evaluate the potential of 2D texture features extracted from magnetic resonance (MR) images for differentiating brain metastasis (BM) and glioblastomas (GBM) following a radiomics approach. METHODS: This retrospective study included 50 patients with BM and 50 with GBM who underwent T1-weighted MRI between December 2010 and January 2017. Eighty-eight rotation-invariant texture features were computed for each segmented lesion using six texture analysis methods. These features were also extracted from the four images obtained after applying the discrete wavelet transform (88 features × 4 images). Three feature selection methods and five predictive models were evaluated. A 5-fold cross-validation scheme was used to randomly split the study group into training (80 patients) and testing (20 patients), repeating the process ten times. Classification was evaluated computing the average area under the receiver operating characteristic curve. Sensibility, specificity and accuracy were also computed. The whole process was tested quantizing the images with different gray-level values to evaluate their influence in the final results. RESULTS: Highest classification accuracy was obtained using the original images quantized with 128 gray-levels and a feature selection method based on the p-value. The best overall performance was achieved using a support vector machine model with a subset of 32 features (AUC = 0.896 ± 0.067, sensitivity of 82% and specificity of 80%). Naïve Bayes and k-nearest neighbors models showed also valuable results (AUC ≈ 0.8) with a lower number of features (<13), thus suggesting that these models may be more generalizable when using external validations. CONCLUSION: The proposed radiomics MRI approach is able to discriminate between GBM and BM with high accuracy employing a set of 2D texture features, thus helping in the diagnosis of brain lesions in a fast and non-invasive way.

Classifying brain metastases by their primary site of origin using a radiomics approach based on texture analysis: a feasibility study.

OBJECTIVE: To examine the capability of MRI texture analysis to differentiate the primary site of origin of brain metastases following a radiomics approach. METHODS: Sixty-seven untreated brain metastases (BM) were found in 3D T1-weighted MRI of 38 patients with cancer: 27 from lung cancer, 23 from melanoma and 17 from breast cancer. These lesions were segmented in 2D and 3D to compare the discriminative power of 2D and 3D texture features. The images were quantized using different number of gray-levels to test the influence of quantization. Forty-three rotation-invariant texture features were examined. Feature selection and random forest classification were implemented within a nested cross-validation structure. Classification was evaluated with the area under receiver operating characteristic curve (AUC) considering two strategies: multiclass and one-versus-one. RESULTS: In the multiclass approach, 3D texture features were more discriminative than 2D features. The best results were achieved for images quantized with 32 gray-levels (AUC = 0.873 ± 0.064) using the top four features provided by the feature selection method based on the p-value. In the one-versus-one approach, high accuracy was obtained when differentiating lung cancer BM from breast cancer BM (four features, AUC = 0.963 ± 0.054) and melanoma BM (eight features, AUC = 0.936 ± 0.070) using the optimal dataset (3D features, 32 gray-levels). Classification of breast cancer and melanoma BM was unsatisfactory (AUC = 0.607 ± 0.180). CONCLUSION: Volumetric MRI texture features can be useful to differentiate brain metastases from different primary cancers after quantizing the images with the proper number of gray-levels. KEY POINTS: • Texture analysis is a promising source of biomarkers for classifying brain neoplasms. • MRI texture features of brain metastases could help identifying the primary cancer. • Volumetric texture features are more discriminative than traditional 2D texture features.

Automatic segmentation of the spine by means of a probabilistic atlas with a special focus on ribs suppression.

PURPOSE: The development of automatic and reliable algorithms for the detection and segmentation of the vertebrae are of great importance prior to any diagnostic task. However, an important problem found to accurately segment the vertebrae is the presence of the ribs in the thoracic region. To overcome this problem, a probabilistic atlas of the spine has been developed dealing with the proximity of other structures, with a special focus on ribs suppression. METHODS: The data sets used consist of Computed Tomography images corresponding to 21 patients suffering from spinal metastases. Two methods have been combined to obtain the final result: firstly, an initial segmentation is performed using a fully automatic level-set method; secondly, to refine the initial segmentation, a 3D volume indicating the probability of each voxel of belonging to the spine has been developed. In this way, a probability map is generated and deformed to be adapted to each testing case. RESULTS: To validate the improvement obtained after applying the atlas, the Dice coefficient (DSC), the Hausdorff distance (HD), and the mean surface-to-surface distance (MSD) were used. The results showed up an average of 10 mm of improvement accuracy in terms of HD, obtaining an overall final average of 15.51 ± 2.74 mm. Also, a global value of 91.01 ± 3.18% in terms of DSC and a MSD of 0.66 ± 0.25 mm were obtained. The major improvement using the atlas was achieved in the thoracic region, as ribs were almost perfectly suppressed. CONCLUSION: The study demonstrated that the atlas is able to detect and appropriately eliminate the ribs while improving the segmentation accuracy.

Semiautomatic computer-aided classification of degenerative lumbar spine disease in magnetic resonance imaging.

BACKGROUND: Computer-aided diagnosis (CAD) methods for detecting and classifying lumbar spine disease in Magnetic Resonance imaging (MRI) can assist radiologists to perform their decision-making tasks. In this paper, a CAD software has been developed able to classify and quantify spine disease (disc degeneration, herniation and spinal stenosis) in two-dimensional MRI. METHODS: A set of 52 lumbar discs from 14 patients was used for training and 243 lumbar discs from 53 patients for testing in conventional two-dimensional MRI of the lumbar spine. To classify disc degeneration according to the gold standard, Pfirrmann classification, a method based on the measurement of disc signal intensity and structure was developed. A gradient Vector Flow algorithm was used to extract disc shape features and for detecting contour abnormalities. Also, a signal intensity method was used for segmenting and detecting spinal stenosis. Novel algorithms have also been developed to quantify the severity of these pathologies. Variability was evaluated by kappa (k) and intra-class correlation (ICC) statistics. RESULTS: Segmentation inaccuracy was below 1%. Almost perfect agreement, as measured by the k and ICC statistics, was obtained for all the analyzed pathologies: disc degeneration (k=0.81 with 95% CI=[0.75..0.88]) with a sensitivity of 95.8% and a specificity of 92.6%, disc herniation (k=0.94 with 95% CI=[0.87..1]) with a sensitivity of 60% and a specificity of 87.1%, categorical stenosis (k=0.94 with 95% CI=[0.90..0.98]) and quantitative stenosis (ICC=0.98 with 95% CI=[0.97..0.98]) with a sensitivity of 70% and a specificity of 81.7%. DISCUSSION: The proposed methods are reproducible and should be considered as a possible alternative when compared to reference standards.

Magnetic resonance microimaging of a swine infarcted heart: Performing cardiac virtual histologies.

The aim of this study is to develop a computer-aided intuitive software tool based on MATLAB to reproduce the functions of a virtual histology over Magnetic Resonance (MR) microimages of small samples of swine's infarcted hearts. The basic characterization consists of selecting regions of interest (ROIs) of that MR microimage and extracting the most important information of these regions. The software tool will implement intuitive and sophisticated tools that allow the user to define ROIs on the different types of images provided by the MR scanner. The final purpose of this tool will be to analyze the acquired data in order to characterize some aspects of the later possible events after a myocardial infarction in swine's hearts and expand the study to human cases.

Automatic segmentation of the spine by means of a probabilistic atlas with a special focus on ribs suppression. Preliminary results.

Spine is a structure commonly involved in several prevalent diseases. In clinical diagnosis, therapy, and surgical intervention, the identification and segmentation of the vertebral bodies are crucial steps. However, automatic and detailed segmentation of vertebrae is a challenging task, especially due to the proximity of the vertebrae to the corresponding ribs and other structures such as blood vessels. In this study, to overcome these problems, a probabilistic atlas of the spine, including cervical, thoracic and lumbar vertebrae has been built to introduce anatomical knowledge in the segmentation process, aiming to deal with overlapping gray levels and the proximity to other structures. From a set of 3D images manually segmented by a physician (training data), a 3D volume indicating the probability of each voxel of belonging to the spine has been developed, being necessary the generation of a probability map and its deformation to adapt to each patient. To validate the improvement of the segmentation using the atlas developed in the testing data, we computed the Hausdorff distance between the manually-segmented ground truth and an automatic segmentation and also between the ground truth and the automatic segmentation refined with the atlas. The results are promising, obtaining a higher improvement especially in the thoracic region, where the ribs can be found and appropriately eliminated.

A fully automated level-set based segmentation method of thoracic and lumbar vertebral bodies in Computed Tomography images.

Spine is a structure commonly involved in several diseases. Identification and segmentation of the vertebral structures are of relevance to many medical applications related to the spine such as diagnosis, therapy or surgical intervention. However, the development of automatic and reliable methods are an unmet need. This work presents a fully automatic segmentation method of thoracic and lumbar vertebral bodies from Computed Tomography images. The procedure can be divided into four main stages: firstly, seed points were detected in the spinal canal in order to generate initial contours in the segmentation process, automating the whole process. Secondly, a processing step is performed to improve image quality. Third step was to carry out the segmentation using the Selective Binary Gaussian Filtering Regularized Level Set method and, finally, two morphological operations were applied in order to refine the segmentation result. The method was tested in clinical data coming from 10 trauma patients. To evaluate the result the average value of the DICE coefficient was calculated, obtaining a 90.86 ± 1.87% in the whole spine (thoracic and lumbar regions), a 86.08 ± 1.73% in the thoracic region and a 95,61 ±2,25% in the lumbar region. The results are highly competitive when compared to the results obtained in previous methods, especially for the lumbar region.

Functional diffusion map: A biomarker of brain metastases response to treatment based on magnetic resonance image analysis.

Validated biomarkers for treatment response in patients suffering from brain metastases are needed in daily clinical practice as they may improve survival by providing reliable prognostic information and allowing alternative therapies. This work presents a new analysis tool for an early and non-invasive evaluation of treatment response in patients with brain metastases. A set of twenty-five metastases from sixteen patients were examined by T1-weighted and diffusion magnetic resonance imaging before starting radiotherapy and at least once after treatment. Diffusion MRI can show a correlation between water diffusion variation within metastasis area and its clinical evolution. Images were co-registered to pretreatment scans. Diffusion changes, resulting in spatially varying changes in apparent diffusion coefficient values of metastatic lesions, were quantified and presented as a functional diffusion map (fDM). These functional maps were compared to two traditional criteria for assessing oncological response. Of the twenty-five metastases analyzed, seven were classified as partial response (PR), eight as stable disease (SD) and nine as progressive disease (PD). Normalized volume values of the metastases for each response group were obtained, disclosing that apparent diffusion coefficient increase was a good predictor of response. Sensitivity was 88%, specificity 100%, positive predictive value 100% and negative predictive value was 94%. Outcome reveals that the implemented tool, based on functional diffusion mapping as evolution biomarker, provides a reliable prediction of metastases response to treatment.

Automatic detection of local arterial input functions through Independent Component Analysis on Dynamic Contrast enhanced Magnetic Resonance Imaging.

Arterial Input Function (AIF) is obtained from perfusion studies as a basic parameter for the calculus of hemodynamic variables used as surrogate markers of the vascular status of tissues. However, at present, its identification is made manually leading to high subjectivity, low repeatability and considerable time consumption. We propose an alternative method to automatically identify local AIF in perfusion images using Independent Component Analysis.