Magnetic Resonance Imaging (MRI) is a non-invasive tool for the clinical assessment of low-prevalence neuromuscular disorders. Automated diagnosis methods might reduce the need for biopsies and provide valuable information on disease follow-up. In this paper, three methods are proposed to classify target muscles in Collagen VI-related myopathy cases, based on their degree of involvement, notably a Convolutional Neural Network, a Fully Connected Network to classify texture features, and a hybrid method combining the two feature sets. The proposed methods were evaluated on axial T1-weighted Turbo Spin-Echo MRI from 26 subjects, including Ullrich Congenital Muscular Dystrophy and Bethlem Myopathy patients at different evolution stages. The hybrid model achieved the best cross-validation results, with a global accuracy of 93.8%, and F-scores of 0.99, 0.82, and 0.95, for healthy, mild and moderate/severe cases, respectively.

Magnetic Resonance Imaging (MRI) is a non-invasive tool for the clinical assessment of low-prevalence neuromuscular disorders. Automated diagnosis methods might reduce the need for biopsies and provide valuable information on disease follow-up. In this paper, three methods are proposed to classify target muscles in Collagen VI-related myopathy cases, based on their degree of involvement, notably a Convolutional Neural Network, a Fully Connected Network to classify texture features, and a hybrid method combining the two feature sets. The proposed methods were evaluated on axial T1-weighted Turbo Spin-Echo MRI from 26 subjects, including Ullrich Congenital Muscular Dystrophy and Bethlem Myopathy patients at different evolution stages. The hybrid model achieved the best cross-validation results, with a global accuracy of 93.8%, and F-scores of 0.99, 0.82, and 0.95, for healthy, mild and moderate/severe cases, respectively.

Muscle texture analysis in Magnetic Resonance Imaging (MRI) has revealed a good correlation with typical histological changes resulting from neuromuscular disorders. In this research, we assess the effectiveness of several features in describing intramuscular texture alterations in cases of Collagen VI-related myopathy. A T1-weighted Turbo Spin-Echo MRI dataset was used (N_subj = 26), consisting of thigh scans from subjects diagnosed with Ullrich Congenital Muscular Dystrophy or Bethlem Myopathy, with different severity levels, as well as healthy subjects. A total of 355 texture features were studied, including attributes derived from the Gray-Level Co-occurrence Matrix, the Run-Length Matrix, Wavelet and Gabor filters. The extracted features were ranked using the Support Vector Machine Recursive Feature Elimination (SVM-RFE) algorithm with Correlation Bias Reduction, prior to cross-validated classification with a Gaussian kernel SVM.

OCT exams are increasingly used for the screening of clinical pathologies in ophthalmology. With an aging population, yet with its associated longevity, these exams are becoming more frequent as a way to assess the health of the retina and surrounding tissues. These exams are expensive, and therefore, are much more common in hospital centres, as they are the ones that serve the majority of the Portuguese population. Doctors sometimes see up to 24 patients in a morning, which is significant when considering that for each patient there may be a request for this exam, along with the time required for its prior or precise execution. These numbers contribute to making the whole process a “burden” for both patients and clinicians. One way to attempt to optimize the ophthalmology field, particularly regarding OCT, is the introduction of Artificial Intelligence, bringing advantages for both doctors and patients, especially by reducing potential omissions by clinicians and decreasing patient wait times. This dissertation aims to study the applicability of an algorithm or set of algorithms and Machine Learning (ML) techniques in the medical field of ophthalmology, specifically in the execution of OCT and the subsequent identification of pathologies that are accepted by the medical community. This work also seeks to develop a model or models that meet the desired conditions and can be applied in real-world situations. The project was divided into five key parts: establishing a cooperation protocol with a hospital entity to gather a dataset based on real and local patients; surveying the state of the art regarding applied algorithms and optimization in published datasets, and identifying the relevant data for feeding/training the model; implementing a model for comparison and establishing a reference scenario; applying the model to the mentioned scenario with an evaluation of the impact of possible parameter adjustments on model performance and estimating potential gains.

Melanoma, the most dangerous skin cancer, is sometimes associated with a nevus, a relatively common skin lesion. To find early melanoma, nevus, and other lesions, dermoscopy is often used. In this context, intelligent methods have been applied in dermoscopic images to support decision making. A typical computer-aided diagnosis method comprises three steps: (1) extraction of features that describe image properties, (2) selection of important features previously extracted, (3) classification of images based on the selected features. In this work, traditional data mining approaches underexploited in dermoscopy were applied: information gain for feature selection and an ensemble classification method based on gradient boosting. The former technique ranks image features according to data entropy, while the latter combines the outputs of single classifiers to predict the image class. After evaluating these approaches in a public dataset, we can observe that the results obtained are competitive with the state-of-the-art. Moreover, the presented approach allows a reduction of the total number of features and types of features to produce similar classification scores.

Early diagnosis is still the most important factor to deal with skin cancer, a disease that challenges physicians and researchers. It has benefited from computer-aided diagnosis methods that successfully combine dermoscopy, Digital Image Processing, and Machine Learning techniques. This paper aims to approximate medical professionals working with dermoscopy to these methods, to join the challenge of melanoma early detection. Accordingly, a proposal for extracting, selecting and combining texture and shape features from dermoscopic images is presented. The Feature Selection task is added to the learning process to potentiate the quality of classification models. Three classical Machine Learning algorithms were applied to differentiate melanoma from non-melanoma images. The models are evaluated by standard performance measures and a multi-criteria decision analysis method. This is the first time such method is used in melanoma diagnosis. As a result, we found a decision tree that performs well and allows the explicit representation and analysis of the knowledge learned from the images. In addition, the competitiveness of our decision models in comparison with literature approaches reviewed in this work encourages further applications of Machine Learning and Feature Selection to assist computer-aided diagnosis.

Texture analysis describes a variety of image analysis techniques that quantify the variation in intensity and pattern. This paper provides an overview of several texture analysis approaches addressing the rationale supporting them, their advantages, drawbacks, and applications. This survey’s emphasis is in collecting and categorising over five decades of active research on texture analysis.Brief descriptions of different approaches are presented along with application examples. From a broad range of texture analysis applications, this survey’s final focus is on biomedical image analysis. An up-to-date list of biological tissues and organs in which disorders produce texture changes that may be used to spot disease onset and progression is provided. Finally, the role of texture analysis methods as biomarkers of disease is summarised.