Rui Fonseca-Pinto

Screening of obstructive sleep apnea using Hilbert-Huang decomposition of oronasal airway pressure recordings

Polysomnographic signals are usually recorded from patients exhibiting symptoms related to sleep disorders such as obstructive sleep apnea (OSA). Analysis of polysomnographic data allows for the determination of the type and severity of sleep apnea or other sleep-related disorders by a specialist or technician. The usual procedure entails an overnight recording several hours long. This paper presents a methodology to help with the screening of OSA using a 5-min oronasal airway pressure signal emanating from a polysomnographic recording during the awake period, eschewing the need for an overnight recording. The clinical sample consisted of a total of 41 subjects, 20 non-OSA individuals and 21 individuals with OSA. A signal analysis technique based on the Hilbert-Huang transform was used to extract intrinsic oscillatory modes from the signals. The frequency distribution of both the first mode and second mode and their sum were shown to differ significantly between non-OSA subjects and OSA patients. An index measure based on the distribution frequencies of the oscillatory modes yielded a sensitivity of 81.0% (for 95% specificity) for the detection of OSA. Two other index measures based on the relation between the area and the maximum of the 1st and 2nd halves of the frequency histogram both yielded a sensitivity of 76.2% (for 95% specificity). Although further tests will be needed to test the reproducibility of these results, the proposed measures seem to provide a fast method to screen OSA patients, thus reducing the costs and the waiting time for diagnosis.

Functional abolition of carotid body activity restores insulin action and glucose homeostasis in rats: key roles for visceral adipose tissue and the liver

Aims/hypothesisWe recently described that carotid body (CB) over-activation is involved in the aetiology of insulin resistance and arterial hypertension in animal models of the metabolic syndrome. Additionally, we have demonstrated that CB activity is increased in animal models of insulin resistance, and that carotid sinus nerve (CSN) resection prevents the development of insulin resistance and arterial hypertension induced by high-energy diets. Here, we tested whether the functional abolition of CB by CSN transection would reverse pre-established insulin resistance, dyslipidaemia, obesity, autonomic dysfunction and hypertension in animal models of the metabolic syndrome. The effect of CSN resection on insulin signalling pathways and tissue-specific glucose uptake was evaluated in skeletal muscle, adipose tissue and liver.MethodsExperiments were performed in male Wistar rats submitted to two high-energy diets: a high-fat diet, representing a model of insulin resistance, hypertension and obesity, and a high-sucrose diet, representing a lean model of insulin resistance and hypertension. Half of each group was submitted to chronic bilateral resection of the CSN. Age-matched control rats were also used.ResultsCSN resection normalised systemic sympathetic nervous system activity and reversed weight gain induced by high-energy diets. It also normalised plasma glucose and insulin levels, insulin sensitivity lipid profile, arterial pressure and endothelial function by improving glucose uptake by the liver and perienteric adipose tissue.Conclusions/interpretationWe concluded that functional abolition of CB activity restores insulin sensitivity and glucose homeostasis by positively affecting insulin signalling pathways in visceral adipose tissue and liver.

On the influence of time-series length in EMD to extract frequency content: Simulations and models in biomedical signals

In this paper, fractional Gaussian noise (fGn) was used to simulate a homogeneously spreading broadband signal without any dominant frequency band, and to perform a simulation study about the influence of time-series length in the number of intrinsic mode functions (IMFs) obtained after empirical mode decomposition (EMD). In this context three models are presented. The first two models depend on the Hurst exponent H, and the last one is designed for small data lengths, in which the number of IMFs after EMD is obtained based on the regularity of the signal, and depends on an index measure of regularity. These models contribute to a better understanding of the EMD decomposition through the evaluation of its performance in fGn signals. Since an analytical formulation to evaluate the EMD performance is not available, using well-known signals allows for a better insight into the process. The last model presented is meant for application to real data. Its purpose is to predict, in function of the regularity signal, the time-series length that should be used when one wants to divide the spectrum into a pre-determined number of modes, corresponding to different frequency bands, using EMD. This is the case, e.g., in heart rate and blood pressure signals, used to assess sympathovagal balance in the central nervous system.

A New Tool for Nonstationary and Nonlinear Signals: The Hilbert-Huang Transform in Biomedical Applications

Time-frequency techniques constitutes a major improvement in signal analysis, namely at the field of biomedical signals in which the interdisciplinary nature of the proposed questions implies the development of new strategies to answer to specific problems. Timefrequency analysis using Wavelets, Wigner-Ville transform and more recently the HilbertHuang Transform (HHT) constitutes the core of these tools with applications in biomedical signals in last years. The non-linearity and non-stationarity nature of these signals puts HHT as a powerful tool to process signals with those properties, avoiding artefacts related to the use of linear and stationary assumptions. Classical spectral analysis using Fourier Transform still the most commonly used method when one wants to measure the global power-frequency distribution (power spectrum) of a given signal. In all areas of knowledge, Fourier-based analysis of time-series data faces constraining limitations. In biomedical signals, the critical constraining factors are the shortness of total data span, the non-stationarity of the data and the nonlinearity of the underlying physiological process. Methods using Short Time Fourier Transform (STFT) are able to extract the spectral information by defining short time windows and locally computing the Fourier transform, thereby coping with non-stationary phenomena. The frequency resolution is inversely proportional to the window length, and changes in time resolution (window length) compromise the frequency resolution. Even with optimized joint time-frequency localization, the trade-off between time and frequency resolution is unavoidable. In spite of these limitations, classical Fourier spectral analysis is still widely used to process biomedical data, for lack of alternatives. The uncritical use of Fourier spectral analysis and the careless adoption of the stationary and linear assumptions may give misleading results. Wavelet theory developed in the 90’s of last century was a significant contribution to tackle the problem of non-stationarity in time-series analysis. In common with Fourier-based analysis such as STFT, wavelet analysis yields a time-frequency representation, the main difference being that the decomposition is not based on sinusoidal functions, but rather

Classification of reticular pattern and streaks in dermoscopic images based on texture analysis.

Abstract. The early detection of melanoma is one of the greatest challenges in clinical practice of dermatology, and the reticular pattern is one of the most important dermoscopic structures to improve melanocytic lesion diagnosis. A texture-based approach is developed for the automatic detection of reticular patterns, whose output will assist clinical decision-making. Feature selection was based on the use of two algorithms by means of the classical graylevel co-occurrence matrix and Laws energy masks optimized on a set of 104 dermoscopy images. The AdaBoost (adaptive boosting) approach to machine learning was used within this strategy. Results suggest superiority of LEM for reticular pattern detection in dermoscopic images, achieving a sensitivity of 90.16% and a specificity of 86.67%. The use of automatic classification in dermoscopy to support clinicians is a strong tool to assist diagnosis; however, the use of automatic classification as a complementary tool in clinical routine requires algorithms with high levels of sensitivity and specificity. The results presented in this work will contribute to achieving this goal.

On the geometric modulation of skin lesion growth: a mathematical model for melanoma

Introduction Early detection of suspicious skin lesions is critical to prevent skin malignancies, particularly the melanoma, which is the most dangerous form of human skin cancer. In the last decade, image processing techniques have been an increasingly important tool for early detection and mathematical models play a relevant role in mapping the progression of lesions. Methods This work presents an algorithm to describe the evolution of the border of the skin lesion based on two main measurable markers: the symmetry and the geometric growth path of the lesion. The proposed methodology involves two dermoscopic images of the same melanocytic lesion obtained at different moments in time. By applying a mathematical model based on planar linear transformations, measurable parameters related to symmetry and growth are extracted. Results With this information one may compare the actual evolution in the lesion with the outcomes from the geometric model. First, this method was tested on predefined images whose growth was controlled and the symmetry known which were used for validation. Then the methodology was tested in real dermoscopic melanoma images in which the parameters of the mathematical model revealed symmetry and growth rates consistent with a typical melanoma behavior. Conclusions The method developed proved to show very accurate information about the target growth markers (variation on the growth along the border, the deformation and the symmetry of the lesion trough the time). All the results, validated by the expected phantom outputs, were similar to the ones on the real images.

Reticular pattern detection in dermoscopy: an approach using Curvelet Transform

Introduction Dermoscopy is a non-invasive in vivo imaging technique, used in dermatology in feature identification, among pigmented melanocytic neoplasms, from suspicious skin lesions. Often, in the skin exam is possible to ascertain markers, whose identification and proper characterization is difficult, even when it is used a magnifying lens and a source of light. Dermoscopic images are thus a challenging source of a wide range of digital features, frequently with clinical correlation. Among these markers, one of particular interest to diagnosis in skin evaluation is the reticular pattern. Methods This paper presents a novel approach (avoiding pre-processing, e.g. segmentation and filtering) for reticular pattern detection in dermoscopic images, using texture spectral analysis. The proposed methodology involves a Curvelet Transform procedure to identify features. Results Feature extraction is applied to identify a set of discriminant characteristics in the reticular pattern, and it is also employed in the automatic classification task. The results obtained are encouraging, presenting Sensitivity and Specificity of 82.35% and 76.79%, respectively. Conclusions These results highlight the use of automatic classification, in the context of artificial intelligence, within a computer-aided diagnosis strategy, as a strong tool to help the human decision making task in clinical practice. Moreover, the results were obtained using images from three different sources, without previous lesion segmentation, achieving to a rapid, robust and low complexity methodology. These properties boost the presented approach to be easily used in clinical practice as an aid to the diagnostic process.

An Adaptive Approach for Skin Lesion Segmentation in Dermoscopy Images Using a Multiscale Local Normalization

Skin cancer is one of the most common malignancies in humans. Early detection of suspicious skin signs is critical to prevent this kind of malignancy, and various disciplines can play a crucial role in its detection. The lesion border is especially relevant for diagnosis, and provides information on the shape of the lesion, growth path, and growth rate. Digital image processing methods can be used to perform automatic lesion border detection; nonetheless, the presence of artifacts may induce artificial borders, thereby jeopardizing the efficiency of automatic detection algorithms. Artifact removal is a necessary pre-processing step to improve the accuracy quality of the border identification. In this work, we present a method to identify and remove artifacts in dermoscopic images. This pre-processing step enhances the output of the segmentation of the lesion. This process is based on several applications of the Local Normalization, which is a method that increases the local contrast between local pixels, improving the overall quality of the image, especially with non-uniform illumination. The process is scale sensitive and uses a multi-scale approach adaptable to every shape and size of skin lesions.

A textured scale-based approach to melanocytic skin lesions in dermoscopy

Melanoma is the most dangerous and lethal form of human skin cancer and the early detection is a fundamental key for its successful management. In recent years the use of automatic classification algorithms in the context of Computer Aided Diagnosis (CAD) systems have been an important tool, by improving quantification metrics and also assisting in the decision regarding lesion management. This paper presents a novel and robust textured-based approach to detect melanomas among melanocytic images obtained by dermoscopy, using Local Binary Pattern Variance (LBPV) histograms after the Bidimensional Empirical Mode Decomposition (BEMD) scale-based decomposition methodology. The results show that it is possible to develop a robust CAD system for the classification of dermoscopy images obtained from different databases and acquired in diverse conditions. After the initial texture-scale based classification a post-processing refinement is proposed using reticular pattern and color achieving to 97.83, 94.44 and 96.00 for Sensitivity, Specificity and Accuracy.

Dermoscopic assisted diagnosis in melanoma: Reviewing results, optimizing methodologies and quantifying empirical guidelines

Abstract 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.