Computer-assisted analysis of endoscopic imagescan be helpful to the automatic diagnosis and classificationof neoplastic lesions. Barrett's esophagus (BE) is a commontype of reflux that is not straightforward to be detected byendoscopic surveillance, thus being way susceptible to erroneousdiagnosis, which can cause cancer when not treated properly. In this work, we introduce the Optimum-Path Forest (OPF) classifier to the task of automatic identification of Barrett'sesophagus, with promising results and outperforming the wellknownSupport Vector Machines (SVM) in the aforementionedcontext. We consider describing endoscopic images by means offeature extractors based on key point information, such as theSpeeded up Robust Features (SURF) and Scale-Invariant FeatureTransform (SIFT), for further designing a bag-of-visual-wordsthat is used to feed both OPF and SVM classifiers. The bestresults were obtained by means of the OPF classifier for bothfeature extractors, with values lying on 0.732 (SURF) - 0.735(SIFT) for sensitivity, 0.782 (SURF) - 0.806 (SIFT) for specificity, and 0.738 (SURF) - 0.732 (SIFT) for the accuracy.
Approximately 50,000 to 60,000 new cases of Parkinson's disease (PD) are diagnosed yearly. Despite being non-lethal, PD shortens life expectancy of the ones affected with such disease. As such, researchers from different fields of study have put great effort in order to develop methods aiming the identification of PD in its early stages. This work uses handwriting dynamics data acquired by a series of tasks and proposes the application of a deep-driven graph-based clustering algorithm known as Optimum-Path Forest to learn a dictionary-like representation of each individual in order to automatic identify Parkinson's disease. Experimental results have shown promising results, with results comparable to some state-of-the-art approaches in the literature.
Appropriate metrics are paramount for machine learning and pattern recognition. In Content-based Image Retrieval-oriented applications, low-level features and pairwise-distance metrics are usually not capable of representing similarity among the objects as observed by humans. Therefore, metric learning from available data has become crucial in such applications, but just a few related approaches take into account the contextual information inherent from the samples for a better accuracy performance. In this paper, we propose a novel approach which combines an unsupervised manifold learning algorithm with the Optimum-Path Forest (OPF) classifier to obtain more accurate recognition rates, as well as we show it can outperform standard OPF-based classifiers that are trained over the original manifold. Experiments conducted in some public datasets evidenced the validity of metric learning in the context of OPF classifiers.