This study proposes to subtype patients with Parkinson's disease (PD) based on speech characteristics using unsupervised learning. The speech of 88 individuals was analyzed with two sets of features: (1) traditional acoustic metrics via Parselmouth (Praat) and (2) embeddings from the self-supervised wav2vec model. Each was clustered with Gaussian Mixture Models, producing two subtypes. Embedding-based clusters showed significant clinical differences in gait, postural stability, freezing, and tremor, with milder cases linked to education, suggesting cognitive reserve effects. Acoustic-based clusters showed limited distinctions. The results indicate that speech embeddings better capture the severity of motor symptoms, supporting speech as an accessible and non-invasive biomarker for stratification, early diagnosis, and personalized treatment in PD.

Parkinson's Disease (PD) is a neurodegenerative disorder marked by diverse motor and non-motor symptoms, driven largely by dopaminergic neuronal loss in the substantia nigra. Traditional approaches to PD treatment have been impeded by the disease's clinical heterogeneity, highlighting the need for more specific subtyping methods. Recent advancements propose utilizing clinical and biological data to uncover biologically relevant PD subtypes, with neuroimaging being a suitable candidate given the disease's impact on the brain. Previous neuroimaging studies on PD subtyping have utilized small datasets and identified age-related subtypes of PD. We address these limitations by leveraging the Parkinson's Progression Markers Initiative dataset and utilizing Gaussian Mixture Models (GMMs) to cluster neuroimaging data, while preserving clinical data solely for validation purposes. The analysis included 747 patients, each characterized by 12 brain volume features obtained through quality-controlled neuroimaging processing pipelines. GMM clustering revealed three distinct PD subtypes with varying severity of motor symptoms but similar non-motor symptoms and cognitive scores. The subtypes identified - mild, intermediate, and severe motor severity - demonstrated significant differences in motor symptom severity, mainly in terms of bradykinesia, rigidity, and tremor. Our approach provides a stratification that could enhance targeted treatment strategies. Future research should investigate the integration of multimodal and longitudinal data to elucidate further the potential of neuroimaging-based subtyping in predicting disease progression and therapeutic outcomes.

Sign language is a visual and gestural communication system used by deaf and hearing-impaired people. Despite numerous deep learning methods proposed for automatic interpretation, a gap persists in developing applications that effectively utilize these models for assisting sign language studies and inclusion. We introduce LERMO (https://lermo.app/), a web game merging machine learning and gamification to enhance sign language fingerspelling. Inspired by Wordle™, LERMO offers an interactive word-guessing game where users can play using a video camera. We create a new dataset of labeled landmark fingerspelling and design our model to ensure optimal speed and efficiency to run on a web browser. We survey approximately 40 users, which find LERMO user-friendly and innovative. From those, 95% believe LERMO could be used to enhance fingerspelling skills.

Recently, machine learning has dominated software engineering research, much of which can be attributed to the success of large language models in handling source code-related tasks. Yet, despite the significant advancements in pre-trained language models for such area, the exploration of their potential in a multi-task learning (MTL) environment remains largely unaddressed. For such, this paper offers a comparative analysis between task-specific and multi-task approaches, focusing on two main tasks: Natural Language Code Search and Unit Test Case Generation. We propose a methodology based on prompt MTL and perform an extensive evaluation, which contrasts the performance of MTL models against their single-task counterparts making use of three pre-trained models across seven datasets. Delving deeper, we conduct an additional exploratory analysis to uncover the underlying causes that explain the observed results, investigating the specificities that govern the application of MTL in this context. Our empirical results sketch a nuanced landscape. MTL does not improve the results when compared to its single-task counterparts. Nevertheless, there are some scenarios for particular models, or when data is scarce, that make MTL and STL achieve quite similar results. The most important finding, however, is that the nature of the pre-training tasks significantly affects the fine-tuning capabilities of MTL, opening space for more guided research on how to pre-train and fine-tune those models.

This study evaluates the potential of the “Segment Anything Model” (SAM) as a robust alternative for medical imaging segmentation in a zero-shot learning context. We evaluate SAM's performance across six diverse medical imaging datasets spanning four different imaging modalities. By employing eight unique prompting strategies we reveal comprehensive insights into SAM's adaptability. The Bounding Box strategy, with its variations, matched or even outperformed existing benchmarks. On the Breast Ultrasound Images dataset, SAM notably outperformed SOTA models, attesting to its capability as a robust zero-shot segmentation tool. Conversely, challenges arose with datasets having indistinct boundaries and inconsistent annotations, as in skin lesion images. The study also establishes a practical set of guidelines aimed at optimizing SAM's clinical usage. The findings underscore SAM's potential as a powerful, versatile tool for medical imaging segmentation, alleviating the burden of manual segmentation and potentially improving ground truth masks for labeling new datasets. With its minimal resource requirements and promising results, SAM represents an exciting advancement in medical imaging analysis.

The COVID-19 pandemic has overwhelmed healthcare systems worldwide. Computed tomography imaging has emerged as an essential tool in diagnosing and monitoring cases, allowing for the detection of pulmonary changes even in the early stages of the disease. This study introduces a radiomics-based predictive model aimed at predicting the need for oxygen support in COVID-19 patients. Utilizing a private dataset collected from a local hospital with 81 patients that underwent two longitudinal chest CT scans, we employed two machine learning algorithms, Random Forest and XGBoost, to analyze radiomic features extracted from lung segmentation. We also explore incorporating clinical features in addition to the radiomic ones and using feature selection techniques to handle the high-dimensionality of the dataset. Our best model achieves an AUC of 0.81 in the test set. Our results indicate that only using radiomic features from the last time point reach a higher performance than with additional data. Therefore, it should be practical to implement a framework to predict the need of oxygen support in a clinical setting, as all the information required by the model comes from a single CT scan.

Kerogen type in source rocks is directly related to its hydrocarbon generation potential. Its determination is often carried out with destructive methods. This study presents a non-destructive technique as an alternative to determine kerogen type using hyperspectral data and machine learning techniques. To present the technique, models were training using Support Vector Machines, K Nearest Neighbors, and Random Forest classifiers on spectral data collected in rock samples acquired from Taubaté Basin, Brazil, of an outcrop with high hydrocarbon generation potential. The models were trained and evaluated using spectral signatures measured with a spectroradiometer and the results were also tested on hyperspectral images of the samples. The experiments described here achieved accuracy above 0.8 with precision and recall above 0.62 and 0.8, respectively, for every kerogen type, indicating the soundness of the classification.

The digital geoscience revolution is modifying the technologies geoscientists use to acquire and process data with the coming of digital outcrop models, hyperspectral data among many methods. These improvements in technology create new challenges in visualization, manipulation, and modeling of the data acquired, opening new research possibilities. In this paper, we present a novel system to visualize, manipulate and correlate geochemical and hyperspectral data, Digital Outcrop Models, and 3D rock samples using state-of-the-art immersive virtual reality techniques. We present a study case using a visualization and data set on an open pit quarry outcrop of a potential analog for hydrocarbon source rocks from Tremembé Formation (Taubaté Basin, Brazil).

This paper presents VizSpectralData, a web based application that runs entirely in the front end and allows spectral data from csv files to be opened, visualized and processed. The system is presented together with the algorithms it implements using real data collected from several carbonate rock samples. It is an open source alternative for simple visualization and processing to proprietary softwares, it is developed in javascript, html and css. It has features to visualize the reflectance, continuum removed spectra, and the derivative of the spectra. It allows to import and export spectral libraries in CSV format.

Very high spatial resolution data seems to reach its maximum for orbital images due to unavoidable atmospheric interactions. At the same time, special hyperspectral cameras are being developed to operate on-board manned or unmanned aircrafts at a fixed optics, which prevents its using for imaging near objects in laboratory conditions. Both limitations can only be surpassed by using super-resolution principles. In this paper, we present a multi-looking approach for enhancing the spatial resolution of images acquired by systems that exhausted their natural ability to provide finer images. The method exploits multiple image takes with controlled spatial differences to produce a higher resolution output. Experiments with static hyperspectral sensor and synthetic data have proven the approach is sound and robust to many applications (e.g., rock samples).

In this paper we propose a new approach to automatically detect and extract fractures as well as estimate the aperture measures from orbital/aerial images. We show, in a first step, the capability of a translation invariant wavelet multiscale decomposition (NDWT) to separate the information related to the fractures from other features in the image. In the second stage, the aperture size (widths of the fractures) in different positions are estimated across scales using curvature analysis. In a third step, the fractures can be automatically extracted using a growing algorithm. Besides the measures of the fracture apertures in an image of Thingvellir in Iceland, we showed the correlation between the fractures of interest extracted automatically and the respective extracted manually were high (0.9) while only 51 % of then were extracted using just curvature analysis and growing algorithm (without NDWT).

In conventional work environments, visualization and integration of correlated data has always been limited by the amount of software windows a regular PC can display. On the other hand, immersive virtual environments present possibilities of interaction that fit scalable and adaptable three-dimensional space, allowing the integration of different types of data, and relating information. This article describes VROffice, an immersive virtual reality office to handle and correlate georeferenced data. VROffice is an immersive and interactive virtual reality office used to handle and correlate georeferenced data, providing access to 2D and 3D elements (i.e rocks, lab samples, organisms) organized as a virtual library. These objects work as data containers ready to store and display data from its location. In the virtual office, data is visualized in a way that allows different forms of interactions, such as observation, free-hand manipulation, inspection of details and object location in a 3D georeferenced space. Users can also organize datasets, perform analysis between different types of objects, combine characteristics, and interpret data in a Geographic Information System (GIS) environment. It enables different forms of interaction with potential to improve insights over observations, instantly turning multivariable datasets into intuitive immersive displays. Different forms of interaction, as well as UI (user interface) and UX (user experience) decisions have been customized to ensure that the experience in using the system is as comfortable as possible. Therefore, in order to demonstrate the possibilities of its use, case studies were created in two different fields of knowledge, geology and biology, which have georeferenced data available.

One of the main difficulties in the inspection of Bridges/Viaducts by observation is inaccessibility or lack of access throughout the structure. Mapping using remote sensors on Unmanned Aerial Vehicles (UAVs) or by means of laser scanning can be an interesting alternative to the engineer as it can enable more detailed analysis and diagnostics. Such mapping techniques also allow the generation of realistic 3D models that can be integrated in Virtual Reality (VR) environments. In this sense, we present the ImSpector, a system that uses realistic 3D models generated by remote sensors embedded in UAVs that implements a virtual and immersive environment for inspections. As a result, the system provides the engineer a tool to carry out field tests directly at the office, ensuring agility, accuracy and safety in bridge and viaduct inspections.

The main goal of this paper was to evaluate the use of a low cost immersive driving simulator to improve the teaching learning process of the Transport Infrastructure undergraduate course. The driving simulator that was developed in a virtual reality environment to assist both the teaching of engineering and the research on road safety. An experiment was conducted in Transport Infrastructure 1 course for Civil Engineering students in a Brazilian university. The students developed a geometric design of a road that was posteriorly modeled in 3D and provided in simulator. Students piloted a vehicle in the immersive simulator in the same road that they designed. Subsequently the usability of the system was assessed by the SUS metric (System Usability Scale). We performed an evaluation with 52 users and the SUS metric that we found was of 73% assuring a degree of usability above average and demonstrating that the immersive system is good to be used as a complementary tool in the learning of transport infrastructure.