The exponential growth in medical imaging volumes necessitates scalable, reliable diagnostic support systems capable of augmenting clinical workflows. This paper presents a systematic quantitative evaluation of state-of-the-art Multimodal Large Language Models (MLLMs) for radiology Visual Question Answering (VQA), a task requiring integrated visual perception and clinical reasoning. We benchmark five leading models-GPT4o, Gemini 2.5 Pro, Qwen 2.5-VL (72B), LLaVA 13B, and Llama 3.2 Vision (90B)-on the VQA-RAD dataset under a rigorous zero-shot protocol with standardized prompts and comprehensive precision-recall-F1 evaluation. Our empirical analysis reveals that Gemini 2.5 Pro achieves superior balanced performance (F1$=0.77$, Accuracy$=0.76$, Recall$=0.81$), while Qwen 2.5-VL attains the highest precision (0.87) at the cost of reduced recall, exhibiting conservative prediction characteristics. These outcomes demonstrate that general-purpose MLLMs can perform competitively with specialized medical models in tasks such as modality and organ recognition but still struggle with abnormality detection and complex clinical reasoning. The findings reinforce that MLLMs currently serve best as assistive decision-support tools rather than autonomous diagnostic agents and highlight the potential of retrieval-augmented and context-aware strategies for improving clinical reliability and interpretability.

Procurement fraud brings severe economic and social damage around the World. Losses in revenue reach the magnitude of trillions of dollars. Control bodies around the World expend great efforts in an attempt to reduce such waste. The audit work involves the analysis of procurement by specialists. The high number of processes analyzed and the time required for fraud analysis come in an inefficient and low fraud detection rate. Several computational models have been proposed in recent years to automate the detection and prediction of procurement fraud. However, most of these models depend on human intervention to extract red flags that the machine should consider in detecting fraud. We propose the use of BERT with NLP techniques for the automatic extraction of red flags used in detecting fraud in procurement. Experimental results show that pre-trained contextualized language models are competitive with other methods.

A recurring problem for energy supply companies is the guarantee of the quality of services, which is regulated in many cases. Even so, there are many lawsuits against energy distribution companies, for several reasons, increasing the operating costs of these companies, in many situations with cases that could be resolved via negotiation. This work proposes a method to predict legal judgment outcome regarding the chance of being won or lost by the company. The idea is to understand in which lawsuits more effort should be made to negotiate with the court. The methodology is divided into five stages: feature extraction, sampling with Borderline SMOTE, feature encoding with Target Encoding, classification with XGBoost, and evaluation. The proposed method was evaluated in a database with more than seventy thousand lawsuits, with different outcomes and types, reaching an accuracy of 78.13%, F1 of 74.34%, and AUC of 77.59%.

The task of predicting energy consumption is a problem of great interest in electric power companies. A minimal error prediction is essential for identifying inconsistencies in the monthly consumption reading process. This paper presents a methodology applied to electric consumption prediction and was performed with and without a hyperparameter optimization strategy using a TCN network. We apply these strategies to indi-vidual electric consumption time series. The TCN approach had superior results when compared to SES, ARIMA, and Gradient Boosting. The results show that the proposed process obtained low efficiency with approximately 1% or less improvement than the use of no optimization. However, the TCN itself showed promising results being the best approach in many of our tests.

A recurring problem for energy supply companies is the quality of guarantees service that is regulated in many cases. Nevertheless, there are many lawsuits against energy distribution companies, for several reasons, which increase these companies' operating costs, in many situations with cases that could be resolved through negotiation. Hence, the main aim of this work is to construct an insightful tool for forecasting court outcomes for energy sector litigation focused on building features gathered from the client's historical partnership and key data for litigation utilizing eXtreme Gradient Boosting (XGBoost) as a classifier, TPE as an optimizer and feature engineering. The idea is to understand how lawsuits more effort should be made to conduct a negotiation outside the court. The proposed method is divided into three steps: (1) data acquisition; (2) feature engineering; (3) classification and optimization when evaluated with a dataset of over 70 thousand lawsuits, with 81 different outcomes reaching TOP-3 accuracy of 84.08%.

Dry eye syndrome is one of the most frequently reported eye diseases in ophthalmological practice. The diagnosis of this disease is a challenging task due to its multifactorial etiology. One of the most applied tests is the manual classification of tear film images captured with the Doane interferometer. The interference phenomena in these images can be characterized as texture patterns, which can be automatically classified into one of the following categories: strong fringes, coalescing strong fringes, fine fringes, coalescing fine fringes, and debris. This work presents a method for classifying tear film images based on texture analysis using phylogenetic diversity indexes and Ripley's K function. The proposed method consists of six main steps: acquisition of the image dataset; segmentation of the region of interest; feature extraction using phylogenetic diversity indexes and Ripley's K function; feature selection using Greedy Stepwise; classification using the algorithms Support Vector Machine (SVM), Random Forest (RF), Naive Bayes (NB), Multilayer Perceptron (MLP), Random Tree (RT) and Radial Basis Function Network (RBFNet); and (6) validation of results. The best result, using the RF classifier, we obtained classification rates higher than 99% of accuracy with 0.843% of standard deviation, 0.999 of the area under the Receiver Operating Characteristics (ROC) curve, 0.995 of Kappa and 0.996 of F-Measure. The experimental results demonstrate that the proposed method is promising and can potentially be used by experts to accurately diagnose dry eye syndrome in tear film images.

Lung cancer is the type of cancer that most often kills after the initial diagnosis. To aid the specialist in its diagnosis, temporal evaluation is a potential tool for analyzing indeterminate lesions, which may be benign or malignant, during treatment. With this goal in mind, a methodology is herein proposed for the analysis, quantification, and visualization of changes in lung lesions. This methodology uses a modified version of the quality threshold clustering algorithm to associate each voxel of the lesion to a cluster, and changes in the lesion over time are defined in terms of voxel moves to another cluster. In addition, statistical features are extracted for classification of the lesion as benign or malignant. To develop the proposed methodology, two databases of pulmonary lesions were used, one for malignant lesions in treatment (public) and the other for indeterminate cases (private). We determined that the density change percentage varied from 6.22% to 36.93% of lesion volume in the public database of malignant lesions under treatment and from 19.98% to 38.81% in the private database of lung nodules. Additionally, other inter-cluster density change measures were obtained. These measures indicate the degree of change in the clusters and how each of them is abundant in relation to volume. From the statistical analysis of regions in which the density changes occurred, we were able to discriminate lung lesions with an accuracy of 98.41%, demonstrating that these changes can indicate the true nature of the lesion. In addition to visualizing the density changes occurring in lesions over time, we quantified these changes and analyzed the entire set through volumetry, which is the technique most commonly used to analyze changes in pulmonary lesions.

Urban roads are very complex. The increase in the flow of vehicles in the cities has contributed to traffic accidents. Researches for accident reduction show that the traffic lights are effective in reducing accidents. Traffic lights can minimize the occurrence of accidents at intersections and crosswalks. The implementation of traffic light signals shows significant advantages, otherwise reveals some problems such as the failure to detect road signs by drivers on urban roads. This fact is related to excessive visual information, the stress of the drivers and/or eyestrain, making the drivers lose their attention. These reasons motivated researches about intelligent vehicles. This work aims to develop a methodology to detect and recognize traffic lights, to be applied in smart vehicles. This methodology can contribute to the Advanced Driver Support Systems (ADAS), which assists drivers, especially those with partial vision impairment. Image processing techniques are used to develop the detection methodology. Back project and global thresholding are combined to find light points. Local thresholding techniques are applied to calculate the symmetry between the radius and the center of the light points to segment the traffic light body. The first step got an average rate of 99% of detection. The features of the traffic lights are extracted using Haralick texture measures, with the inclusion of color and shape information. The data generated by the feature extraction step were pre-processed using the SMOTE technique to balance the database. The recognition and identification of the traffic lights state are made by an artificial neural network using Multilayer-Perceptron (MLP). The backpropagation learning algorithm is used in the network training. The validation results show an average recognition rate of 98%.

Dry eye is a common disease that affects the tear film and the ocular surface causing a great variety of symptoms impairing the patient daily activities. The diagnosis of this disease requires a number of tests to evaluate different physiological characteristics. One of the tests captures the appearance of the tear film using the Doane interferometer, which generates images that can be categorized into five groups: strong fringes, coalescing strong fringes, line fringes, coalescing line fringes, and debris. The images can be manually classified by specialists into one of these five groups. The use of automatic systems for the diagnosis of dry eye can assist experts in the classification of these images, contributing to more accurate exams. Therefore, this work presents a method for automatic classification of images from the the lipid layer of the tear film using phylogenetic diversity indexes to extract texture features of the images. The proposed method presents promising results, reaching accuracy rates higher than 96%.

Female breast cancer is a major cause of deaths in occidental countries. Computer-aided Detection (CAD) systems can aid radiologists to increase diagnostic accuracy. In this work, we present a comparison between two classifiers applied to the separation of normal and abnormal breast tissues from mammograms. The purpose of the comparison is to select the best prediction technique to be part of a CAD system. Each region of interest is classified through a Support Vector Machine (SVM) and a Bayesian Neural Network (BNN) as normal or abnormal region. SVM is a machine-learning method, based on the principle of structural risk minimization, which shows good performance when applied to data outside the training set. A Bayesian Neural Network is a classifier that joins traditional neural networks theory and Bayesian inference. We use a set of measures obtained by the application of the semivariogram, semimadogram, covariogram, and correlogram functions to the characterization of breast tissue as normal or abnormal. The results show that SVM presents best performance for the classification of breast tissues in mammographic images. The tests indicate that SVM has more generalization power than the BNN classifier. BNN has a sensibility of 76.19% and a specificity of 79.31%, while SVM presents a sensibility of 74.07% and a specificity of 98.77%. The accuracy rate for tests is 78.70% and 92.59% for BNN and SVM, respectively.

Recently, digital photography is definitely becoming very popular. Thus, there is a large increase in the number of captured photos, which demands better methods to manage this data. An approach is to allow the user to annotate some information in those photos, for instance, the people who appear in them. In this paper, two solutions are proposed in order to assist the user in the process of annotating people in photos, using mobile devices. The first one, called on-line, accesses a server to produce the results and the other one, called off-line, utilizes the mobile device itself to compute the results. Finally, a comparative study between the solutions, using parameters such as hit rate and performance, is presented.