https://repositorio.ufu.br/handle/123456789/18921 2026-04-18T01:20:57Z https://repositorio.ufu.br/handle/123456789/48484 Title: Classificação de imagens histológicas utilizando algoritmo polinomial de base hermite e regularizador Abstract: Diagnosis support systems are crucial for the analysis of histological images, but their performance depends on the extraction of robust features. Although fractal geometry and eXplainable Artificial Intelligence (XAI) descriptors hold relevance, recent studies point to the superiority of features extracted by deep learning. Therefore, this work proposes a classification system for histological images based on the Hermite Polynomial (HP) algorithm, associated with multi-source descriptors (fractal geometry, deep learning, and XAI). To handle the high dimensionality of the data, the system integrates a LASSO regularizer for feature selection. The computational complexity inherent to HP, in turn, is overcome through a parallel implementation that distributes the training and evaluation process of multiple feature subsets among CPU cores. The experimental evaluation, conducted on binary and multi-class databases, demonstrated the superiority of descriptors extracted from the ResNet-50 (RN50) network in a comparative analysis with fractal and XAI-based approaches. With the RN50 descriptors, the methodology achieved a performance with accuracy (ACC) of up to 100% and an imbalanced accuracy measure (IAM) of 1.00 in binary data classification scenarios. In multi-class databases, the ACC value was higher than 98% and the IAM was 0.89. Additionally, in robustness tests against noise, the proposed system maintained its superiority, presenting significantly higher accuracy than other machine learning algorithms even with 50% noise fin the attributes, in both binary and multi-class scenarios. The results consolidate the combination of the HP classifier with regularization and RN50 descriptors as a high-precision and robust approach, with the potential to assist the decision-making of specialists in clinical practice. 2026-02-06T00:00:00Z https://repositorio.ufu.br/handle/123456789/48048 Title: HESMM - um modelo para o desenvolvimento de pessoas com pensamento crítico de alto nível Abstract: Teaching in the area of Computing has evolved over time, forced in part by frequent innovations in information technology. To meet the growing need for human talent with a high capacity for logical-mathematical reasoning, among some initiatives, Programming Marathons emerged. These are competitions in which participants focus on solving computational challenges using algorithm resolution and implementation techniques. These events, practiced from regional to global scope and experienced by thousands of students, have become very popular in the university community and even in high school because they use a demanding and efficient format based on problems validated by teachers assisted by software robots. With the intention of contributing to the improvement of teaching in the area of Computing, an effort was made to organize, train, participate and collaborate in this movement. It is hoped that the experience gained can support new proposals for education. To generate a result with a high level of consistency, we selected marathon runners with significant results and, using qualitative research as a method, we worked on collecting information and subsequently analyzing the responses through a questionnaire. From this, the premises were outlined and served as the basis for defining a computing teaching model suitable for the new world named HESMM- High-end Shapers and Makers Model of Critical Thinking People. As this work evolved, it became clear that, based on the experience of former competitors who stood out in programming competitions, we can identify routine patterns and generate precepts for proposals to improve current teaching models. The results highlight that the discipline in the study, with adequate planning of schedules and practice in simulators, such as online judge systems, are fundamental for solid learning aligned with contemporary demands. 0008-08-08T00:00:00Z https://repositorio.ufu.br/handle/123456789/47982 Title: Recomendação automática de atividades e rastreamento de competências do sujeito complexo em ambientes virtuais de ensino e aprendizagem Abstract: A persistent challenge in Virtual Learning Environments (VLEs) is the lack of mechanisms to track, over time, higher-order competencies—such as autonomy, cooperation and metacognition—despite the abundance of interaction traces recorded in these platforms. Such tracking can support formative and personalized study guidance. In this context, this thesis develops an automatic pedagogical activity recommender to monitor and foster complex subject competencies in Distance Education. The work was developed in two integrated stages. First, multiple Moodle activities were constructed and refined through competency labeling based on seven complex-subject competencies, supported by quantitative analysis, resulting in an activity×competency matrix validated by educators experienced with Moodle. Second, the M-COMPASS recommender was designed by adapting Dynamic Key-Value Memory Networks (DKVMN), incorporating dynamic memory and attention mechanisms. The main modifications include integrating students’ behavioral signals via SQL queries and introducing competency-level attention, which also grounds a new student model. Experiments conducted in three courses within a Computing Teacher Education program provided consistent evidence: (i) adherence to recommendations is associated with better performance (positive intra-module and overall correlations); (ii) competency-level dose–response patterns, with gains observed both in the short term and cumulatively across modules; and (iii) training stability, indicated by a regular decrease in loss and interpretable individual cases via competency attention and radar charts. The analyses also supported an instrument-improvement cycle (Ridge+LOO diagnostics over SQL queries), reducing redundancies and correcting constant columns. Overall, the thesis contributes a full Moodle→SQL→student model→explainable recommendation pipeline aligned with Complex Thinking Theory, and a DKVMN adaptation for multiple competencies per activity with pedagogical justifications. 2025-12-08T00:00:00Z https://repositorio.ufu.br/handle/123456789/47971 Title: Abordagem computacional para auxiliar a análise e classificação de câncer de próstata por meio de imagens Abstract: The analysis of whole-slide histological images (WSI) in prostate cancer diagnosis presents significant computational challenges, such as gigapixel image processing and tissue morphological variability. This work proposes a methodology that integrates a preprocessing stage based on morphological operations with a convolutional neural network architecture of the Mask R-CNN type for instance segmentation. Applied to the PANDA dataset, the approach included the construction of a curated database with patch generation and class balancing. The model, implemented using the Detectron2 framework, achieved an accuracy of 97.87% in classifying Gleason patterns, confirming that targeted preprocessing enhances network learning compared to end-to-end approaches. Validation was also performed using the Karolinska subset. The results indicate the potential of the proposed methodology as a tool for digital pathology, demonstrating its generalization capability and establishing a new reference for applications addressing challenges in medical imaging. 2025-11-13T00:00:00Z