uBibliorum
Digital Repository of the University of Beira Interior
Recent Submissions
Effects of physical exercise concomitant with chemotherapy on cardiac functionality of women with breast cancer
Publication . Antunes, Pedro Miguel da Silva; Esteves, Maria Dulce Leal; Ascensão, António Alexandre Moreira Ribeiro de; Sampaio, Francisco Pedro Morais Dias de Almeida
Anthracycline-containing chemotherapy is commonly used in the treatment of hematologic cancers and solid tumors, including breast cancer. Despite its wellrecognized clinical efficacy, the use of anthracyclines is limited by dose-dependent and cumulative cardiotoxic risk. Additionally, breast cancer treatment is often multimodal, which can further increase the cardiac damage. The cardiac toxicity poses significant challenges, influencing therapeutic decision-making and exacerbate patients’ morbidity and mortality. Longitudinal studies have shown that cardiovascular diseases are one of the leading causes of death among women with breast cancer. As a result, various international agencies, such as the American Society of Clinical Oncology and the European Society of Cardiology, emphasize the importance of implementing primary prevention interventions. Exercise training is an accessible and cost-effective intervention with well-documented benefits for overall health. Preclinical studies have shown promising results in mitigating the cardiotoxic impact of anthracyclines. However, whether this potential cardioprotective effect translates to humans is a question that remains to be clarified. The objectives of this thesis were to conduct a literature review on this topic (Study 1 and 2) and, as the core of this doctoral program, to design a single-center, randomized controlled trial with two arms (exercise group vs. usual care group) (Study 3). The main objective of this trial was to analyze the impact of a supervised exercise training program on established markers of cardiac toxicity (i.e., left ventricular ejection fraction, global longitudinal strain, and cardiac biomarkers) in women with breast cancer undergoing curative anthracycline-containing chemotherapy. Secondary objectives included analyzing the safety of exercise training and its effect on cardiorespiratory fitness, physical functionality, and health-related quality of life. The program consisted of three sessions per week of combined training (i.e., aerobic and resistance training) during each participant's chemotherapy length (approximately 5- months). Through the literature review, supported by a systematic review with metaanalysis of randomized controlled trials (Study 2), we found that evidence on the effect of physical exercise on established parameters of cardiac toxicity was limited, as only four eligible studies were identified. Most of these studies had significant limitations, including designs not specifically aimed at evaluating the efficacy of physical exercise on the parameters of interest, small sample sizes, short training programs, and brief followup periods. The results of the meta-analysis did not show a significant effect of exercise training compared to non-exercise. However, an exploratory analysis, which included studies with training programs consisting of at least 36 sessions (n=3), revealed a superior effect of exercise training on resting left ventricular ejection fraction. This suggests that the length of the exercise training programs might be associated with the effect of exercise training on cardiac function. The results of the randomized controlled trial demonstrated that exercise training did not significantly prevent the decline in established markers of cardiac toxicity compared to control (Study 4). However, it was found that exercising was safe—no serious adverse events were reported during the exercise training sessions—and it significantly prevented the decline in cardiorespiratory fitness, physical functionality, and health-related quality of life (Study 4 and 6). Additionally, we found that the response of cardiorespiratory fitness among participants randomized to the exercise group was heterogeneous, which justified the analysis of potential predictors through an exploratory study (Study 5). The results revealed that the parameters associated with the response in cardiorespiratory fitness to training were baseline body mass index and total aerobic training metabolic equivalent of task per hour (MET-hour). In summary, the results of this thesis suggest that adding exercise training to standard breast cancer care does not provide benefits on established markers of cardiac toxicity in women undergoing anthracycline-containing chemotherapy. However, the findings of experimental works indicate that implementing a supervised exercise training program during chemotherapy is safe and can be a viable and effective approach for improving cardiorespiratory fitness, preventing the decline in physical functionality and healthrelated quality of life in this challenging patient population. Additionally, our study suggests that the total energy expenditure of aerobic exercise (i.e., MET-hour) may be a predictor of cardiorespiratory response.
Coopetition Technology-Enabled Service Ecosystems: A Framework to Improve the Portuguese Ornamental Stone Sector
Publication . Silva, Agostinho Manuel Antunes da; Cardoso, António João Marques
This doctoral thesis critically examines the design, implementation, and effectiveness of coopetition networks among small and medium-sized enterprises (SMEs) within the dynamic and complex realms of digital and global supply chains. The study addresses the challenges that SMEs face, including financial constraints, operational inefficiencies, scalability issues, customer satisfaction demands, and pressures related to ecological sustainability. To address these challenges, the research introduces and empirically tests the Coopetition Network for Value Cocreation (CN-VCC) framework, subsequently operationalized through the Industrial Internet of Things (IIoT) as the Coopetition Technology-enabled Service Ecosystem (CNEcoTech). The Design Science Research (DSR) methodology, integrated with Service-Dominant (S-D) Logic, underpins the framework's development and validation, with a specific focus on the Portuguese ornamental stone sector (OS-PT), an industry of significant economic importance that is yet to be fully challenged by digital transformation and global market integration. Following the DSR stages of designing and developing an artefact, implementing a Proof-ofConcept (PoC), and an Experimental Pilot Project, the thesis assesses the CN-EcoTech artefacts's impact across multiple dimensions: operational efficiency, scale, customer satisfaction, and ecological sustainability. The findings reveal that engagement levels within the CN-EcoTech framework correlate with significant improvements in OS.PT SMEs. Specifically, SMEs fully engaged in the framework demonstrated enhanced deal closures, robust customer relationships, and superior operational outcomes compared to their less-engaged counterparts. This differential highlights the potential of deep integration into technology-based coopetition frameworks to overcome traditional and emerging market challenges. The thesis confirms that the CN-EcoTech framework facilitates long-term collaboration among SMEs and effectively addresses the challenges of digital and global supply chains. It extends beyond the immediate empirical setting of the OS-PT sector, suggesting broader applicability to similar SME-dominated industries facing digital integration challenges. Academically, this research augments the understanding of coopetition dynamics and provides a structured approach to evaluating and implementing technology-enabled service ecosystems. However, more importantly, it offers actionable insights for industry stakeholders and policymakers, empowering them with strategies to enhance SME competitiveness and sustainability in line with the United Nations Sustainable Development Goals (SDGs). This thesis lays a foundation for future research to explore the scalability of the CN-EcoTech framework across diverse sectors. It offers a robust model for enhancing SME resilience and innovation capacity in a globally interconnected economy.
Portable, multi-task, on-the-edge and low-cost computer vision framework based on deep learning: Precision agriculture application
Publication . Assunção, Eduardo Timóteo; Proença, Hugo Pedro Martins Carriço; Gaspar, Pedro Miguel de Figueiredo Dinis Oliveira
Precision agriculture is a new concept that has been introduced worldwide to increase production, reduce labor and ensure efficient management of fertilizers and irrigation processes. Computer vision is an essential component of precision agriculture and plays an important role in many agricultural tasks. It serves as a perceptual tool for the mechanical interface between robots and environments or sensed objects, as well as for many other tasks such as crop yield prediction. Another important consideration is that some vision applications must run on edge devices, which typically have very limited processing power and memory. Therefore, the computer vision models that are to run on edge devices must be optimized to achieve good performance. Due to the significant impact of Deep Learning and the advent of mobile devices with accelerators, there has been increased research in recent years on computer vision for general purpose applications that have the potential to increase the efficiency of precision agriculture tasks. This thesis explore how deep learning models running on edge devices are affected by optimizations, i.e., inference accuracy and inference time. Lightweight models for weed segmentation, peach fruit detection, and fruit disease classification are cases of studies. First, a case study of peach fruit detection with the well-known Faster R-CNN object detector using the breakthrough AlexNet Convolutional Neural Network (CNN) as the image feature extractor is performed. A detection accuracy of 0.90 was achieved using metric Average Precision (AP). The breakthrough AlexNet CNN is not an optimized model for use in mobile devices. To explore a lightweight model, a case study of peach fruit disease classification is next conducted using the MobineNet CNN. The MobileNet was trained on a small dataset of images of healthy, rotten, mouldy, and scabby peach fruit and achieved a performance of 0.96 F1. Lessons learned from this work led to using this model as a baseline CNN for other computer vision applications (e.g., fruit detection and weed segmentation). Next, a study was conducted on robotic weed control using an automated herbicide spot sprayer. The DeepLab semantic segmentation model with the MobileNet backbone was used to segment weeds and determine spatial coordinates for the mechanism. The model was optimized and deployed on the Jetson Nano device and integrated with the robotic vehicle to evaluate real-time performance. An inference time of 0.04 s was achieved, and the results obtained in this work provide insight into how the performance of the semantic segmentation model of plants and weeds degrades when the model is adapted through optimization for operation on edge devices. Finally, to extend the application of lightweight deep learning models and the use of edge devices and accelerators, the Single Shot Detector (SSD) was trained to detect peach fruit in three different varieties and was deployed in a Raspberry Pi device with an integrated Tensor Unity Processor (TPU) accelerator. Some variations of MobileNet as a backbone were explored to investigate the tradeoff between accuracy and inference time. MobileNetV1 yielded the best inference time with 21.01 Frame Per Second (FPS), while MobileDet achieved the best detection accuracy (88.2% AP). In addition, an image dataset of three peach cultivars from Portugal was developed and published. This thesis aims to contribute to future steps in the development of precision agriculture and agricultural robotics, especially when computer vision needs to be processed on small devices.
Fatores de risco genético para adenomas hipofisários: Uma análise nacional, multicêntrica, genética e clínica
Publication . Gaspar, Leonor Isabel Mesquita ; Lemos, Manuel Carlos Loureiro de; Gonçalves, Catarina Inês Nunes Pires
Os adenomas hipofisários representam, aproximadamente, 10-15% do total dos tumores intracranianos. A prevalência destes tumores foi estimada em 1:1000 na população geral, sendo mais frequentemente diagnosticados entre os 40-60 anos de idade. Estes tumores são monoclonais, tipicamente benignos e de crescimento lento, no entanto podem estar associados a um aumento da morbilidade e mortalidade através da sobreprodução hormonal e dos efeitos de massa resultantes da compressão das estruturas adjacentes ao tumor. Os tumores hipofisários mais frequentes são os prolactinomas, seguido pelos adenomas hipofisários não funcionantes. Os mecanismos subjacentes à tumorigénese hipofisária não são ainda totalmente conhecidos, pelo que uma melhor compreensão desta questão ajudará a gerir a doença. O aumento do risco associado a mutações em genes como o AIP, MEN1, CDKN1B e PRKAR1A, fornece evidências de uma predisposição genética para adenomas hipofisários familiares. A grande maioria dos adenomas hipofisários (cerca de 95%) ocorre num contexto esporádico e na ausência de predisposição genética conhecida. No entanto, três polimorfismos (rs2359536, rs10763170 e rs17083838) foram significativamente associados a adenomas hipofisários esporádicos na população Chinesa Han. O objetivo geral desta tese foi realizar um estudo de âmbito multicêntrico nacional, acerca dos fatores de risco genético para o desenvolvimento de adenomas hipofisários familiares e esporádicos, de forma a ampliar o conhecimento sobre a tumorigénese hipofisária. Numa primeira fase desta tese, foi construída uma base de dados com todas as variantes germinativas identificadas no gene AIP publicadas em casos esporádicos e familiares de adenomas hipofisários, até à data, a nível mundial. Nesta revisão, foram identificadas e avaliadas, ao nível da sua patogenicidade, um total de 158 mutações germinativas entre 562 doentes com adenomas hipofisários esporádicos ou familiares. Estas variantes estavam localizadas em toda a região codificadora e nas regiões de splicing do gene AIP. A patogenicidade de todas as variantes germinativas publicadas foi categorizada de acordo com os critérios da American College of Medical Genetic and Genomics (ACMG), utilizando todos os dados disponíveis. Do número total de doentes, 35,4% apresentavam variantes patogénicas e 24,0% apresentavam variantes provavelmente patogénicas. Na segunda fase desta tese foi determinada a frequência de mutações germinativas do gene AIP em doentes portugueses com macroadenomas hipofisários esporádicos de início precoce. Para isso, foi sequenciado o gene AIP em 218 doentes com macroadenomas hipofisários esporádicos diagnosticados antes dos 40 anos. Foram identificadas variantes raras em heterozigotia neste gene em 18 (8,3%) doentes. No entanto, apenas quatro (1,8%) doentes apresentavam variantes patogénicas. Estas variantes compreendiam duas mutações já conhecidas (p.Arg81* e p.Leu115Trpfs*41) e duas mutações novas (p.Ser53Thrfs*36, e p.Glu246*). Estes quatro doentes tinham sido diagnosticados com somatotrofinoma em idades compreendidas entre os 14 e os 25 anos. A frequência de variantes patogénicas no gene AIP em doentes com idade inferior a 30 anos foi de 3,4% e com idade inferior a 18 anos foi de 5%, respetivamente. A frequência de mutações no gene AIP nesta coorte de doentes portugueses foi inferior à de outros estudos. A identificação de novas variantes no gene AIP expande o espetro das causas genéticas dos adenomas hipofisários e pode ajudar a compreender o papel das mutações neste gene nos mecanismos moleculares subjacentes à tumorigénese hipofisária. A terceira fase desta tese consistiu em identificar mutações germinativas num conjunto específico de 29 genes, descritos na literatura como tendo mutações germinativas em doentes com adenomas hipofisários, numa coorte de doentes portugueses diagnosticados com adenomas hipofisários esporádicos de início precoce. Para isso, foi feita a sequenciação completa do exoma em 225 doentes com macroadenomas hipofisários esporádicos diagnosticados até aos 40 anos de idade. Foram identificadas 154 variantes raras em 25 dos 29 genes. Destas foram identificadas três variantes patogénicas e 13 variantes provavelmente patogénicas, nos genes AIP, CDH23, MEN1, MSH2, PMS2, SDHB, TP53 e VHL, em 7,1% dos doentes. Nos doentes diagnosticados com idades inferiores a 30 e 18 anos, a frequência de mutações foi de 9,0% e 12%, respectivamente. Esta é, até à data, a maior análise multigénica de doentes com macroadenomas hipofisários esporádicos de início jovem. Confirmámos que o AIP é o gene mais frequentemente envolvido, mas também descobrimos causas genéticas mais raras de adenomas hipofisários, incluindo a primeira confirmação independente de um papel do gene CDH23. Na última fase desta tese foi avaliada a associação de três polimorfismos comuns próximos dos genes NEBL (rs2359536), PCDH15 (rs10763170) e CDK8 (rs17083838) à suscetibilidade a adenomas hipofisários esporádicos na população portuguesa. Foram determinadas as frequências genotípicas e alélicas de 570 casos e 546 controlos. O alelo minor CDK8 rs17083838 (alelo A) foi significativamente associado a adenomas hipofisários esporádicos. As variantes NEBL rs2359536 e PCDH15 rs10763170 não foram associadas a risco geral para a doença, embora tenha sido observada uma associação significativa entre o alelo minor PCDH15 rs10763170 (alelo T) e somatotrofinomas. Estes resultados sugerem que a variante CDK8 rs17083838, e possivelmente a variante PCDH15 rs10763170, podem aumentar a suscetibilidade a adenomas hipofisários esporádicos na população portuguesa. Concluindo, diferentes estratégias foram desenvolvidas e implementadas, ao longo desta tese, de forma a determinar quais os fatores de risco genético mais associados ao desenvolvimento de adenomas hipofisários esporádicos e familiares. Estes resultados são importantes sob o ponto de vista científico não só para uma melhor compreensão do panorama genético dos adenomas hipofisários, como também abrem portas para novas estratégias de rastreio genético direcionadas, oferecendo conhecimentos fundamentais para a gestão personalizada dos macroadenomas hipofisários de início precoce.
How Can Deep Learning Aid Human Behavior Analysis?
Publication . Roxo, Tiago Filipe Dias dos Santos ; Proença, Hugo Pedro Martins Carriço; Inácio, Pedro Ricardo Morais
With the increase of available surveillance data and robustness of state-of-the-art deep learning models, various recent research topics focus on human biometric assessment, tracking and person re-identification. However, one other area of work not extensively explored that can combine surveillance and visual-based models is assessing human behavior. The lack of work in this topic is not surprising given the inherent difficulties on categorizing human behavior in such conditions, in particular without subject cooperation. Based on the psychology literature, human behavior analysis typically requires controlled experimental environments, with subject cooperation and assessing features via grid-based survey. As such, it is not clear on how deep learning models can aid psychology experts in human behavior analysis, which is where this thesis intents to contribute to the body of knowledge. We extensively review psychology literature to define a set of features that have been proven as influential towards human behavior and that can be assessed via camera in surveillance-like conditions. This way, we define human behavior via subject profiling using seven behavioral features: interaction, relative position, clothing, soft biometrics, subject proximity, pose, and use of handheld devices. Note that this analysis does not categorize human behavior into specific states (e.g. aggressive, depressive) but rather creates a set of features that can be used to profile subjects, usable to aid/complement behavioral experts and to compare behavioral traits between subjects in a scene. Furthermore, to motivate the development of works in these areas, we review state-of-the-art approaches and datasets to highlight the limitation of certain areas and discuss the topics worth exploring for future works. After defining a set of behavioral features, we start by exploring the limitation of current biometric models in surveillance conditions, in particular the resilience of gender inference approaches. We demonstrate that these models underperform in surveillance-like data, using PAR datasets, highlighting the limitations of training in cooperative settings to perform in wilder conditions. Supported by the findings of our initial experiments, complementing face and body information arouse as a viable strategy to increase model robustness in these conditions, which lead us to design and propose a new model for wild gender inference based on this premise. This way, we extend the knowledge of an extensive discussed literature topic (gender classification) by exploring its application in settings where current models do not typically perform (surveillance). We also explore the topic of human interaction, namely Active Speaker Detection, in particular towards more uncooperative scenarios such as surveillance conditions. Contrary to the gender/biometrics topic, this is a lesser explored area where works are mainly based on assessing active speakers via face and audio information in cooperative conditions and with good audio and image quality (movie settings). As such, to clearly demonstrate the limitations of state-of-the-art ASD models we start by creating a wilder ASD dataset (WASD), composed of different categories with increasing challenges towards ASD, namely with audio and image quality degradation, and containing uncooperative subjects. This dataset highlighted the limitations of current models to deal with unconstrained scenarios (e.g. surveillance conditions), while also displaying the importance of body information in conditions where audio quality is subpar and face access is not guaranteed. Following this premise, we design the first model that complements audio, face, and body information to achieve state-of-the-art performance in challenging conditions, in particular surveillance settings. Furthermore, this model also proposed a novel way to combine data via SE blocks, which allowed to provide reasoning behind model’s decision by visual interpretability. The use of SE blocks was also extended to other models and ASD-related areas to highlight the viability of this approach for model-agnostic interpretability. Although this initial model was superior to the state-of-the-art in challenging data, its performance in cooperative settings was not as robust. As such, we develop a new model that simultaneously combines face and body information in visual data extraction which, in conjunction with pretraining in challenging data, leads to state-of-the-art performance in both cooperation and challenging conditions (such as surveillance settings). These works pave a new way to assess human interaction in more challenging data and with model interpretability, serving as baselines for future works.