Teses de Mestrado ESTiG
Permanent URI for this community
Browse
Browsing Teses de Mestrado ESTiG by Sustainable Development Goals (SDG) "03:Saúde de Qualidade"
Now showing 1 - 10 of 19
Results Per Page
Sort Options
- Aplicação de identificação de episódios de fibrilação atrialPublication . Guerreiro, Nathan Antonio; Teixeira, João Paulo; Dajer, Maria EugeniaAs Doenças Cardiovasculares (DCVs) causam cerca de 18 milhões de mortes por ano, segundo a Organização Mundial da Saúde (OMS). Na Europa, mais de 10 milhões de pessoas são afetadas anualmente, com 3 milhões de óbitos registrados em 2021. No Brasil, são aproximadamente 400 mil mortes por ano, resaltando arritmias cardíacas. A fibrilação atrial (FA), arritmia mais comum, é caracterizada por ritmo cardíaco irregular. Diante desse cenário e alinhado ao papel social do engenheiro e ao Objetivo de Desenvolvimento Sustentável (ODS) “Garantir saúde e bem-estar para todos” da Organização das Nações Unidas (ONU), este trabalho apresenta o desenvolvimento de uma interface gráfica do usuário (GUI) para aquisição e classificação de eletrocardiogramas (ECG). O sistema foi implementado em MATLAB, integrando aquisição em tempo real com a plataforma BITalino c (derivação I, via Bluetooth), detecção dos picos R e classificação automática de episódios de FA com redes neurais LSTM. São extraídas quatro características principais a cada 60 ciclos cardíacos: intervalos RR e entropias de Shannon das ondas T, U e P. Após normalização, essas variáveis compõem os vetores de entrada da rede, que classifica como Outro Ritmo, Ritmo Normal e Ritmo FA. A aplicação permite ainda a visualização dos sinais em tempo real e a geração automática de relatórios em PDF. A validação com sinais da base de dados MIT-BIH Atrial Fibrillation demonstrou que a interface é funcional, e a acurácia de 98,17%, obtida em estudo anterior, evidencia seu potencial como ferramenta auxiliar na análise de ECGs em ambientes clínicos e domiciliares.
- Binary classification of cardiac pathologies using deep learning: a PTB-XL dataset approachPublication . Chaabani, Mohamed Khalil; Teixeira, João Paulo; Slim , Mohamed AymenCardiovascular diseases, including myocardial infarction, remain among the leading causes of mortality worldwide. Timely and accurate diagnosis is critical for effective treatment but often requires labour-intensive manual analysis of clinical-grade electrocardiograms (ECGs). This dissertation proposes a novel deep learning-based approach for binary classification of cardiac pathologies, using the PTB-XL dataset. The final model architecture integrates EfficientNetB3 for spatial feature extraction and a Linformer block to capture long-range dependencies between ECG leads, the results prove its adaptability for ECG image classification tasks. Extensive experimentation and iterative model development were conducted to reach the final design. Early trials involved exploring different hyperparameter tuning like Optuna and Adam optimizer and a wide range of hyperparameter configurations, including different learning rates, dropout rates, batch sizes, and numbers of Linformer layers. These experiments were critical in finding the optimal combination of parameters that balanced computational efficiency and model accuracy. Comprehensive details of these trials and evaluations are provided in the report. The preprocessing pipeline involves selecting the ECGs and converting them to 11 images (aVR was excluded) each representing a lead, converting RGBA ECG images to RGB format and applying normalization to ensure compatibility with model input requirements. This preprocessing step addresses the unique format of the dataset and prepares it for high-performance neural network training. Initial results from the finalized model architecture have demonstrated promising performance, achieving an AUC (Area Under the Curve) of 85.02% and a F1-score of 78.94%. The achieved results are comparable to recent state-of-the-art models reported on the PTB-XL dataset, which typically range between 85% and 95% AUC for similar binary classification tasks. These results indicate that the model's AUC of 85.02% is promising but on the edge of the current state-of-the-art. These findings indicate strong potential for the system to support clinical decision-making by automating the classification of ECG data. Ongoing research aims to extend the current binary classification framework to multi-class scenarios, further enhancing its clinical applicability. Additionally, efforts are being made to improve the model for faster inference times, enabling real-time ECG analysis and improving its feasibility for deployment in healthcare settings.
- Development of cosmetic functional formulations incorporating hyaluronic acidPublication . Silva, Thaís Rossetto Cordeiro da; Barreiro, Filomena; Echart, Arantzazu Santamaria Echart; Düsman, ElisângelaThe cosmetic industry has been expanding continuously, driven by demand for innovative, sustainable, and high-performance formulations. In this context, this work aimed to develop particles composed of hyaluronic acid (HA) and Collagen type I (Col) to act as Pickering stabilisers in cosmetic emulsions. The particles were prepared using different methods (direct mixing - mix and droping - drop), with the most promising dispersions obtained after high-shear homogenisation: P_17_Mix_Ultra and P_17_Drop_Ultra. These formulations exhibited particle sizes of 12.30 ± 3.73 μm and 2.55 ± 0.08 μm, respectively. Both showed highly negative zeta potentials (–37.73 ± 1.50 mV for the mix and –37.72 ± 0.56 mV for the drop), confirming colloidal stability. Wettability tests demonstrated strong oil affinity, with contact angles of 128.30 ± 1.05° and 137.22 ± 3.20°. The drop method stood out for producing smaller, more homogeneous particles. Pickering emulsions were then prepared using sweet almond oil, Miglyol 812, and olive oil. Olive oil originated the most stable systems, and emulsions with a higher oil fraction (70:30 oil-to-aqueous ratio) presented superior stability compared to those with lower oil content. Analyses of emulsion type, colourimetry, creaming index, droplet size, microscopy, zeta potential, and rheology confirmed the better performance of olive oil-based formulations obtained by the drop method. The formulation E_70_OO_17d (drop method, olive oil, 70:30 ratio) achieved the most promising overall results, with a creaming index of 10% after 30 days, homogeneous droplets, and rheological behaviour exhibiting viscoelastic properties comparable to those of commercial products. In conclusion, HA-Col particles proved to be efficient Pickering stabilisers, enabling the development of stable emulsions. The most promising formulation, E_70_OO_17d, was particularly suitable for moisturising and anti-ageing cosmetic applications, where hyaluronic acid provides hydration and regeneration while Col enhances elasticity and repair.
- Development of functional emulsions based on humic acids for cosmetic applicationPublication . Lopes, Nayra Piscoso Saes; Barreiro, Filomena; Santamaria-Echart , Arantzazu; Sipoli , Caroline Casagrande; Junior, Bogdan DemczukThe cosmetics market continues to grow, especially dermocosmetics, which combine aesthetic and therapeutic care. This growth is driving the development of more effective and sustainable formulations, such as Pickering emulsions, which utilize solid particles instead of synthetic surfactants. These latter are responsible for causing environmental impacts and adverse effects on the skin, including irritation and allergies. Humic acid, a natural active ingredient with antioxidant, anti-inflammatory, and photoprotective properties, is an innovative alternative for this type of system. In this context, this work aims to study Pickering emulsions stabilized with humic acid nanoparticles for the development of a dermocosmetic product. Different conditions for producing the particles were analyzed, including the solvent and anti-solvent used, concentration, and pH. The optimized nanoparticles were obtained by acid precipitation, using 0.5 M sodium hydroxide to solubilize the humic acid (final concentration of 10 g/L) and 0.5 M citric acid as the precipitating agent. The optimized particles were used to formulate Pickering emulsions, with an oil volume percentage of 65% showing the best results. The humic acid particles had an average particle size of 73.9 ± 9.7 nm and a three-phase contact angle (particles-water-sweet almond oil) of 65 ± 3.5º, revealing a hydrophilic character. The optimized emulsion consisted of round-shaped droplets with an average size of 44.2 ± 0.3 μm, resulting in an emulsified layer of over 95%, which indicated high stability after 30 days of storage. In addition, the emulsion exhibited a brown color, 14.8% inhibition of antioxidant activity, and, according to rheological analysis, a pseudoplastic, non-Newtonian fluid behavior with gel-like characteristics. It had a smooth texture to the touch, was easy to spread, and had low tackiness. These results demonstrate not only the feasibility of using humic acid particles as stabilizers in Pickering emulsions but also highlight their multifunctional potential for cosmetic applications. This is an unprecedented approach in the field of dermocosmetics, with scientific and technological relevance. Overall, this study presents an innovative, ecological, and effective alternative to the use of synthetic surfactants, offering new perspectives for the development of sustainable cosmetic products based on physical stabilization technologies utilizing nanoparticles.
- Development of functional films based on antimicrobial photodynamic inactivationPublication . Carrasqueiras, Miguel Fernandes; Crugeira, Pedro; Barreiro, FilomenaMicrobial contamination remains a major food safety challenge, requiring innovative and sustainable control strategies beyond traditional thermal or chemical methods. Antimicrobial photodynamic inactivation (aPDI) has emerged as a promising alternative, based on the activation of photosensitizers that induce the formation of reactive oxygen species (ROS) capable of eliminating microbial cells. In this work, κ-carrageenan-based functional films incorporating curcumin solid dispersions at 1% and 5 wt% concentrations, designated CF_1%_1Y and CF_5%_1Y, were evaluated as photoactive systems for food applications. The solid dispersion approach improved the stability, amorphization, and homogeneous distribution of curcumin in the hydrophilic polymeric matrix. The FTIR and DSC analyses confirmed the preservation of the polymer network and the amorphous state of curcumin. The absorption spectrum at 445 nm, characteristic of curcumin, was identified by UV-Vis spectrophotometry. Evaluation of the mechanical properties demonstrated the structural integrity and strength of the CF_5%_1Y film, yielding a tensile modulus of 196.5 ± 15.6 MPa, a tensile strength at break of 3.22 MPa, and a strain at break of 4.03%. In planktonic cells, CF_5%_1Y under LED irradiation (450 ± 10 nm), at an energy density of 100 J/cm², reductions of 99.9% for Escherichia coli and 96.7% for Listeria monocytogenes were achieved, confirming the synergistic action of light, oxygen, and curcumin in ROS generation. Decontamination tests performed on cream cheese and mayonnaise matrices confirmed the antimicrobial properties of aPDI, previously identified in planktonic cells. When using irradiated CF_5%_1Y films, microbial reductions of 98.5% for E. coli and 99.4% for L. monocytogenes were observed in the mayonnaise matrix. In contrast, in cream cheese, reductions reached 91.2% for E. coli and 92.8% for L. monocytogenes. In 15-day preservation tests, samples of mayonnaise and cream cheese with irradiated CF_5%_1Y films maintained the integrity of the food matrices, with constant microbial concentrations of 1.7-1.8 log CFU/g. At the same time, the controls showed increased contamination, reaching approximately 4.3 log CFU/g in mayonnaise and 2.3 log CFU/g in cream cheese. These results demonstrate the effectiveness of functionalization and photoactivity, corroborating the development of stable, efficient, and sustainable films with antimicrobial properties for food decontamination and preservation, offering an eco-friendly path.
- Estrogen removal through adsorption using carbon materials prepared from biomass wastePublication . Exposto, Bruno Marques; Queiroz, A.M.; Ribeiro, António E.; Brito, PauloEndocrine disruptors are class of micropollutants that can influence and deregulate the endocrine system in humans and animals. Endocrine disruptors can consist of natural estrogenic hormones, such as E1 and E2, and synthetic estrogenic hormones, such as EE2, that are not easily removed by conventional treatment processes in water and sewage treatment plants, becoming dangerous emerging pollutants. This work addresses the removal of such compounds from aqueous matrices through adsorption processes onto activated carbon produced from biomass waste. The materials selected were almond shell in natura (ASiN) and cork in natura (CiN). These were activated through carbonization at 550ºC for 1 h, obtaining carbonized almond shell (CAS) and carbonized cork (CC). These materials were characterized by carbonization yield, moisture and ash content, PZC value, quantification of acidic and basic sites and FTIR analysis. For the adsorption assays, the impact of conditions such as adsorbent type, temperature and medium pH was assessed in relation to removal performance. Kinetic and isotherm assays were also performed in batch mode, together with the assessment of the thermodynamic data. The results show that equilibrium time was 16 h for ASiN, 4 h for CAS, 6 h for CiN and 24 h for CC. The adsorption process was regulated through the Elovich model for all adsorbents except for CiN, where the best model was PSO. Moreover, cork-based adsorbents presented higher adsorption capacities and activation through carbonization did not contribute to an increase in estrogen removal. Therefore, CiN was selected as the best material for estrogen removal to the posterior assays. A temperature increase did not affect CiN’s adsorption capacity, and Ea was determined to be 136.15, 80.84 and 146.81 kJ/mol for E2, EE2 and E1, respectively. Moreover, pH change only negatively affects adsorption when in extreme basic media. CiN’s adsorption process has been characterized by the Sheindorf-Rebuhn-Sheintuch multicomponent isotherm for all estrogens, with maximum adsorption capacities of 3.75, 7.23 and 3.82 mg/g achieved for estrogens E2, EE2 and E1, respectively, at 25ºC. Average removal percentage increased with the adsorbent’s dosage until it reached a maximum of 66.1% at 1 g/L. Both the adsorption capacity and removal percentage decreased with the increase in temperature. Thermodynamic data was determined and negative values for ΔGº and ΔHº and positive values for ΔSº were obtained, meaning that adsorption was characterized as spontaneous, exothermic and entropically favourable. It was verified that estrogen adsorption occurred through a combination of chemisorption and physisorption phenomena. Chemisorption occurred mainly through pEDA interactions and hydrogen bridges between estrogen and CiN’s functional groups and aromatic rings, while physisorption occurred mainly through pore filling. pEDA bonding assisted by hydrophobic interactions could be an explanation for the higher adsorption rates verified for EE2 in CiN.
- Explicabilidade em modelos de IA aplicados à seleção de recursos humanosPublication . Neto, Reginaldo Gregório de Souza; Teixeira, João Paulo; Foleis, Juliano HenriqueEste trabalho tem como objetivo propor um sistema de apoio à decisão para o processo de recrutamento e seleção de candidatos ao cargo de consultor de vendas, por meio da predição automatizada de notas atribuídas a currículos. Foram empregados algoritmos de aprendizado de máquina supervisionado, incluindo KNN, Random Forest, SVM e redes neurais MLP, para prever a avaliação de um recrutador humano a partir de 14 atributos técnicos extraídos de currículos reais. Como diferencial, este estudo incorporou técnicas de explicabilidade como SHAP, LIME e TreeInterpreter, promovendo a explicabilidade na análise preditiva, possibilitando identificar a importância de cada variável tanto local quanto globalmente. A base de dados foi padronizada, normalizada e complementada com dados sintéticos a fim de mitigar desequilíbrios. A avaliação dos modelos foi conduzida com base em métricas como MAE, MSE, RMSE e R2, além da análise de resíduos e da matriz de confusão. O melhor desempenho entre as redes neurais foi obtido com a MLP (23-12-6-1), treinada com 50% de dados sintéticos, que alcançou MAE de 0,23, MSE de 0,28, RMSE de 0,53 e R2 = 0,97. No entanto, o modelo com maior desempenho geral foi a Random Forest com 1000 árvores, que atingiu MAE de 0,14, MSE de 0,16, RMSE de 0,40 e R2 = 0,98.
- Fire behaviour of composite slabs with steel deck and mineral woolPublication . Oliveira, Israel Mateus Melo; Piloto, P.A.G.; Balsa, Carlos; Alvarenga, LucianaEste trabalho investiga o comportamento térmico de lajes mistas com chapa de aço colaborante (steel deck) submetidas a situação de incêndio, com foco na influência da temperatura sobre o momento fletor positivo. A pesquisa tem como principal objetivo desenvolver uma fórmula simplificada que permita estimar, com precisão, a temperatura média nos elementos que compõem a seção transversal da laje – como o perfil metálico e a armadura, considerando a inexistência de diretrizes normativas que tratem diretamente desse aspecto para estruturas mistas. A avaliação precisa dessas temperaturas é essencial para possibilitar as análises de resistência à flexão em situação de incêndio, uma vez que a degradação térmica dos materiais altera significativamente a capacidade resistente da laje. A metodologia adotada fundamentou-se em simulações térmicas tridimensionais conduzidas no software ANSYS, as quais foram validadas com base em dados experimentais da literatura. A validação demonstrou boa concordância entre os modelos numéricos e os resultados de referência, com erros médios quadráticos (RMSE) aceitáveis para os diferentes pontos analisados. Em seguida, foram realizados 72 estudos paramétricos, nos quais se variaram a espessura do concreto, a espessura da lã de rocha, o diâmetro da armadura e o tempo de exposição ao fogo. A partir desses resultados, foi desenvolvida uma equação linear ajustada por regressão, capaz de prever a temperatura média nos componentes com elevada precisão. Comparada aos dados obtidos no ANSYS, a fórmula proposta apresentou RMSE inferior a 5% em todos os casos analisados, evidenciando sua eficácia como ferramenta auxiliar no dimensionamento térmico de lajes mistas em situação de incêndio.
- Green extraction of artemisinin from Artemisia annua L. and evaluation of the antimalarial activityPublication . Corso, Luan Barichello; Martins, Mónia; Pinho, Simão; Ferreira, Olga; Zuber, AndréMalaria remains one of the greatest global public health challenges, affecting millions of people, especially in tropical and subtropical regions. One of the main obstacles to treating the disease is the low water solubility of antimalarial compounds, which compromises their bioavailability. In this work, terpenes are proposed as alternative solvents and pharmaceutical excipients for processing and formulating these drugs, aiming for higher sustainability and safety in line with green chemistry principles. First, the COSMO-RS model was used to investigate approximately 8000 systems that combine solvents from different classes with antimalarial drugs (artemisinin, quinine, quinidine, tetracycline, artemether, dapsone, and pyrimethamine). Focusing on artemisinin and the evaluated terpene candidates, thymol and its mixtures with α-pinene, β-pinene, γ-terpinene, and p-cymene showed the greatest potential, and were selected as solvents to extract artemisinin from the plant Artemisia annua L. Heat extractions with magnetic stirring were performed using a carousel system under the conditions: 50 °C, 600 rpm, 1 hour, and 1:10 solid-liquid ratio. Besides artemisinin (ART), dihydroartemisinic acid (DHAA) was also identified in significant amounts. The extraction yields using conventional solvents were 1.15 mg ART/gplant and 5.53 mg DHAA/gplant for water and 7.11 mg ART/gplant and 9.42 mg DHAA/gplant for ethanol. Higher global yields were obtained using pure terpenes (α-pinene and β-pinene), and all the equimolar thymol mixtures. The highest values were 10.43 mg ART/gplant for the thymol:α-pinene mixture and 11.86 mg DHAA/gplant for α-pinene. These values are consistent with the maximum amount available in the plant (11.2 ± 0.8 mg ART/gplant, 13.4 ± 1.0 mg DHAA/gplant), obtained by performing five consecutive extraction cycles with an Accelerated Solvent Extractor (100 °C, 5 min, 1:10 S/L ratio). A selected set of extracts (water, dichloromethane, ethanol, p-cymene, γ-terpinene, α-pinene, β-pinene) was subject to in vitro antimalarial activity assays against Plasmodium falciparum (strain 3D7-GFP). All extracts obtained with terpenes were proven to be significant inhibitors, with IC50 values ranging from 5.19 nM (γ-terpinene) to 8.17 nM (α-pinene), close to the value of the standard artemisinin (IC50 = 4.20 nM). Terpenes maintained high activity, highlighting them and reinforcing their predictions as a green alternative for the extraction and formulation of antimalarial compounds.
- Green solvents and AI-driven models for recycling 3D printing wastePublication . Costa, Samuel Felipe Martins; Abranches , Dinis Oliveira; Ferreira , Olga; Patrício, Patrícia Santiago de OliveiraThe increasing adoption of 3D printing, particularly using polylactic acid (PLA), has led to a significant rise in plastic waste, calling for the development of sustainable recycling solutions. Among recycling approaches, the physical method is gaining more space, especially with advances in the use of green solvents. Therefore, this study examines the application of green solvents and artificial intelligence (AI)-driven models for the dissolution and recovery of PLA from 3D printing waste. In particular, the research focuses on identifying environmentally friendly solvents, based on qualitative PLA dissolution data, using machine learning (ML) techniques to find and predict the best solvents for dissolving PLA while minimizing contamination from additives and other polymers. Among the solvents initially investigated, dimethylformamide (DMF), chloroform (CLFM), dimethyl carbonate (DC), and isosorbide dimethyl (IDE) achieved complete dissolution of PLA after 24 h at 50 °C. Dissolution behavior was further examined above and below the PLA glass transition temperature (Tg = 55 - 60 °C), with only ethyl acetate (EtAce) changing from a poor solvent to a good solvent with increasing temperature. The Hansen Solubility Parameters (HSP) and the infinite dilution activity coefficients (γ∞) predicted by COSMO-RS were employed to rationalize the dissolution behavior, showing unsatisfactory discrimination between good and poor solvents. Subsequently, ML models were applied to the experimental dataset to identify additional suitable solvents. The results demonstrated excellent predictive performance, correctly classifying good and poor solvents for PLA and identifying new good solvents as acetonitrile (ACN), methyl acetate (MeAce), and dichloromethane (DCM). Overall, by integrating solvent-based recycling with AI-driven optimization, this work showed potential solvents to enhance the circular economy of PLA-based materials, promoting more sustainable and effective waste management practices.