7 research outputs found
A functional microsatellite of the macrophage migration inhibitory factor gene associated with meningococcal disease
No full textMacrophage migration inhibitory factor (MIF) is an abundantly expressed proinflammatory cytokine playing a critical role in innate immunity and sepsis and other inflammatory diseases. We examined whether functional MIF gene polymorphisms (-794 CATT(5-8) microsatellite and -173 G/C SNP) were associated with the occurrence and outcome of meningococcal disease in children. The CATT(5) allele was associated with the probability of death predicted by the Pediatric Index of Mortality 2 (P=0.001), which increased in correlation with the CATT(5) copy number (P=0.04). The CATT(5) allele, but not the -173 G/C alleles, was also associated with the actual mortality from meningoccal sepsis [OR 2.72 (1.2-6.4), P=0.02]. A family-based association test (i.e., transmission disequilibrium test) performed in 240 trios with 1 afflicted offspring indicated that CATT(5) was a protective allele (P=0.02) for the occurrence of meningococcal disease. At baseline and after stimulation with Neisseria meningitidis in THP-1 monocytic cells or in a whole-blood assay, CATT(5) was found to be a low-expression MIF allele (P=0.005 and P=0.04 for transcriptional activity; P=0.09 and P=0.09 for MIF production). Taken together, these data suggest that polymorphisms of the MIF gene affecting MIF expression are associated with the occurrence, severity, and outcome of meningococcal disease in children.-Renner, P., Roger, T., Bochud, P.-Y., Sprong, T., Sweep, F. C. G. J., Bochud, M., Faust, S. N., Haralambous, E., Betts, H., Chanson, A.-L., Reymond, M. K., Mermel, E., Erard, V., van Deuren, M., Read, R. C., Levin, M., Calandra, T. A functional microsatellite of the macrophage migration inhibitory factor gene associated with meningococcal disease
In vitro activity of daptomycin and vancomycin lock solutions on staphylococcal biofilms in a central venous catheter model
No full textBackground. Catheter lock solutions are used for prevention and management of catheter-related bloodstream infections. We investigated the activity of daptomycin and vancomycin lock solutions against Staphylococcus aureus and Staphylococcus epidermidis in an in vitro central venous catheter (CVC) model. Methods. Biofilm-producing reference strains of S. aureus and S. epidermidis were evaluated. After 24 h of bacterial growth in a CVC model, daptomycin and vancomycin bactericidal activity (+/- preservative-containing heparin sodium) were separately evaluated as a lock solution using 0.5, 1 and 35 mg/ml. Calcium carbonate (50 mg/l) was added to all lock solutions containing daptomycin. Each CVC was drained, flushed and sonicated at 72 h to assess CFU/ml. Results. After 72 h of exposure in the catheter lock solutions, daptomycin and vancomycin at 0.5, 1 and 5 mg/ml demonstrated bactericidal activity (\u3e3.0 log10 CFU/ml) against S. aureus and S. epidermidis (P≤0.001). Heparin lock solution alone produced a non-significant reduction in S. aureus and S. epidermidis (1.92±0.07 and 1.65±0.03 log10 CFU/ml, respectively). Daptomycin 5 mg/ml lock solution +/- heparin eradicated (limit of detection 2.0 log10 CFU/ml) S. epidermidis at 72 h as did the vancomycin 5 mg/ml plus heparin. S. aureus was only eradicated from the daptomycin 5 mg/ml catheter lock-solution. Conclusions. Our CVC model demonstrated that 72 h of exposure to 5 mg/ml lock solutions of daptomycin (plus calcium), +/- heparin or 5 mg/ml of vancomycin plus heparin demonstrate promise in treating catheter infections with biofilm-producing S. epidermidis. Similarly, 5 mg/ml of daptomycin (plus calcium) as a lock solution shows great promise in treating S. aureus catheter infections. © The Author [2007]. Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved
Management of candidaemia and invasive candidiasis in critically ill patients
No full textCritically ill patients in the intensive care unit (ICU) are at increased risk of encountering bloodstream infections (BSIs) with Candida spp., associated with an elevated crude mortality rate. This supports the significance of early detection of infection and identification of the most effective management approach. A review of the various antifungal treatments and an evaluation of the diverse management approaches for invasive candidiasis in critically ill patients is necessary for guiding evidence-based decision-making. Different early detection schemes for invasive candidiasis are well documented in the literature. Other than the common use of blood cultures, new methods entail the use of risk prediction scores and biomarker tests. Regarding management strategies, different options are currently supported. These include prophylaxis, empirical therapy, pre-emptive therapy, and treatment of culture-documented infections. The choice of treatment is greatly dependent on several factors related to the patient and-or to the surrounding environment. Attention needs to be given to previous exposure to azoles, epidemiological data on dominant Candida spp. in local ICUs, severity of illness and associated morbidities. This paper summarises the most recent literature as well as the guidelines issued by the Infectious Diseases Society of America. The objective is to identify the best diagnosis and management approaches for serious Candida infections in critically ill patients. In addition, this article addresses an important aspect associated with managing candidaemia in critically ill patients pertaining to the decision for intravenous catheter removal. © 2013 Elsevier B.V. and the International Society of Chemotherapy.AbeleHorn M, 1996, INFECTION, V24, P426, DOI 10.1007-BF01713042; Anaissie EJ, 1996, CLIN INFECT DIS, V23, P964; Andes DR, 2012, CLIN INFECT DIS, V54, P1110, DOI 10.1093-cid-cis021; Arendrup MC, 2011, J CLIN MICROBIOL, V49, P325, DOI 10.1128-JCM.01811-10; Arendrup MC, 2008, CLIN MICROBIOL INFEC, V14, P487, DOI 10.1111-j.1469-0691.2008.01954.x; Avni T, 2011, J CLIN MICROBIOL, V49, P665, DOI 10.1128-JCM.01602-10; Bassetti M, 2010, CRIT CARE, V14, DOI 10.1186-cc9239; Bennett JE, 2006, NEW ENGL J MED, V355, P1154, DOI 10.1056-NEJMct060052; Blumberg HM, 2001, CLIN INFECT DIS, V33, P177, DOI 10.1086-321811; Bougnoux ME, 2008, INTENS CARE MED, V34, P292, DOI 10.1007-s00134-007-0865-y; Brass EP, 2010, CLIN INFECT DIS, V51, P304, DOI 10.1086-653936; Calandra T, 2004, CLIN INFECT DIS, V39, pS185, DOI 10.1086-421955; Cruciani M, 2005, INTENS CARE MED, V31, P1479, DOI 10.1007-s00134-005-2794-y; De Waele JJ, 2003, CLIN INFECT DIS, V37, P208, DOI 10.1086-375603; Diekema DJ, 2002, J CLIN MICROBIOL, V40, P1298, DOI 10.1128-JCM.40.4.1298-1302.2002; Driessen M, 1996, PEDIATR INFECT DIS J, V15, P1107, DOI 10.1097-00006454-199612000-00011; Dupont H, 2003, CRIT CARE MED, V31, P752, DOI 10.1097-01.CCM.0000053525.49267.77; Ellis M, 2009, J MED MICROBIOL, V58, P606, DOI 10.1099-jmm.0.006452-0; EORTC International Antimicrobial Therapy Cooperative Group, 1989, AM J MED, V86, P668; Gafter-Gvili A, 2008, MAYO CLIN PROC, V83, P1011, DOI 10.4065-83.9.1011; Garbino J, 2002, INTENS CARE MED, V28, P1708, DOI 10.1007-s00134-002-1540-y; Garnacho-Montero J, 2003, CRIT CARE MED, V31, P2742, DOI 10.1097-01.CCM.0000098031.24329.10; Gudlaugsson O, 2003, CLIN INFECT DIS, V37, P1172, DOI 10.1086-378745; Hachem R, 2008, CANCER, V112, P2493, DOI 10.1002-cncr.23466; Horn DL, 2009, CLIN INFECT DIS, V48, P1695, DOI 10.1086-599039; Ibrahim EH, 2000, CHEST, V118, P146, DOI 10.1378-chest.118.1.146; Karageorgopoulos DE, 2011, CLIN INFECT DIS, V52, P750, DOI 10.1093-cid-ciq206; Kett DH, 2011, CRIT CARE MED, V39, P665, DOI 10.1097-CCM.0b013e318206c1ca; KUJATH P, 1993, INFECTION, V21, P376, DOI 10.1007-BF01728917; Kullberg BJ, 2005, LANCET, V366, P1435, DOI 10.1016-S0140-6736(05)67490-9; Kuse ER, 2007, LANCET, V369, P1519, DOI 10.1016-S0140-6736(07)60605-9; Lamaris GA, 2008, EUR J CLIN MICROBIOL, V27, P343, DOI 10.1007-s10096-007-0443-9; Leon C, 2006, CRIT CARE MED, V34, P730, DOI 10.1097-01.CCM.0000202208.37364.7D; Leroy O, 2009, CRIT CARE MED, V37, P1612, DOI 10.1097-CCM.0b013e31819efac0; Lipsett PA, 2004, CLIN INFECT DIS, V39, pS193, DOI 10.1086-421956; Martin GS, 2003, NEW ENGL J MED, V348, P1546, DOI 10.1056-NEJMoa022139; Mermel LA, 2009, CLIN INFECT DIS, V49, P1, DOI 10.1086-599376; Mills Edward J, 2009, Ann Clin Microbiol Antimicrob, V8, P23, DOI 10.1186-1476-0711-8-23; Mora-Duarte J, 2002, NEW ENGL J MED, V347, P2020, DOI 10.1056-NEJMoa021585; Morgan J, 2005, INFECT CONT HOSP EP, V26, P540, DOI 10.1086-502581; NollaSalas J, 1997, INTENS CARE MED, V23, P23, DOI 10.1007-s001340050286; Nucci M, 2010, CLIN INFECT DIS, V51, P295, DOI 10.1086-653935; Ostrosky-Zeichner L, 2007, EUR J CLIN MICROBIOL, V26, P271, DOI 10.1007-s10096-007-0270-z; Ostrosky-Zeichner L, 2003, ANTIMICROB AGENTS CH, V47, P3149, DOI 10.1128-AAC.47.10.3149-3154.2003; Paphitou NI, 2005, MED MYCOL, V43, P235, DOI 10.1080-13693780410001731619; Pappas PG, 2004, CLIN INFECT DIS, V38, P161, DOI 10.1086-380796; Pappas PG, 2007, CLIN INFECT DIS, V45, P883, DOI 10.1086-520980; Pappas PG, 2009, CLIN INFECT DIS, V48, P503, DOI 10.1086-596757; Pereira GH, 2010, MED MYCOL, V48, P839, DOI 10.3109-13693780903580121; Pfaller MA, 2011, J CLIN MICROBIOL, V49, P396, DOI 10.1128-JCM.01398-10; Phillips P, 1997, EUR J CLIN MICROBIOL, V16, P337, DOI 10.1007-BF01726360; Piarroux R, 2004, CRIT CARE MED, V32, P2443, DOI 10.1097-01.CCM.0000147726.62304.7F; PITTET D, 1994, ANN SURG, V220, P751, DOI 10.1097-00000658-199412000-00008; PIZZO PA, 1982, AM J MED, V72, P101, DOI 10.1016-0002-9343(82)90594-0; Playford EG, 2006, J ANTIMICROB CHEMOTH, V57, P628, DOI 10.1093-jac-dkl491; Prella M, 2005, DIAGN MICR INFEC DIS, V51, P95, DOI 10.1016-j.diagmicrobio.2004.08.015; Ramage G, 2002, ANTIMICROB AGENTS CH, V46, P3634, DOI 10.1128-AAC.46.11.3634-3636.2002; Reboli AC, 2007, NEW ENGL J MED, V356, P2472, DOI 10.1056-NEJMoa066906; Rex JH, 2003, CLIN INFECT DIS, V36, P1221, DOI 10.1086-374850; REX JH, 1994, NEW ENGL J MED, V331, P1325, DOI 10.1056-NEJM199411173312001; Shorr AF, 2005, CRIT CARE MED, V33, P1928, DOI 10.1097-01.CCM.0000178352.14703.49; Tortorano AM, 2006, INT J ANTIMICROB AG, V27, P359, DOI 10.1016-j.ijantimicag.2006.01.002; Vincent JL, 2009, JAMA-J AM MED ASSOC, V302, P2323, DOI 10.1001-jama.2009.1754; Walsh TJ, 2002, CLIN INFECT DIS, V34, P600, DOI 10.1086-338715; Walsh TJ, 2002, NEW ENGL J MED, V347, P2070, DOI 10.1056-NEJMe020142; WEY SB, 1989, ARCH INTERN MED, V149, P2349, DOI 10.1001-archinte.149.10.2349; Wisplinghoff H, 2004, CLIN INFECT DIS, V39, P309, DOI 10.1086-421946; Zaoutis TE, 2005, CLIN INFECT DIS, V41, P1232, DOI 10.1086-49692212
Clinical and microbiological characterization of bloodstream infections in cancer patients in a health care institution in Barranquilla (Colombia), 2023
No full textIntroducción: Las infecciones del torrente sanguíneo (IS) son una importante causa de enfermedad y muerte en pacientes oncológicos. Los principales factores de riesgo para desarrollar infección del torrente sanguíneo en este tipo de pacientes son la neutropenia, quimioterapia y uso de catéter intravascular. Se adquieren con mayor frecuencia en el ámbito intrahospitalario y en la mayoría no se logra determinar un foco primario de la infección. Los pacientes oncológicos cursan con un riesgo de aproximadamente 20%, de ser colonizados por gérmenes multirresistentes.
Objetivo: Describir las características clínicas y microbiológicas de pacientes con cáncer e infecciones del torrente sanguíneo hospitalizados en unidad de cuidados intensivos de una institución de salud de Barranquilla (Colombia), durante el período 2023.
Métodos: Estudio descriptivo, de corte transversal. Se evaluó la normalidad de los datos a través de la prueba de Kolmogórov-Smirnov. Se utilizaron frecuencias absolutas y relativas para describir las variables categóricas. Para el análisis de las variables categóricas, se utilizó la prueba de Chi-cuadrado o el test exacto de Fisher. Se utilizó un modelo de regresión logística multivariado con el fin de identificar posibles factores de riesgo y factores protectores asociados a mortalidad. Se consideró un valor p <0.05 para significancia estadística. El software estadístico utilizado fue R-CRAN versión 4.3.0.
Resultados: Se observó una mayor prevalencia de Klebsiella spp. en IS primarias y de microorganismos resistentes en IS secundarias. En IS secundarias, se observó una mayor prevalencia de microorganismos potencialmente productores de carbapenemasas (33%) en comparación con IS primarias (17%). El mecanismo de resistencia AMPc y los microorganismos meticilino resistentes fueron más frecuentes en IS secundarias (19% vs 2.4%) y (15% vs 7.3%), respectivamente. Se encontró que factores como la estancia en UCI de menos de 15 días, la neutropenia y la presencia de mieloma múltiple se asociaron con un mayor riesgo de mortalidad.
Conclusión: En la población con cáncer e IS analizada existe mayor aislamiento por microorganismos resistentes, principalmente Klebsiella spp., la neutropenia y el mieloma múltiple se asociaron a mayor riesgo de mortalidad.Universidad Libre Seccional Barranquilla -- Facultad de Ciencias de la Salud y Exactas y Naturales -- Especialización en Medicina InternaIntroduction: Blood stream infections (BSI) are a major cause of illness and death in cancer patients. The main risk factors for developing bloodstream infection in this type of patient are neutropenia, chemotherapy, and intravascular catheter use. They are most often acquired in the hospital and in most cases the primary focus of infection cannot be determined. Oncological patients are at approximately 20% risk of being colonized by multi-resistant germs.
Objective: Describe the clinical and microbiological characteristics of patients with cancer and bloodstream infections hospitalized in an intensive care unit of a health institution in Barranquilla, Colombia, during the period 2023.
Methods: Descriptive study, cross-cutting. The normality of the data was evaluated using the Kolmogorov-Smirnov test. Absolute and relative frequencies were used to describe categorical variables. For the analysis of the categorical variables, the Chi- square test or the exact Fisher test was used. A multivariate logistic regression model was used to identify possible risk factors and protective factors associated with Mortality. A p value <0.05 was considered for statistical significance. The statistical software used was R-CRAN version 4.3.0
Results: An increased prevalence of Klebsiella spp. was observed in primary BSI and of resistant microorganisms in secondary BSI. In secondary BSI, a higher prevalence of potentially carbapenemase-producing microorganisms (33%) compared to primary BSI (17%) was observed. The AMPc resistance mechanism, and methicillin-resistant microorganisms were more common in secondary BSIs (19% vs. 2.4%) and (15% vs. 7.3%), respectively. Factors such as UCI stay of less than 15 days, neutropenia, and the presence of multiple myeloma were found to be associated with an increased risk of mortality.
Conclusion: In the population with cancer and IS analyzed, there is a greater isolation of resistant microorganisms, mainly Klebsiella spp., neutropenia and multiple myeloma were associated with a higher risk of mortality
Этиотропная терапия COVID-19: критический анализ и перспективы
Get PDFThe COVID-19 outbreak started in December 2019 in China has spread over all countries of the world within few month acquiring a pandemic nature, the incident population counting millions. The pathogenic mechanisms of the new coronaviral infection caused by never-before-seen virus SARS-CoV2 are yet to be studied. Various drugs are used for COVID-19 treatment and guidelines are continuously revised as new experience is acquired. In the current pandemic situation, it is important to provide specialists with latest information concerning efficacy and safety drugs for COVID-19 patients and promising research in this field.The purpose of the review is to critically analyze published data on outcomes of COVID-19 treatment with various drugs including potentially promising drugs.The search has been carried out through such databases as PubMed, Scopus, Cyberleninka, https://www.globalclinicaltrialsdata.com, https://clinicaltrials.gov, Cochrane Library; mostly, randomized clinical trials-2020 and papers dedicated to candidate drugs have been considered. The paper is structured based on the drug’s action mechanism and contains parts dedicated to antiviral, immunomodulatory, and antibacterial therapies. Looking for a new promising target in COVID-19 treatment, the authors focus their attention on matrix metalloproteinases (MMP), which abundance results in the destruction of extracellular matrix, epithelial and endothelial basal membranes and leads to secondary lung tissue injury. The paper provides a theoretic justification of MMP inhibitor use by an example of doxycycline and offers an efficacy study protocol for the new approach to COVID-19 therapy.Conclusion: as of now, there are no drugs which efficacy for COVID 19 has been proven. Drugs possessing multiple mechanisms of action are employed beside their specified indications, often in combinations; in this situation, additive side effects with adverse consequences for the patient can hardly be avoided. Administration of drugs with unproven efficacy may be justified only in clinical trials followed by subsequent analysis and publication of findings demonstrating that in case of success, recommendations for a majority of COVID-19 patients could be confidently issued.Эпидемия COVID-19, начавшаяся в декабре 2019 года в Китае, за несколько месяцев распространилась на все страны мира, приняв характер пандемии, число заболевших исчисляется миллионами. Механизмы патогенеза новой коронавирусной инфекции, вызванной неизвестным ранее вирусом SARS-CoV2, остаются недостаточно изученными. Для лечения COVID-19 применяют препараты разных групп, по мере появления опыта рекомендации регулярно пересматриваются. В условиях текущей пандемии важно предоставить специалистам актуальную информацию об эффективности и безопасности лечебных препаратов, применяемых для лечения пациентов с COVID-19, и о перспективных исследованиях в этой области. Цель обзора — критический анализ опубликованных результатов лечения COVID-19 с использованием различных групп препаратов для выбора наиболее перспективных лекарственных средств.Поиск источников провели по базам данных PubMed, Scopus, Cyberleninka, Clinical Trials, Cochrane Library и др., рассматривали преимущественно рандомизированные клинические исследования 2020 года, а также работы по изучению препаратов-претендентов. Материал статьи структурирован по механизму действия препаратов, содержит разделы противовирусной, иммуномодулирующей, антибактериальной терапии. В поиске новой перспективной мишени в лечении COVID-19 концентрировали внимание на матриксных металлопротеиназах (ММР), избыток которых ведет к разрушению внеклеточного матрикса, базальных мембран эпителия и эндотелия, способствует вторичному повреждению легочной ткани. В работе теоретически обосновали применение ингибиторов MMP на примере доксициклина, предложили протокол исследования для оценки эффективности нового подхода к лечению COVID-19.Заключение. Лекарственных средств с доказанной эффективностью в отношении COVID 19 в настоящее время нет. Препараты с разными механизмами действия применяются не по показаниям, часто в комбинациях, в этих условиях трудно избежать суммирования побочных эффектов с неблагоприятными последствиями для пациента. Применение препаратов с недоказанной эффективностью оправдано лишь в рамках клинических исследований с последующим анализом и публикацией результатов, чтобы в случае успеха с уверенностью рекомендовать их большинству пациентов с COVID-19. Ключевые слова: COVID-19; противомалярийные средства; ингибиторы вирусных протеаз; противопаразитарные препараты; ингибиторы интерлейкинов; ингибиторы янус-киназ; интерфероны; плазма реконвалесцентов; кортикостероиды; прокальцитонин; антибиотики; новая мишень; матриксные металлопротеиназы, доксициклин
Regresión y aprendizaje multimodal como ayuda al diagnóstico en oftalmologı́a e histopatologı́a
No full textilustraciones, diagramasThe main contribution of this thesis is the development of probabilistic machine learning models to support disease diagnosis from medical data sources. We show how a probabilistic approach offers great versatility in exploiting all available information about the target task. Based on the mathematical formalism of quantum mechanics, we develop and apply machine learning models that allow us to handle the flow of information using density matrices in different ways. We develop mechanisms that can naturally encode not only categorical but also ordinal information, and can also merge different data modalities. Furthermore, we show that the proposed models are naturally interpretable, which allows and facilitates their use in sensitive domains such as health applications. In particular, our models are tested in the diagnosis of several eye diseases and prostate cancer. First, we show the effectiveness and benefit of using regression models in the diagnosis of eye diseases of genetic origin. We then demonstrate the importance of including disease grading information and performing discrete regression to improve the performance of the binary diagnosis of diabetic retinopathy and prostate cancer. We show that a probabilistic interpretation of the results provides information on the uncertainty of the models, which can also be used in training processes. Finally, the proposed framework allows us to encode information using kernel functions, which in turn allows us to naturally introduce flexible information fusion mechanisms and thus to address multimodal tasks. Overall, we show that incorporating ordinal and multimodal information using probabilistic kernel-based frameworks allows learning better data representations, which improves the performance of the models and provides them with a higher level of interpretability.La principal contribución de esta tesis es el desarrollo de modelos probabilísticos de aprendizaje de máquina para apoyar el diagnóstico de enfermedades a partir de información médica. Mostramos cómo un enfoque probabilístico ofrece una gran versatilidad al momento de aprovechar toda la información disponible sobre la tarea objetivo. Basándonos en el formalismo matemático de la mecánica cuántica, desarrollamos y aplicamos modelos de aprendizaje que nos permiten manejar el flujo de información utilizando matrices de densidad de diferentes maneras. Desarrollamos mecanismos que pueden codificar de forma natural no sólo información categórica, sino también ordinal, y que también pueden fusionar distintas modalidades de información. Además, demostramos que los modelos propuestos son naturalmente interpretables, lo que permite y facilita su aplicación en dominios sensibles como las aplicaciones médicas. Precisamente, en este trabajo probamos nuestros modelos en tareas específicas de diagnóstico de enfermedades oculares y cáncer de próstata. En primer lugar, mostramos la eficacia y el beneficio de usar modelos de regresión en el diagnóstico de enfermedades oculares de origen genético. A continuación, demostramos la importancia de incluir información sobre el estadio de las enfermedades y realizar una regresión discreta para mejorar el rendimiento del diagnóstico binario de la retinopatía diabética y el cáncer de próstata. Demostramos que la interpretación probabilística de los resultados proporciona información sobre la incertidumbre de los modelos, que puede utilizarse también en los procesos de entrenamiento. Por último, los modelos propuestos nos permiten codificar la información mediante funciones kernel, que a su vez nos permiten introducir de forma natural mecanismos de fusión de información, flexibles y versátiles, y con estos abordar tareas multimodales. En conjunto, demostramos que la incorporación de información ordinal y multimodal mediante modelos probabilísticos basados en funciones de kernel permite aprender mejores representaciones de los datos, lo que mejora el rendimiento de los modelos y les proporciona un mayor nivel de interpretabilidad. (Texto tomado de la fuente).DoctoradoDoctor en IngenieríaSistemas Inteligente
Regresión y aprendizaje multimodal como ayuda al diagnóstico en oftalmologı́a e histopatologı́a
No full textilustraciones, diagramasThe main contribution of this thesis is the development of probabilistic machine learning models to support disease diagnosis from medical data sources. We show how a probabilistic approach offers great versatility in exploiting all available information about the target task. Based on the mathematical formalism of quantum mechanics, we develop and apply machine learning models that allow us to handle the flow of information using density matrices in different ways. We develop mechanisms that can naturally encode not only categorical but also ordinal information, and can also merge different data modalities. Furthermore, we show that the proposed models are naturally interpretable, which allows and facilitates their use in sensitive domains such as health applications. In particular, our models are tested in the diagnosis of several eye diseases and prostate cancer. First, we show the effectiveness and benefit of using regression models in the diagnosis of eye diseases of genetic origin. We then demonstrate the importance of including disease grading information and performing discrete regression to improve the performance of the binary diagnosis of diabetic retinopathy and prostate cancer. We show that a probabilistic interpretation of the results provides information on the uncertainty of the models, which can also be used in training processes. Finally, the proposed framework allows us to encode information using kernel functions, which in turn allows us to naturally introduce flexible information fusion mechanisms and thus to address multimodal tasks. Overall, we show that incorporating ordinal and multimodal information using probabilistic kernel-based frameworks allows learning better data representations, which improves the performance of the models and provides them with a higher level of interpretability.La principal contribución de esta tesis es el desarrollo de modelos probabilísticos de aprendizaje de máquina para apoyar el diagnóstico de enfermedades a partir de información médica. Mostramos cómo un enfoque probabilístico ofrece una gran versatilidad al momento de aprovechar toda la información disponible sobre la tarea objetivo. Basándonos en el formalismo matemático de la mecánica cuántica, desarrollamos y aplicamos modelos de aprendizaje que nos permiten manejar el flujo de información utilizando matrices de densidad de diferentes maneras. Desarrollamos mecanismos que pueden codificar de forma natural no sólo información categórica, sino también ordinal, y que también pueden fusionar distintas modalidades de información. Además, demostramos que los modelos propuestos son naturalmente interpretables, lo que permite y facilita su aplicación en dominios sensibles como las aplicaciones médicas. Precisamente, en este trabajo probamos nuestros modelos en tareas específicas de diagnóstico de enfermedades oculares y cáncer de próstata. En primer lugar, mostramos la eficacia y el beneficio de usar modelos de regresión en el diagnóstico de enfermedades oculares de origen genético. A continuación, demostramos la importancia de incluir información sobre el estadio de las enfermedades y realizar una regresión discreta para mejorar el rendimiento del diagnóstico binario de la retinopatía diabética y el cáncer de próstata. Demostramos que la interpretación probabilística de los resultados proporciona información sobre la incertidumbre de los modelos, que puede utilizarse también en los procesos de entrenamiento. Por último, los modelos propuestos nos permiten codificar la información mediante funciones kernel, que a su vez nos permiten introducir de forma natural mecanismos de fusión de información, flexibles y versátiles, y con estos abordar tareas multimodales. En conjunto, demostramos que la incorporación de información ordinal y multimodal mediante modelos probabilísticos basados en funciones de kernel permite aprender mejores representaciones de los datos, lo que mejora el rendimiento de los modelos y les proporciona un mayor nivel de interpretabilidad. (Texto tomado de la fuente).DoctoradoDoctor en IngenieríaSistemas Inteligente
