In den letzten Jahren wurde die Sentinel Node Biopsie in Studien zur Diagnostik und Behandlung des N0-Halses als Stagingverfahren untersucht. Ziel der vorliegenden Studie war der Vergleich der postoperativen Lebensqualität und des funktionellen Status bei Patienten mit oralen und oropharyngealen Karzinomen, die sich einer Sentinelbiopsie, bzw. einer selektiven Neck Dissection bei N0-Hals unterzogen haben. 57 Patienten mit primär operativ behandeltem Karzinom wurden im Rahmen der Studie untersucht, 29 nach Sentinel Node Biopsie und 28 nach selektiver Neck Dissection. Es wurde die postoperative gesundheitsbezogene (EORTC QLQ-C30) und krankheitsspezifische Lebensqualität (EORTC QLQ-H&N35) erfasst. Patientenbezogene psychosoziale Variablen wurden anhand der „Hospital Anxiety and Depression Scale“ (HADS) und der Kurzform des „Progredienzangst-Fragebogens“ (PA-F-KF) ermittelt. Die Erhebung des klinischen Gesundheitszustandes erfolgte unter besonderer Berücksichtigung des funktionellen Status nach zervikaler Lymphknotendissektion. Bezüglich der gesundheitsbezogenen Lebensqualität sowie der Angst und Depression existierten keine signifikanten Unterschiede. Hinsichtlich der krankheitsspezifischen Lebensqualität zeigte sich, dass die Patienten nach Sentinel Node Biopsie weniger Schluckbeschwerden angaben (p=0,037). Ebenso zeigte sich bei der klinischen Untersuchung eine geringere Beeinträchtigung durch die zervikalen Narben. Die Sentinel Node Biopsie bietet, neben dem geringeren operativen Aufwand, der schnelleren Rekonvaleszenz und der geringeren funktionellen Minderung der Patienten, eine tendenzielle Verbesserung der krankheitsspezifischen Lebensqualität. Als Grund für die subjektiv weniger beeinträchtigte Schluckfunktion kann die fehlende Präparation der suprahyoidalen Muskulatur und der entsprechend innervierenden Nerven in Erwägung gezogen werden.

Einhergehend mit schnellem Bevölkerungs- und Wirtschaftswachstum erlebt die Welt innerhalb der letzten Jahrzehnte eine schnelle Akkumulation langlebiger Ressourcen in Gebäuden und Infrastruktur, auch gesellschaftlicher Materialbestand genannt. Im 21. Jahrhundert wird die Fortsetzung dieser Entwicklung zur großen Herausforderung für den sozioökonomischen Stoffwechsel der Erde und zum Erreichen biophysikalischer Grenzen führen. Siedlungen sind von besonderem Interesse, da Menschen dort Nachfrage nach Leistungen wie Nahrung oder Mobilität generieren und mit ihnen interagieren. Zukünftig wird neben einer globalen Entwicklungsperspektive auf Materialbestände und Bevölkerung auch ein räumlich explizites, hochauflösendes Verständnis lokaler Muster und Prozesse von Relevanz für eine datenbasierte Antwort auf Herausforderungen des globalen Wandels sein. Diese Arbeit präsentiert einen Workflow zur Kartierung und Quantifizierung von Materialbeständen und Bevölkerungsverteilung und -dynamik mittels hochaufgelöster mehrdimensionaler Siedlungskartierung mit Multisensor-Erdbeobachtungsdaten auf nationaler Ebene. Der erste Abschnitt demonstriert das Potenzial der Verwendung von Sentinel-1 und -2 Zeitreihendaten mit Methoden des maschinellen Lernens für die Kartierung von Siedlungsstrukturen, d.h. Subpixel-Landbedeckung, Gebäudehöhe und Gebäudetyp. Der zweite Abschnitt quantifiziert Schlüsselparameter des sozioökonomischen Metabolismus, d. h. Bevölkerung und Materialbestand, anhand zuvor generierter Datensätze zur Siedlungsstruktur. Der dritte Abschnitt nutzt das Landsat-Datenarchiv und Zeitreihenanalyse, um räumliche Muster und Dynamiken von Bevölkerung und Materialbeständen in Deutschland seit 1985 zu quantifizieren. Frei verfügbare und global konsistente Erdbeobachtungsdaten und Techniken des maschinellen Lernens haben großes Potenzial, das räumlich explizite hochaufgelöste Verständnis sozioökologischer Variablen basierend auf mehrdimensionaler Siedlungskartierung zu verbessern. / During the recent decades of the Anthropocene, the world has experienced rapid growth of population and economic activity. This went along with a considerable accumulation of long-lived resources, for example in buildings and infrastructure, i.e., societal material stock. In the 21st century, a continuation of this development will be a major challenge to the Earth’s socio-economic metabolism, as some limitations of the Earth’s biophysical basis might be reached. Settlements are of particular interest, because they are the places where people generate demand for, and interact with services. Both an overarching perspective on the global long-term development of material stock and population as well as a spatially explicit, high-resolution understanding of local patterns and processes will be of particular relevance for a more data-informed response to challenges of global change. This dissertation presents a workflow to map and quantify material stocks and population distribution and dynamics by means of multi-dimensional settlement mapping with decameter resolution multi-source Earth Observation data on a national scale. The first part demonstrates the potential of using Sentinel-1 and -2 time series imagery with machine learning regression and classification for settlement structure mapping, including sub-pixel land cover, building height and building type mapping. The second part quantifies key parameters of the socio-economic metabolism, i.e., population and material stock, using previously generated datasets on settlement structure. The third part uses the Landsat data archive and Change-Aftereffect-Trend analysis to quantify spatial-temporal patterns and dynamics of population and material stock development in Germany since 1985. Findings demonstrate that freely available and globally consistent Earth Observation data and machine learning techniques have great potential to improve the spatially explicit high-resolution understanding of socio-metabolic variables based on multi-dimensional settlement mapping in a seamless workflow.

This study aims to create a flood extent map with Sentinel imagery and to evaluate impacts on agricultural land in the lagoon region of central Vietnam. In this study, remote sensing images, obtained from 2017 to 2019, were used to simultaneously map the land cover status of a flood in the Quang Dien district. This study highlights flooded areas from Sentinel-2 images by calculating some indicators such as the Land Surface Water Index (LSWI) and the Enhanced Vegetation Index (EVI). Comparisons between the floodplain samples (GPS point-based) and flood mapping results, with the ground-truth data, indicate that the overall accuracy and Kappa coefficients were 97.9% and 0.62 respectively for 2017; the values for 2019 were 95.7% and 0.77 for the same coefficients. Land use maps overlying the flood-affected maps show that approximately 11% of the agriculture land area was affected by floods in 2019 comparison to a 10% in 2017. Wet rice was the most affected crop with the flooded area accounting for more than 70% of the district under each flood event. The most affected communes are: Quang An, Quang Phuoc and Quang Thanh. This study provides valuable information for flood disaster planning, mitigation and recovery activities in Vietnam. / Mục tiêu của nghiên cứu là lập bản đồ phân bố ngập lụt với hình ảnh vệ tinh Sentinel và đánh giá ảnh hưởng ngập lụt đến sử dụng đất nông nghiệp ở vùng đầm phá miền Trung, Việt Nam. Trong nghiên cứu này, ảnh viễn thám thu nhận giai đoạn 2017-2019 được sử dụng để xây dựng bản đồ hiện trạng sử dụng đất tại thời điểm bị ngập nước trên địa bàn huyện Quảng Điền. Nghiên cứu đã xác định được vùng ngập lụt ở huyện Quảng Điền bằng phương pháp phân loại chỉ số mặt nước (Land Surface Water Index – LSWI) và chỉ số khác biệt thực vật (Enhanced Vegetation Index-EVI) từ ảnh Sentinel-2. Xác định vùng nước lũ bị che khuất bởi mây bằng mô hình số hóa độ cao (DEM). Kết quả phân loại vùng ngập lụt được so sánh với giá trị tham chiếu mặt đất cho thấy độ chính xác tổng thể và hệ số Kappa đạt được trong năm 2017 là 97,9% và 0,62; trong khi năm 2019 đạt 95,7% và 0.77. Bản đồ sử dụng đất chồng lên bản đồ lũ lụt cho thấy khoảng 11% diện tích đất nông nghiệp bị ảnh hưởng bởi lũ lụt năm 2019 so với 10% năm 2017. Cây lúa nước là cây trồng bị ảnh hưởng nặng nề nhất, với diện tích bị ngập lụt chiếm hơn 70% diện tích lúa của huyện. Các xã bị ngập lớn là xã Quảng An, Quảng Phước và Quảng Thành. Nghiên cứu này cung cấp thông tin có giá trị cho các hoạt động lập kế hoạch, giảm nhẹ và phục hồi thiên tai lũ lụt ở Việt Nam.

Grasländer sind wichtige Elemente der zentraleuropäischen Landschaft und stellen essenzielle Ökosystemdienstleistungen bereit. Dürren, welche durch den globalen Klimawandel zunehmen, haben negative Auswirkungen auf die Vitalität und Produktivität von Grasland. Satellitenmissionen wie Sentinel-2 und Landsat liefern große, bisher ungenutzte Möglichkeiten für das Grasland Monitoring. Ansätze auf Basis quantitativer Parameter, z.B. Prozentanteile von photosynthetisch aktiver Vegetation (PV), nicht photosynthetisch aktiver Vegetation (NPV) und Boden sind bisher für die Anwendung in zentraleuropäischen Grasländern nicht erforscht. Das Ziel der Arbeit war es, das Verständnis von Dürreeinflüssen auf zentraleuropäische Grasländer durch die Entwicklung eines fernerkundungsbasierten Monitoring Frameworks zu verbessern. Der erste Teil dieses Frameworks umfasste die Ableitung konsistenter Zeitreihen von PV-, NPV-, und Bodenanteilen. Der zweite Teil umfasste die Quantifizierung von Dürreeffekten anhand dieser Zeitreihen. Die Ergebnisse zeigten einen großflächigen, massiven und langanhaltenden Rückgang von Graslandvitalität in extremen Dürrejahren (z.B. 2003, 2018-2020). Robuste statistische Zusammenhänge bestätigten die starke Kopplung von Graslandvitalität und Dürre, insbesondere bei gleichzeitigen Hitzewellen. Zudem beeinflussten Bodeneigenschaften sowie klimatische und hydrologische Bedingungen die Dürresensitivität. Die Ergebnisse unterstreichen den Wert von generalisierten Entmischungsansätzen basierend auf Sentinel-2/Landsat Zeitreihen für großflächiges, quantitatives Monitoring von Grasland. Die Ergebnisse deuten darauf hin, dass durch den Klimawandel verstärkte Dürreereignisse in Zukunft erheblichen Einfluss auf die Vitalität von Grasländern in Zentraleuropa haben werden. Die hier gewonnenen Informationen liefern wichtige Beiträge zur Verbesserung von Dürremonitoring und können die Maßnahmenentwicklung zur Verringerung von Dürreschäden im Grasland unterstützen. / Grasslands are vital landscape elements in Central Europe providing essential ecosystem services. Drought events, which are increasing with global climate change, negatively affect grassland vitality and productivity. Satellite remote sensing missions such as Sentinel-2/Landsat offer untapped potential for monitoring grassland vitality. However, workflows for grassland monitoring based on fractional cover of photosynthetic vegetation (PV), non-photosynthetic vegetation (NPV), and soil, remain largely unexplored. The goal of this thesis was to advance the understanding of drought impacts on Central European grasslands by developing a framework for monitoring grassland vitality. The framework included the retrieval of consistent PV, NPV, and soil fractional cover time series from Landsat/Sentinel-2, which was achieved by implementing and generalizing an unmixing workflow. Second, drought impacts were quantified and evaluated based on fractional cover time series. Results showed large-scale, severe, and long-lasting negative impacts on grassland vitality in extreme drought years (e.g., in 2003, and 2018-2020). Robust statistical links confirmed the overall consistent coupling of grassland vitality to drought, specifically to compounding droughts and heatwaves. Spatiotemporal patterns of grassland drought sensitivity revealed that underlying factors such as soil features, and climatic and hydrological conditions modulate drought impacts on local to regional scales. Findings of this thesis emphasize the value of generalized unmixing workflows based on Sentinel-2/Landsat time series for quantitative grassland monitoring across large areas. Furthermore, results suggest that droughts amplified by climate change will pose substantial challenges for grassland vitality across Central European grasslands in the future. The findings provide a steppingstone towards improved drought monitoring and can thus inform adaptation efforts to alleviate drought impacts on grasslands.

[pt] Imagens de sensoriamento remoto permitem o monitoramento e mapeamento de culturas de maneira precisa, apoiando práticas de agriculturaeficientes e sustentáveis com o objetivo de garantir a segurança alimentar.No entanto, a identificação do tipo de cultura a partir de dados de sensoriamento remoto em regiões tropicais ainda são consideradas tarefas comalto grau de dificuldade. As favoráveis condições climáticas permitem o uso,planejamento e o manejo da terra com maior flexibilidade, o que implica emculturas com dinâmicas mais complexas. Além disso, a presença constantede nuvens dificulta o uso de imagens ópticas, tornando as imagens de radar uma alternativa interessante para o mapeamento de culturas em regiõestropicais. Os modelos de campos aleatórios condicionais (CRFs) têm sidousados satisfatoriamente para explorar o contexto temporal e espacial naclassificação de imagens de sensoriamento remoto. Estes modelos oferecemuma alta precisão na classificação, no entanto, dependem de atributos extraídos manualmente com base em conhecimento especializado do domínio.Neste contexto, os métodos de aprendizado profundo, tais como as redesneurais convolucionais (CNNs), provaram ser uma alternativa robusta paraa classificação de imagens de sensoriamento, pois podem aprender atributosótimos diretamente dos dados. Este trabalho apresenta um modelo híbridobaseado em aprendizado profundo e CRF para o reconhecimento de culturas em áreas de regiões tropicais caracterizadas por ter uma dinâmicaespaço–temporal complexa. O framework proposto consiste em dois módulos: uma CNNs que modela o contexto espacial e temporal dos dados deentrada, e o CRF que modela a dinâmica temporal considerando a dependência entre rótulos para datas adjacentes. Estas dependências podem seraprendidas ou desenhadas por um especialista nas práticas de agriculturalocal. Comparações entre diferentes variantes de como modelar as transiçõestemporais são apresentadas usando sequências de imagens SAR de duas municipalidades no Brasil. Os experimentos mostraram melhorias significativasatingindo ate 30 por cento no F1 score por classe e ate 12 por cento no F1 score medio em relação ao modelo de base que não inclui dependências temporais duranteo processo de aprendizagem. / [en] Remote sensing imagery enables accurate crop mapping and monitoring, supporting efficient and sustainable agricultural practices to ensure food security. However, accurate crop type identification and crop area estimation from remote sensing data in tropical regions are still challenging tasks. Compared to the characteristic conditions of temperate regions, the more favorable weather conditions in tropical regions permit higher flexibility in land use, planning, and management, which implies complex crop dynamics. Moreover, the frequent cloud cover prevents the use of optical data during large periods of the year, making SAR data an attractive alternative for crop mapping in tropical regions. To exploit both spatial and temporal contex, conditional random fields (CRFs) models have been used successfully in the classification of RS imagery. These approaches deliver high accuracies; however, they rely on features engineering manually designed based on domain-specific knowledge. In this context, deep learning methods such as convolutional neural networks (CNNs) proved to be a robust alternative for remote sensing image classification, as they can learn optimal features and classification parameters directly from raw data. This work introduces a novel end-to-end hybrid model based on deep learning and conditional random fields for crop recognition in areas characterized by complex spatio-temporal dynamics typical of tropical regions. The proposed framework consists of two modules: a CNN that models spatial and temporal contexts from the input data and a CRF that models temporal dynamics considering label dependencies between adjacent epochs. These dependencies can be learned or designed by an expert in local agricultural practices. Comparisons between data-driven and prior-knowledge temporal constraints are presented for two municipalities in Brazil, using multi-temporal SAR image sequences. The experiments showed significant improvements in per class F1 score of up to 30 percent and up to 12 percent in average F1 score against a baseline model that doesn t include temporal dependencies during the learning process.

Introduction: Lyme disease is caused by the tick-borne spirochete Borrelia burgdorferi. Research has shown that dogs can be used as sentinels for human infection of Lyme disease. The purpose of this 5-year, retrospective study was to determine if there was any evidence that the incidence of canine Lyme disease has increased between 2000 and 2005 in Accomack and Northampton counties. An increased incidence in Lyme disease in dogs may indicate an increased present or future risk of Lyme disease in humans.Methods: Cases of canine Lyme disease were identified via practice invoicing systems and dogs that received doxycycline were entered into the database. Demographic information and the absence or presence of clinical signs such as fever, lameness, articular swelling, lymphadenomegaly, anorexia, general malaise and improvement after antibiotic use were collected. Testing history also was recorded.Results: Cases of canine Lyme disease that met any definition were identified (n=1048). Over the 5-year period the number of positive ELISA test results increased and the frequency of clinical signs decreased. The incidence of disease meeting the practitioner's definition increased until 2004 when the incidence dropped from 105.33 cases per 1,000 dogs to 56.93 cases per 1,000. The incidence of disease based on the study probable definition remained fairly constant with a high in 2002 of 2.94 cases per 1,000 dogs.Discussion: Trends of canine Lyme disease coincided with the introduction and use of the in-house ELISA test. Practitioners could identify more dogs exposed to Borrelia burgdorferi. The areas with the highest frequency of canine cases of Lyme disease also had the highest frequency of human cases reported to the Virginia Department of Health. Further study could identify animals that tested positive and later developed clinical signs. Using dogs as sentinels for human infection allows public health workers to identify endemic areas regardless of human case reports.

En av Sveriges största industrier är skogsindustrin. Att sköta stora skogsinnehav medför vissa svårigheter, t.ex. så kan i händelse av en storm kan delar av skogen bli vindfälld. Det är då viktigt att upptäcka och ta bort de fallna träden eftersom det annars kan leda till granbarkborrangrepp. En metod för att upptäcka den vindfällda skogen är att ta flygbilder över området, vilket kan bli både dyrt och tidsödande. Därför testas i denna studie detektering av stormfälld skog i Sentinel 2 bilder. Sentinel 2 har valts ut eftersom den har både en hög spatial- och temporal upplösning samt att bilderna är tillgängliga gratis. Tidigare studier på området har använt satellitbilder med en lägre spatial upplösning eller data från andra typer av fjärranalys. De flesta av dessa metoder är ganska komplexa eller väldigt specifika för ett särskilt fall. Metoden som tas fram i denna studie ska vara enkel att implementera även för personer utan någon djupare kunskap inom fjärranalys. Bilddifferens med olika index såsom NDVI, NDMI och GreenNDVI testas. Även oövervakad klassificering testas. Noggrannheten har utvärderats med två-stegs metoden med en noggrannhet på 85 % men även en konfusionsmatris tillämpas för att utvärdera noggrannheten av områden där ingen förändring inträffat. Bilddifferens med NDVI och GreenNDVI klarar två-stegs testet när ett statistiskt bestämt tröskelvärde används, NDVI får högst användarnoggrannhet. Felmatrisen visar dock att det finns många stormfällen i ytorna som blivit klassade som ingen förändring, den oövervakade klassificeringen får inte det problemet i samma utsträckning. Bilddifferens i NDVI med statistiskt bestämt tröskelvärde bedöms vara den mest effektiva metoden för att detektera stormfälld skog.

O diagnóstico precoce de tumores e metástase constitui atualmente o elemento de maior impacto dentro das políticas de saúde públicas contra o câncer. No câncer de mama e melanoma, a técnica de biópsia de linfonodo sentinela, para o diagnostico de metástase, tem sido muito utilizada evitando a dissecção total dos nodos da região anatômica afetada, e permitindo definir com precisão o procedimento terapêutico a utilizar. O objetivo principal deste trabalho centrou-se no desenvolvimento de conjugados radiomarcados de dextran-manose para diagnóstico, utilizando o núcleo de tecnécio altamente estável, [99mTc(CO)3]+. A cisteína, ligante tridentado, foi incorporada na estrutura dos conjugados, como agente quelante do Tecnécio-99m. As condições de marcação definidas para os produtos avaliados garantiram altos valores de pureza radioquímica (>90%) e atividade específica (>59,9 MBq/nmol) assim como uma alta estabilidade in vitro. Os conjugados de dextran-cisteína-manose demonstraram uma captação superior (4 vezes maior) nos nodos linfáticos em relação aos homólogos que não possuíam manose na estrutura. O conjugado de dextran-cisteína-manose de 30 kDa radiomarcado (99mTc-DCM2) foi o traçador com melhor desempenho biológico entre os avaliados à diferentes atividades injetadas. Demonstrou-se que concentrações superiores a 1 M favorecem a retenção do produto nos nodos linfáticos. As comparações com radiofármacos já utilizados no Brasil (Dextran-500 e Fitato) para detecção de linfonodo sentinela evidenciaram a superioridade do 99mTc-DCM2. / Early diagnosis of tumors and metastasis is the current cornerstone in public health policies directed towards the fights against cancer. In breast cancer and melanoma, the sentinel lymph node biopsy has been widely used for diagnoses of metastasis. The minor impact in patient of this technique compared with total nodes dissection and the accurate definition of therapeutic strategies have powered its spreading. The aim of this work was the development of radiolabeled dextran-mannose conjugates for diagnosis using the stable technetium core [99mTc(CO)3]+. Cysteine, a trident ligand, was attached to the conjugates backbone, as a chelate for 99mTc labeling. Radiolabeling conditions established for all products considered in this study showed high radiochemical purities (> 90%) and specific activities (>59,9 MBq/nmol) as well and high stability obtained through in vitro tests. The lymphatic node uptake increased significantly (4-folds) when mannose units were added to the conjugates compared with those without this monosaccharide. The radiolabeled cysteine-mannose-dextran conjugate with 30 kDa (99mTc-DCM2) showed the best performance at different injected activities among the studied tracers. Concentrations of this radiocomplex higher than 1 M demonstrated an improvement of lymph node uptakes. Comparisons of 99mTc-DCM2 performance with commercial radiopharmaceuticals in Brazil market for lymph node detection showed its upper profile.

A linfocintilografia tem contribuído muito para a visualização da drenagem linfática e linfonodo sentinela (LNS) acometido e o estado patológico do LNS é considerado o fator prognóstico mais importante do melanoma. Os objetivos deste trabalho foram descrever os padrões de drenagem linfática em pacientes operados por melanoma, avaliar as características clínicas, demográficas e cirúrgicas destes pacientes, assim como a contribuição da Linfocintilografia na localização dos LNS com vistas à biópsia do mesmo. Foram avaliados 29 pacientes com melanoma operados, e a Linfocintilografia pré-operatória foi realizada após 60 minutos da injeção de Fitato marcado com Tecnécio 99m (99mTc), em 4 pontos cardeais a 1,0 cm da lesão. Encontramos maior acometimento por melanoma na população feminina e idosa, maioria em região de tronco posterior. A maioria (37,04%) drenou para 2 LNS/LNNS, 88,89% dos pacientes drenaram apenas para região ipsilateral à lesão, e 88,89% também drenaram apenas para um único território de drenagem. Encontramos padrões de drenagem linfática inesperados em melanoma localizados nas regiões de cabeça e pescoço, tronco anterior e tronco posterior. A técnica de linfocintilografia foi eficaz para evidenciar LNS/LNNS em 93,10% dos pacientes e dos 23 exames anatomopatológicos realizados, 6 (26,09%) apresentaram comprometimento metastático nos LNS. / Lymphoscintigraphy has contributed much to the visualization of lymphatic drainage and sentinel lymph node (SLN) and involved SLN pathological state is considered the most important prognostic factor of melanoma. Our objectives were to describe the lymphatic drainage patterns, to evaluate the clinical, demographic, and surgical data, as well as assessing the contribution of lymphoscintigraphy for SLN localization in a sample of patients operated for melanoma. We evaluated 29 patients operate with melanoma, and the Preoperative lymphoscintigraphy was performed after 60 minutes of injection of phytate labeled with Technetium 99m (99mTc) in 4 cardinal points to 1,0 cm of the lesion. We found greater involvement by melanoma in the elderly female population and, most located in the posterior trunk region. The majority (37.04%) drained for 2 SLN/ SLNN, 88.89% of the patients drained only to the ipsilateral region, and also 88,89% had lymphatic drainage to only a single drainage area. We found unexpected lymphatic drainage patterns in melanoma localized in head and neck, trunk, anterior and posterior trunk. The technique of lymphoscintigraphy was effective to show SLN/SLNN in 93.10% of patients, and from 23 pathological examinations performed, 6 (26.09%) exhibited metastatic involvement in the SLN.

Cette thèse a permis la mise en place d’un système d’alerte précoce des épidémies de paludisme basé sur le système de surveillance sentinelle proposant différents seuils épidémiques et un modèle de prédiction. Les données collectées quotidiennement par SMS sont automatiquement stockées sur un serveur dédié. Concomitamment, le système acquiert systématiquement et de manière automatique des données satellitaires météorologiques sur chaque site sentinelle en lien avec les changements de prévalence du paludisme, tels que la température, les précipitations et l'indice de végétation (eng. NDVI). Une base de données des interventions contre le paludisme a également été créée. Ce système a déjà démontré sa capacité à détecter une épidémie de paludisme dans le sud-est du pays en 2014. Deuxièmement, nous avons réalisé une étude pour évaluer la relation entre la durée de l'efficacité de la campagne de masse des moustiquaires imprégnées d'insecticide à effet longue durée (MILD) et les épidémies de paludisme identifiées à Madagascar de 2009 à 2015 par le système de surveillance sentinelle. Cette étude a montré que la différence entre l'efficacité théorique et l’efficacité réelle peut entraîner des lacunes dans la couverture des services pendant les années suivantes, contribuant au rebond du paludisme et souligne la nécessité de mise en place de mécanisme de distribution continue de moustiquaires. Ce travail vise à maximiser l'utilité d'un système de surveillance sentinelle dans des milieux à ressources limitées, guider les changements dans l'orientation des programmes de lutte et fournir des exemples pratiques pour son utilisation dans d'autres systèmes ou contextes. / We describe a Malaria Early Warning System (MEWS) using various epidemic thresholds and a forecasting component with the support of recent technologies to improve the performance of a sentinel MEWS. Malaria-related data from sentinel sites collected by Short Message Service are automatically stored in a database hosted on a server at Institut Pasteur de Madagascar. Concomitantly our system routinely and automatically acquires site specific satellite weather data related to changes in malaria prevalence such as temperature, rainfall and Normalized Difference Vegetation Index (NDVI). A Malaria Control Intervention data base has also been. This system has already demonstrated its ability to detect a malaria outbreak in southeastern part of Madagascar in 2014. In a second time, we conducted a study to assess the relationship between the effectiveness of mass campaign of long-lasting insecticidal nets (LLIN) over time and malaria outbreaks identified in Madagascar from 2009 to 2015 through the Sentinel surveillance system. This study showed that the difference between efficacy and effectiveness may result in gaps in service coverage during the subsequent years contributing to malaria rebound well before the replacement of the LLINs and highlights the need of continuous distribution mechanism of LLINs.This work aims to maximize the usefulness of a sentinel surveillance system to predict and detect epidemics in limited-resource environments, to guide any changes in the orientation of malaria control programs and to provide practical examples and suggestions for use in other systems or settings.