Bedingt durch die Internationalisierung der Märkte, einen sich zuspitzenden Preiswettbewerb, immer kürzere Produktlebenszyklen und nicht zuletzt durch die zunehmende Nutzung innovativer Informations- und Kommunikationstechnologien sehen sich viele Unternehmen einer erhöhten Marktdynamik und einem verschärften Wettbewerb gegenüber. Für die Konzeption von Management Support Systemen (MSS) bedeutet dies, dass neben den internen Informationen auch Informationen aus externen Quellen eingebunden, mit den internen verknüpft und hinsichtlich auffälliger Konstellationen für die Entscheidungsunterstützung aufbereitet werden müssen. Im Rahmen des von der Deutschen Forschungsgemeinschaft geförderten und kooperativ mit dem Bereich Wirtschaftsinformatik I (Prof. Dr. Dr. h. c. mult. Peter Mertens) an der Universität Erlangen-Nürnberg bearbeiteten Forschungsprojektes INTEX („Integration von Controlling- und Marktforschungsdaten in einem Expertisesystem“) wurde dazu ein Konzept entwickelt und prototypisch realisiert [MeUh01].

As part of Volvo Penta's initiative to further the development of predictive maintenance in their field test environments, this thesis compares neural networks in an effort to predict the occurrence of three common diagnostics trouble codes using field test data. To quantify the neural networks' performances for comparison a number of evaluation metrics were used. By training a multitude of differently configured feedforward neural networks with the processed field test data and evaluating the resulting models, it was found that the resulting models perform better than that of a baseline classifier. As such it is possible to use Volvo Penta's field test data along with neural networks to achieve predictive maintenance. It was also found that Long Short-Term Memory (LSTM) networks with methodically selected hyperparameters were able to predict the diagnostic trouble codes with the greatest performance among all the tested neural networks.

Diabetic Retinopathy (DR) is one of the principal sources of blindness due to diabetes mellitus. It can be identified by lesions of the retina, namely microaneurysms, hemorrhages, and exudates. DR can be effectively prevented or delayed if discovered early enough and well-managed. Prior studies on diabetic retinopathy typically extract features manually but are time-consuming and not accurate. In this research, we propose a research framework using advanced retina image processing, deep learning, and a boosting algorithm for high-performance DR grading. First, we preprocess the retina image datasets to highlight signs of DR, then follow by a convolutional neural network to extract features of retina images, and finally apply a boosting tree algorithm to make a prediction based on extracted features. Experimental results show that our pipeline has excellent performance when grading diabetic retinopathy images, as evidenced by scores for both the Kaggle dataset and the IDRiD dataset. / Master of Science

The incidence of cutaneous malignant melanoma (CMM) is steadily increasing. Most of the patients have thin CMM with a good prognosis and a 5-year survival of about 90%. The prognosis is highly related to tumour thickness and clinical stage at diagnosis. Effective systemic treatment for patients with metastatic disease has only recently been available. This thesis aims to increase knowledge of trends in tumour thickness, prognostic factors, socioeconomic differences and medical costs in patients with CMM. The population-based Swedish melanoma register is the main source of data in all papers in the thesis. Papers I-III include patients from all of Sweden while paper IV is delimited to the County of Östergötland. Cox regression and logistic regression are the main multivariable methods used. Paper IV is focused on stage-specific costs of CMM by comparing direct healthcare costs to a general population. For men, there has been a shift over time towards thinner tumours at diagnosis accompanied by an improved survival. Women are still diagnosed with considerably thinner tumours and they experience a better survival than men. Tumour ulceration, tumour thickness and Clark’s level of invasion all showed significant independent long-term prognostic information in T1 CMMs. By combining these factors, three distinct prognostic subgroups were identified. Lower level of education was associated with reduced CMM-specific survival, which may at least partially be attributed to a more advanced stage at diagnosis. The direct healthcare costs for CMM patients were significantly higher than for the general population, independent of clinical stage. CMM patients diagnosed in clinical stage III-IV were associated with particularly high costs. Even though the survival among Swedish patients with CMM is among the highest in the world and still seems to improve, the results of this thesis emphasise the need of improved early detection strategies. This may be of particular concern in men, older women, and groups with a low level of education. The results also imply that the costs for the management of CMM patients may be reduced if early detection efforts are successful and lead to a more favourable stage distribution. The finding of a better risk stratification of thin CMMs may help to improve the management of this large patient group.

This dissertation examines sociodemographic determinants and preventive health behaviors among women 40-64 years of age who participated in the Virginia Department of Health’s Every Woman’s Life breast cancer screening program. Utilizing secondary data, this research sought to explore patterns of breast cancer incidence, mammography screening utilization and sources of health information among low-income women. The Virginia Department of Health provided a large sample size (N=34,942) on which to perform binary logistic regression analyses. Sociodemographic determinants and preventive health behaviors were analyzed as potential influencing factors in the diagnosis of breast cancer, the stage at the time of diagnosis and source of health information. Additionally, frequencies across all variables were explored and compared to state and national statistics, where appropriate. In this study, cancer and preventive health disparities reported in the literature persist within this sample of low income women. The binary regression analyses demonstrated that there are marginally worse outcomes for each level of decreasing income. Those with the most “wealth” were less likely to be diagnosed with invasive breast cancer and were more likely to obtain health information from a health provider. Additionally, it was determined that those without a prior mammogram were more likely to be diagnosed with breast cancer and the cancer was more likely to be invasive. The aims of the Every Woman’s Life program align with Affordable Care Act (2010) to strengthen health care and eliminate cancer disparities. Highlighting program characteristics and presenting these analyses allows policymakers, program officials and practitioners an opportunity to tailor health promotion activities while considering all tiers of influence.

Since the field of computer networks has rapidly grown in the last two decades, congestion control of traffic loads within networks has become a high priority. Congestion occurs in network routers when the number of incoming packets exceeds the available network resources, such as buffer space and bandwidth allocation. This may result in a poor network performance with reference to average packet queueing delay, packet loss rate and throughput. To enhance the performance when the network becomes congested, several different active queue management (AQM) methods have been proposed and some of these are discussed in this thesis. Specifically, these AQM methods are surveyed in detail and their strengths and limitations are highlighted. A comparison is conducted between five known AQM methods, Random Early Detection (RED), Gentle Random Early Detection (GRED), Adaptive Random Early Detection (ARED), Dynamic Random Early Drop (DRED) and BLUE, based on several performance measures, including mean queue length, throughput, average queueing delay, overflow packet loss probability, packet dropping probability and the total of overflow loss and dropping probabilities for packets, with the aim of identifying which AQM method gives the most satisfactory results of the performance measures. This thesis presents a new AQM approach based on the RED algorithm that determines and controls the congested router buffers in an early stage. This approach is called Dynamic RED (REDD), which stabilises the average queue length between minimum and maximum threshold positions at a certain level called the target level to prevent building up the queues in the router buffers. A comparison is made between the proposed REDD, RED and ARED approaches regarding the above performance measures. Moreover, three methods based on RED and fuzzy logic are proposed to control the congested router buffers incipiently. These methods are named REDD1, REDD2, and REDD3 and their performances are also compared with RED using the above performance measures to identify which method achieves the most satisfactory results. Furthermore, a set of discrete-time queue analytical models are developed based on the following approaches: RED, GRED, DRED and BLUE, to detect the congestion at router buffers in an early stage. The proposed analytical models use the instantaneous queue length as a congestion measure to capture short term changes in the input and prevent packet loss due to overflow. The proposed analytical models are experimentally compared with their corresponding AQM simulations with reference to the above performance measures to identify which approach gives the most satisfactory results. The simulations for RED, GRED, ARED, DRED, BLUE, REDD, REDD1, REDD2 and REDD3 are run ten times, each time with a change of seed and the results of each run are used to obtain mean values, variance, standard deviation and 95% confidence intervals. The performance measures are calculated based on data collected only after the system has reached a steady state. After extensive experimentation, the results show that the proposed REDD, REDD1, REDD2 and REDD3 algorithms and some of the proposed analytical models such as DRED-Alpha, RED and GRED models offer somewhat better results of mean queue length and average queueing delay than these achieved by RED and its variants when the values of packet arrival probability are greater than the value of packet departure probability, i.e. in a congestion situation. This suggests that when traffic is largely of a non bursty nature, instantaneous queue length might be a better congestion measure to use rather than the average queue length as in the more traditional models.

This study proposes an accuracy comparison of two of the best performing machine learning algorithms in natural language processing, the Bayesian Network and the Long Short-Term Memory (LSTM) Recurrent Neural Network, in detecting Alzheimer’s disease symptoms in conversation transcripts. Because of the current global rise of life expectancy, the number of seniors affected by Alzheimer’s disease worldwide is increasing each year. Early detection is important to ensure that affected seniors take measures to relieve symptoms when possible or prepare plans before further cognitive decline occurs. Literature shows that natural language processing can be a valid tool for early diagnosis of the disease. This study found that mild dementia and possible Alzheimer’s can be detected in conversation transcripts with promising results, and that the LSTM is particularly accurate in said detection, reaching an accuracy of 86.5% on the chosen dataset. The Bayesian Network classified with an accuracy of 72.1%. The study confirms the effectiveness of a natural language processing approach to detecting Alzheimer’s disease.

Les capacités physiques et cognitives diminuent considérablement à cause du vieillissement. Les problèmes de santé liées au vieillissement présentent une grande charge pour la santé publique. Aujourd’hui, les services gériatriques ne sont pas suffisants pour détecter les problèmes de santé dans les premiers stades de leur évolution, dans le but d’améliorer l’évaluation et l’intervention médicale des personnes âgées. Traduire ces besoins gériatriques à travers les services existants est essentiel pour améliorer leur impact. Dans le cadre de ma thèse, je propose une approche technologique qui utilise des technologies non-intrusives pour analyser le comportement des personnes âgés sur de longues périodes et détecter des possibilités d’évolution de maladies physiques ou cognitives. Des références gériatriques internationales me permettent d’identifier des indicateurs de changement de comportement qui peuvent être suivis à travers de technologies non-intrusives sans interférer avec le comportement naturel des personnes âgées. J’analyse ces indicateurs en considérant plusieurs dimensions spatio-temporelles, et utilisant des techniques de détection de changement qui différencient les changements transitoires et continus dans le comportement suivi. Je valide mon approche proposée à travers un déploiement réel de 3 ans dans une maison de retraite et des maisons individuelles.Je propose une méthodologie de détection précoce et non-intrusive des possibilités de changement dans l’état de santé. Des entretiens personnels avec les personnes âgées, les membres de la famille, les médecins gériatres et les infirmières de la maison de retraite me permettent d’identifier leurs besoins gériatriques. Les parties prenantes de mes services proposés ont besoin d’une information fiable à propos des changements possibles dans l’état de santé le plus tôt possible, sans suivre les personnes d’une manière intrusive. Afin de traduire ces besoins gériatriques, je propose une approche technologique qui utilise des technologies non-intrusives pour suivre les personnes âgées pendant des semaines et des mois, et identifier des changements possible dans leur comportement fortement liés à des problèmes physiques ou cognitifs.Mon service web ChangeTracker implémente ma méthodologie. ChangeTracker analyse le comportement des personnes âgées en ligne et détecte des changements possibles chaque jour. Je développe des algorithmes qui convertissent les données de capteurs brutes en données inférées en relation avec l’état de santé de la personne suivie. Des techniques de détection de changement (par ex., des techniques statistiques, probabilistes et d’apprentissage) distinguent les changements temporaires et continus dans le comportement de la personne.Une validation réelle de mon approche a lieu dans 3 villes françaises Montpellier, Lattes et Occagnes. Les résultats expérimentaux de 25 participants validement ma détection précoce et non-intrusive des changements de santé. Les 25 participants vivent seuls à domicile ou dans une maison de retraite. Dans mon cas d’étude, j’installe des capteurs de mouvement dans chaque chambre de la maison et des capteurs de contact sur chaque porte principale. Ces capteurs collectent mes données de suivi pendant 3 ans. Mes algorithmes analysent ces données, calculent des indicateurs gériatriques significatifs, et détectent des changements possibles en corrélation avec l’état de santé. / Aging process is associated with serious decline in physical and cognitive abilities. Aging-related health problems present growing burden on public health and economy. Nowadays, existing geriatric services have limitations in terms of early detecting possible health changes toward better adaptation of medical assessment and intervention for elderly people. Bridging the gap between these geriatric needs and existing services is a major enabler to improve their impact. In this thesis, proposed technological approach employs unobtrusiveInternet of Things (IoT) technologies for long-term behavior monitoring and early detection of possible changes. Proposed methodology identifies geriatric indicators that can be monitored via unobtrusive IoT technologies, and are associated with physical and cognitive problems. This thesis develops data processing algorithms that convert raw sensor data into geriatric indicators. These geriatric indicators are analyzed on a daily basis, in order to early detect possible changes. This thesis evaluates and adapts further statistical, probabilistic and machine-learning techniques for long-term change detection. Adapting these techniques discards transient deviations, and retains permanent changes in monitored behavior. Real 3-year deployments in nursing home and individual houses validate proposed approach. Medical clinic geriatrician and nursing home team validate medical relevance of detected changes.

O câncer de mama é a neoplasia maligna mais prevalente entre as mulheres no mundo e representa 23% de todos os cânceres femininos. Buscou-se avaliar a implementação de um programa de rastreamento mamográfico para as mulheres que vivem na área de Barretos usando uma unidade móvel (UM) e uma unidade fixa (UF). Um total de 54.238 mulheres com idade entre 40 a 69 anos reside nesta área e são elegíveis para a participação no programa. Os dados epidemiológicos das mulheres foram examinadas entre 01 de abril de 2003 e 31 de março de 2005. A análise estatística foi constituída pela avaliação das freqüências dos parâmetros clínicos e as características do tumor usando o teste de Qui-quadrado com correção de Bonferroni, com valor de confiança de p<0,05. Um total de 17.964 mulheres (media de 51 anos de idade) foram efetivamente examinadas por mamografia, o que representou 33,1% de todas as mulheres elegíveis (18,6 exames por dia na UF e 26,3 na UM). Setenta e seis casos foram diagnosticados como câncer de mama (41, ou 54%, no UM), o que representa 4,2 casos de câncer de mama para cada 1.000 exames. Foi observada diferença significativa na detecção de câncer entre mulheres com idade entre 50-59 e 60-69 anos (p<0,001) e com idade entre 40-49 e 60-69 anos (p<0,001). Não foram observadas diferenças entre 40 a 49 e 50-59 anos (p = 0,164). O programa de rastreamento mamografico é viável no território nacional e os resultados preliminares são animadores / Breast cancer is the most prevalent malignancy among women worldwide and enrolls 23% of all female cancers. We sought to evaluate the implementation of a screening program for women who living in Barretos county area using a mobile unit (MU) and a fixed unit (FU). A total of 54,238 women aged 40 to 69 years is living in this area and are eligible for breast screening. Epidemiologic-based data supported the study design and the women were examined from April 01, 2003 to March 31, 2005. Statistical analysis supported the evaluation of clinical parameters frequencies and tumor characteristics using Chi-test and Bonferroni correction test, with confidence value of p<0,05. Overall of 17,964 women (media of 51 years old) were effectively examined by mammogram which represented 33,1% of all eligible women (18,6 in RA and 26,3 exams in MU per day). Seventy-six cases were diagnosed as breast cancer (41, or 54%, at MU), which represents 4,2 cases of breast cancer for each 1.000 exams. It was observed significant difference of cancer detection between women aged 50 to 59 and 60 to 69 yrs (p<0, 001) and between women aged 40 to 49 and 60 to 69 yrs (p<0,001). No differences were observed between aged 40 to 49 and 50 to 59 yrs (p=0,164). The program for mammogram screening is feasible to be implementing in Brazil territory and the preliminary results are encouraging

Introdução: O câncer de próstata é a neoplasia maligna não-cutânea mais frequente nos homens brasileiros. Seu rastreamento é tema controverso na literatura, e a maioria das entidades médicas não recomenda sua realização a partir dos 70 anos. Não existem estudos sobre suas características nessa faixa etária da população brasileira, que não é submetida a rastreamento ativo sistemático. Objetivos: Avaliar a prevalência e critérios de agressividade do câncer de próstata diagnosticado por rastreamento ativo em homens com idade inferior e superior a 70 anos no Brasil. Pacientes e métodos: Estudo transversal retrospectivo incluindo 17.571 voluntários no Brasil, submetidos a rastreamento ativo através de toque retal e dosagem sérica do antígeno prostático específico (PSA), entre janeiro de 2004 e dezembro de 2007. Os critérios de indicação para a biópsia foram: PSA>4,0ng/ml, ou PSA entre 2,5 e 4,0ng/ml com relação PSA livre/total <=15%, ou toque retal suspeito. Todos os homens rastreados foram divididos em dois grupos etários: grupo A, entre 45 e 69 anos; grupo B, acima de 70 anos. Os grupos foram comparados com relação a prevalência e critérios de agressividade da doença (valor do PSA sérico, escore de Gleason da biópsia e estadiamento clínico TNM). Resultados e discussão: A prevalência do câncer de próstata na nossa amostra foi de 3,71%. O grupo dos homens com mais de 70 anos apresentou prevalência da doença 2,9 vezes maior (RP 2,90; p <0,001), o valor médio de PSA foi mais elevado nos acometidos (17,28ng/ml no grupo B versus 9,54ng/ml no grupo A), assim como ocorreu maior chance de haver portadores de câncer com PSA acima de 10,0ng/ml (OR 2,63; p=0,003). No grupo de homens com mais de 70 anos também houve uma prevalência 3,59 vezes maior do padrão histológico mais agressivo (Gleason 8-10: RP 3,59; p<0,001) e maior prevalência de doença metastática (RP 4,95; p<0,05). Conclusão: O rastreamento do câncer de próstata nos homens com idade acima de 70 anos e expectativa de vida superior a 10 anos pode ser relevante no Brasil. Neste grupo etário detectamos uma maior prevalência desta doença, quando comparado ao grupo de idade entre 45 e 69 anos. Nosso estudo também demonstrou que o grupo de homens com mais de 70 anos possui maior probabilidade de apresentar doença de alto risco ao diagnóstico (PSA sérico mais elevado e em faixas de maior risco; escore de Gleason 8 a 10 e disseminação metastática à distância mais frequentes) / Background: Prostate cancer (PC) is the leading non-cutaneous malignancy among Brazilian men. PC may present as an indolent or aggressive life-threatening disease. There is no consensus in the literature regarding PC screening, and most medical organizations do not recommend it over the age of 70 years old. There are no studies in the literature addressing this topic in the Brazilian population. Objectives: To compare the prevalence and the aggressiveness of prostate cancer diagnosed, by active screening, in men under and over 70 years. Patients and methods: We performed a retrospective cross-sectional study including 17,571 volunteers. Screening was performed by digital rectal examination and prostatespecific antigen (PSA) measurement. Individuals who met the criteria for PC suspicion (PSA>4.0ng/ml, or PSA 2.5-4.0ng/ml with free/total PSA ratio <=15%, or suspicious digital rectal examination) underwent prostate biopsy. Those diagnosed with cancer were staged. The screened men were stratified by age in two groups: group A, between 45 and 69 years old, and group B, over 70 years old. The groups were compared regarding PC prevalence and its aggressiveness criteria (seric PSA value, Gleason score from biopsy and TNM staging). Results and discussion: The prevalence of prostate cancer was 3.71% in all population. The group of men over 70 years old had disease prevalence 2.9 times higher (RP 2.90; p<0.001); higher mean PSA value in men diagnosed with prostate cancer (17.28ng/ml vs. 9.54ng/ml); and greater likelihood to present PC when PSA level was above 10.0ng/ml (OR 2.63; p=0.003), when compared with men between 45 and 69 years old. The group of men aged over 70 years also presented a prevalence of histologic aggressive disease 3.59 times higher (Gleason 8-10: RP 3.59, p<0.001) and greater prevalence of metastatic disease (RP 4,95; p<0,05). Conclusion: Our study reveals that men over 70 years old presented a higher prevalence of prostate cancer and a higher probability to present high-risk disease at diagnosis (higher PSA; Gleason score 8-10 and metastatic disease more frequent), when compared to men aged 45-69 years. Screening for prostate cancer in men aged over 70 years and life expectancy over 10 years may be relevant in Brazil