Po radikalaus gydymo 20–50 proc. pacientų, kuriems nustatytas galvos–kaklo srities vėžys lokoregioninis atkrytis nustatomas per pirmus dvejus metus. Literatūroje paskelbtų tyrimų rezultatai taikant pakartotinę nuotolinę spindulinę terapiją dėl galvos-kaklo vėžio atkryčio prasti: 2-jų metų bendras išgyvenimas siekė 15,2–40 proc., vėlyvųjų 3-4 laipsnio komplikacijų dažnis buvo 1,4–47 proc., 5 laipsnio - 7,6 proc. Retrospektyvinių ir II fazės tyrimų rezultatai naudojant didelės dozės galios brachiterapiją galvos-kaklo srities vėžio atkryčiui gydyti: 2-jų metų bendras išgyvenimas siekė 19–63 proc., vėlyvųjų 3-4 laipsnio komplikacijų dažnis buvo 4–22,2 proc. Tyrimų metu skirtos 3–4 Gy frakcijos iki 30–40 Gy suminės dozės. Iki šiol neatlikti tyrimai lyginantys nuotolinės spindulinės terapijos ir didelės dozės galios brachiterapijos gydymo veiksmingumą ir saugumą. Šioje disertacijoje palyginti skirtingi spindulinio gydymo metodai gydant galvos-kaklo srities vėžio atkrytį: kontrolinei grupei taikytas nuotolinis konforminis spindulinis gydymas (25 frakcijos po 2 Gy, suminė dozė 50 Gy), tiriamajai grupei - hipofrakcionuota didelės dozės galios brachiterapija skiriant naują frakcionavimo režimą – po 2,5 Gy per frakciją po dvi frakcijas per dieną, iki 30 Gy suminės dozės. Toks frakcionavimo režimas pasirinktas siekiant sumažinti spindulinių reakcijų dažnį ir sunkumo laipsnį, o suminė dozė yra biologiškai ekvivalentiška suminėms dozėms, kurios buvo naudotos ankstesniuose tyrimuose. / After radical treatment of head and neck cancer 20–50% of patients are diagnosed with the locoregional recurrence during first two years. In the literature the results of studies, using reirradiation by three-dimensional radiotherapy for head and neck cancer recurrence, according to a 2-year overall survival and toxicity, are poor: overall survival reached 15.2–40%, the grade 3 - 4 toxicity reached 1.4–47% and grade 5 - 7.6%. The results of phase II and retrospective studies using the high-dose-rate brachytherapy for treatment of head and neck cancer relapse were: 2-year overall survival was 19–63%; grade 3 - 4 late toxicity 4–22.2%. In these studies 3–4 Gy per fraction up to 30–40 Gy total dose were administered. So far, the randomized study, comparing the high-dose-rate brachytherapy with the three-dimensional radiotherapy, treating head and neck cancer relapse, hasn’t been conducted. We compared different radiotherapy methods: three-dimensional conformal radiotherapy was administered to the control group (25 fractions of 2 Gy, total dose of 50 Gy); the hypofractionated high-dose-rate brachytherapy was administered to the experimental group, while applying a new regime of fractionation: 2.5 Gy per fraction, two fractions per day, up to 30 Gy total dose. Such fractionation regimen was selected in order to reduce the rate and grade of toxicity, while the total dose is biologically equivalent to the total doses, which have been used in previous studies.

The management of rectal cancer has changed in many countries over the last two decades and resulted in improved survival for the majority of rectal cancer patients. In this thesis some surgical strategies and histopathological aspects to improve and clarify the management of rectal cancer patients are investigated. Even in the era of TME surgery and radiotherapy, a higher local recurrence rate and shorter survival for rectal cancer patients operated with abdominoperineal resection is reported. In the first paper we describe a new strategy with partial anterior en bloc resection of either the prostate or the vagina, resulting in very low local recurrence rates and excellent long-term survival. Histopathological examination of the specimen lays the foundation for decision making on oncological therapy. A positive circumferential resection margin (CRM) has, in previous papers, been related to a high risk of local recurrence. In the second paper we show that a CRM ≤ 1 mm was not correlated with an increased risk of local recurrence when patients were managed in a multidisciplinary setting with preoperative radiotherapy and optimal TME surgery. As the complexity of rectal cancer management is increasing, demands on organizational structure are growing. In paper three we could show that long-term survival was increased for all rectal cancer patients after the centralization to a single unit. Whether or not to resect the primary rectal tumour in patients with metastatic disease is an ongoing debate in the literature. In paper four, we studied the national management of rectal cancer patients with primary metastatic disease. Nineteen per cent of rectal cancer patients present with Stage IV disease and, at a national level, there is a clear shift to a more selective and restrictive approach. The 30-day mortality was low for patients that underwent a resectional surgery, for patients having an exploratory laparotomy, however, it was high. Overall survival was improved over time even though up to one fourth of patients received no surgical treatment.

Sleep disordered breathing (SDB) is a highly prevalent but an under-diagnosed disease. Among adults in the ages between 30 to 60 years, 24% of males and 9% of females show conditions of SDB, while 82% of men and 93% of women with moderate to severe SDB remain undiagnosed. Polysomnography (PSG) is the reference diagnostic test for SDB. During PSG, a number of physiological signals are recorded during an overnight sleep and then manually scored for sleep/wake stages and SDB events to obtain the reference diagnosis. The manual scoring of SDB events is an extremely time consuming and cumbersome task with high inter- and intra-rater variations. PSG is a labour intensive, expensive and patient inconvenient test. Further, PSG facilities are limited leading to long waiting lists. There is an enormous clinical need for automation of PSG scoring and an alternative automated ambulatory method suitable for screening the population. During the work of this thesis, we focus (1) on implementing a framework that enables more reliable scoring of SDB events which also lowers manual scoring time, and (2) implementing a reliable automated screening procedure that can be used as a patient-friendly home based study. The recordings of physiological measurements obtained during patients’ sleep of- ten suffer from data losses, interferences and artefacts. In a typical sleep scoring session, artifact-corrupted signal segments are visually detected and removed from further consideration. We developed a novel framework for automated artifact detection and signal restoration, based on the redundancy among respiratory flow signals. The signals focused on are the airflow (thermistor sensors) and nasal pressure signals that are clinically significant in detecting respira- tory disturbances. We treat the respiratory system as a dynamical system, and use the celebrated Takens embedding theorem as the theoretical basis for sig- nal prediction. In this study, we categorise commonly occurring artefacts and distortions in the airflow and nasal pressure measurements into several groups and explore the efficacy of the proposed technique in detecting/recovering them. Results we obtained from a database of clinical PSG signals indicated that theproposed technique can detect artefacts/distortions with a sensitivity >88% and specificity >92%. This work has the potential to simplify the work done by sleep scoring technicians, and also to improve automated sleep scoring methods. During the next phase of the thesis we have investigated the diagnostic ability of single – and dual–channel respiratory flow measuring devices. Recent studies have shown that single channel respiratory flow measurements can be used for automated diagnosis/screening for sleep disordered breathing (SDB) diseases. Improvements for reliable home-based monitoring for SDB may be achieved with the use of predictors based on recurrence quantification analysis (RQA). RQA essentially measures the complex structures present in a time series and are relatively independent of the nonlinearities present in the respiratory measurements such as those due to breathing nonlinearities and sensor movements. The nasal pressure, thermistor-based airflow, abdominal movement and thoracic movement measurements obtained during Polysomnography, were used in this study to implement an algorithm for automated screening for SDB diseases. The algorithm predicts SDB-affected measurement segments using twelve features based on RQA, body mass index (BMI) and neck circumference using mixture discriminant analysis (MDA). The rate of SDB affected segments of data per hour of recording (RDIS) is used as a measure for the diagnosis of SDB diseases. The operating points to be chosen were the prior probability of SDB affected data segments (π1) and the RDIS threshold value, above which a patient is predicted to have a SDB disease. Cross-validation with five-folds, stratified based on the RDI values of the recordings, was used in estimating the operating points. Sensitivity and specificity rates for the final classifier were estimated using a two-layer assessment approach with the operating points chosen at the inner layer using five-fold cross-validation and the choice assessed at the outer layer using repeated learning-testing. The nasal pressure measurement showed higher accuracy compared to other respiratory measurements when used alone. The nasal pressure and thoracic movement measurements were identified as the best pair of measurements to be used in a dual channel device. The estimated sensitivity and specificity (standard error) in diagnosing SDB disease (RDI ≥ 15) are 90.3(3.1)% and 88.3(5.5)% when nasal pressure is used alone and together with the thoracic movement it was 89.5(3.7)% and 100.0(0.0)%. Present results suggest that RQA of a single respiratory measurement has potential to be used in an automated SDB screening device, while with dual-channel more reliable accuracy can be expected. Improvements may be possible by including other RQA based features and optimisation of the parameters.

Sleep disordered breathing (SDB) is a highly prevalent but an under-diagnosed disease. Among adults in the ages between 30 to 60 years, 24% of males and 9% of females show conditions of SDB, while 82% of men and 93% of women with moderate to severe SDB remain undiagnosed. Polysomnography (PSG) is the reference diagnostic test for SDB. During PSG, a number of physiological signals are recorded during an overnight sleep and then manually scored for sleep/wake stages and SDB events to obtain the reference diagnosis. The manual scoring of SDB events is an extremely time consuming and cumbersome task with high inter- and intra-rater variations. PSG is a labour intensive, expensive and patient inconvenient test. Further, PSG facilities are limited leading to long waiting lists. There is an enormous clinical need for automation of PSG scoring and an alternative automated ambulatory method suitable for screening the population. During the work of this thesis, we focus (1) on implementing a framework that enables more reliable scoring of SDB events which also lowers manual scoring time, and (2) implementing a reliable automated screening procedure that can be used as a patient-friendly home based study. The recordings of physiological measurements obtained during patients’ sleep of- ten suffer from data losses, interferences and artefacts. In a typical sleep scoring session, artifact-corrupted signal segments are visually detected and removed from further consideration. We developed a novel framework for automated artifact detection and signal restoration, based on the redundancy among respiratory flow signals. The signals focused on are the airflow (thermistor sensors) and nasal pressure signals that are clinically significant in detecting respira- tory disturbances. We treat the respiratory system as a dynamical system, and use the celebrated Takens embedding theorem as the theoretical basis for sig- nal prediction. In this study, we categorise commonly occurring artefacts and distortions in the airflow and nasal pressure measurements into several groups and explore the efficacy of the proposed technique in detecting/recovering them. Results we obtained from a database of clinical PSG signals indicated that theproposed technique can detect artefacts/distortions with a sensitivity >88% and specificity >92%. This work has the potential to simplify the work done by sleep scoring technicians, and also to improve automated sleep scoring methods. During the next phase of the thesis we have investigated the diagnostic ability of single – and dual–channel respiratory flow measuring devices. Recent studies have shown that single channel respiratory flow measurements can be used for automated diagnosis/screening for sleep disordered breathing (SDB) diseases. Improvements for reliable home-based monitoring for SDB may be achieved with the use of predictors based on recurrence quantification analysis (RQA). RQA essentially measures the complex structures present in a time series and are relatively independent of the nonlinearities present in the respiratory measurements such as those due to breathing nonlinearities and sensor movements. The nasal pressure, thermistor-based airflow, abdominal movement and thoracic movement measurements obtained during Polysomnography, were used in this study to implement an algorithm for automated screening for SDB diseases. The algorithm predicts SDB-affected measurement segments using twelve features based on RQA, body mass index (BMI) and neck circumference using mixture discriminant analysis (MDA). The rate of SDB affected segments of data per hour of recording (RDIS) is used as a measure for the diagnosis of SDB diseases. The operating points to be chosen were the prior probability of SDB affected data segments (π1) and the RDIS threshold value, above which a patient is predicted to have a SDB disease. Cross-validation with five-folds, stratified based on the RDI values of the recordings, was used in estimating the operating points. Sensitivity and specificity rates for the final classifier were estimated using a two-layer assessment approach with the operating points chosen at the inner layer using five-fold cross-validation and the choice assessed at the outer layer using repeated learning-testing. The nasal pressure measurement showed higher accuracy compared to other respiratory measurements when used alone. The nasal pressure and thoracic movement measurements were identified as the best pair of measurements to be used in a dual channel device. The estimated sensitivity and specificity (standard error) in diagnosing SDB disease (RDI ≥ 15) are 90.3(3.1)% and 88.3(5.5)% when nasal pressure is used alone and together with the thoracic movement it was 89.5(3.7)% and 100.0(0.0)%. Present results suggest that RQA of a single respiratory measurement has potential to be used in an automated SDB screening device, while with dual-channel more reliable accuracy can be expected. Improvements may be possible by including other RQA based features and optimisation of the parameters.

Sleep disordered breathing (SDB) is a highly prevalent but an under-diagnosed disease. Among adults in the ages between 30 to 60 years, 24% of males and 9% of females show conditions of SDB, while 82% of men and 93% of women with moderate to severe SDB remain undiagnosed. Polysomnography (PSG) is the reference diagnostic test for SDB. During PSG, a number of physiological signals are recorded during an overnight sleep and then manually scored for sleep/wake stages and SDB events to obtain the reference diagnosis. The manual scoring of SDB events is an extremely time consuming and cumbersome task with high inter- and intra-rater variations. PSG is a labour intensive, expensive and patient inconvenient test. Further, PSG facilities are limited leading to long waiting lists. There is an enormous clinical need for automation of PSG scoring and an alternative automated ambulatory method suitable for screening the population. During the work of this thesis, we focus (1) on implementing a framework that enables more reliable scoring of SDB events which also lowers manual scoring time, and (2) implementing a reliable automated screening procedure that can be used as a patient-friendly home based study. The recordings of physiological measurements obtained during patients’ sleep of- ten suffer from data losses, interferences and artefacts. In a typical sleep scoring session, artifact-corrupted signal segments are visually detected and removed from further consideration. We developed a novel framework for automated artifact detection and signal restoration, based on the redundancy among respiratory flow signals. The signals focused on are the airflow (thermistor sensors) and nasal pressure signals that are clinically significant in detecting respira- tory disturbances. We treat the respiratory system as a dynamical system, and use the celebrated Takens embedding theorem as the theoretical basis for sig- nal prediction. In this study, we categorise commonly occurring artefacts and distortions in the airflow and nasal pressure measurements into several groups and explore the efficacy of the proposed technique in detecting/recovering them. Results we obtained from a database of clinical PSG signals indicated that theproposed technique can detect artefacts/distortions with a sensitivity >88% and specificity >92%. This work has the potential to simplify the work done by sleep scoring technicians, and also to improve automated sleep scoring methods. During the next phase of the thesis we have investigated the diagnostic ability of single – and dual–channel respiratory flow measuring devices. Recent studies have shown that single channel respiratory flow measurements can be used for automated diagnosis/screening for sleep disordered breathing (SDB) diseases. Improvements for reliable home-based monitoring for SDB may be achieved with the use of predictors based on recurrence quantification analysis (RQA). RQA essentially measures the complex structures present in a time series and are relatively independent of the nonlinearities present in the respiratory measurements such as those due to breathing nonlinearities and sensor movements. The nasal pressure, thermistor-based airflow, abdominal movement and thoracic movement measurements obtained during Polysomnography, were used in this study to implement an algorithm for automated screening for SDB diseases. The algorithm predicts SDB-affected measurement segments using twelve features based on RQA, body mass index (BMI) and neck circumference using mixture discriminant analysis (MDA). The rate of SDB affected segments of data per hour of recording (RDIS) is used as a measure for the diagnosis of SDB diseases. The operating points to be chosen were the prior probability of SDB affected data segments (π1) and the RDIS threshold value, above which a patient is predicted to have a SDB disease. Cross-validation with five-folds, stratified based on the RDI values of the recordings, was used in estimating the operating points. Sensitivity and specificity rates for the final classifier were estimated using a two-layer assessment approach with the operating points chosen at the inner layer using five-fold cross-validation and the choice assessed at the outer layer using repeated learning-testing. The nasal pressure measurement showed higher accuracy compared to other respiratory measurements when used alone. The nasal pressure and thoracic movement measurements were identified as the best pair of measurements to be used in a dual channel device. The estimated sensitivity and specificity (standard error) in diagnosing SDB disease (RDI ≥ 15) are 90.3(3.1)% and 88.3(5.5)% when nasal pressure is used alone and together with the thoracic movement it was 89.5(3.7)% and 100.0(0.0)%. Present results suggest that RQA of a single respiratory measurement has potential to be used in an automated SDB screening device, while with dual-channel more reliable accuracy can be expected. Improvements may be possible by including other RQA based features and optimisation of the parameters.

Diss. (sammanfattning) Stockholm : Karol. inst., 2004. / Härtill 5 uppsatser.

Diss. (sammanfattning) Linköping : Linköpings universitet, 2004. / Härtill 5 uppsatser.

Contexte & Motivation:Les installations électriques des bâtiments se détériorent au fil du temps et leur gravité et leur taux de détérioration dépendent de facteurs environnementaux (chaleur, humidité, réactions chimiques corrosives et vieillies isolations) ou d'actions externes indésirables telles que la manipulation humaine erronée, qui conduit à des charges ou des câbles/réseaux endommagés.L'European Fire Academy (EFA) et de nombreuses compagnies d'assurance indiquent que 25% des incendies de bâtiments sont d'origine électrique. Ces incendies peuvent être déclenchés par des circuits surchargés, des court-circuits, des courants de fuite à la terre, des surtensions et / ou des défauts d'arc électrique dans les connexions et les câbles.Les protections électriques classiques telles que les disjoncteurs et les disjoncteurs différentiels offrent une protection insuffisante. Par exemple, en cas de défauts d'arcs en série, la valeur du courant de défaut d'arc reste inférieure à la valeur du courant nominal, car elle est limitée par la résistance du carbone généré par le défaut d'arc. Dans ce cas, toute protection existante peut détecter ce type de faute.Détection de défaut d'arc : approche de traitement du signalDans le cadre de ce travail, l'objectif a été de détecter chaque instant d'arc, ce qui, pour un réseau alternatif, permettrait d'identifier correctement chaque arc dans chaque demi-cycle de réseau où il se produit.En fonction de la caractéristique numérique utilisée à des fins de détection, nous avons introduit différentes classes de méthodes:• Caractéristiques énergétiques (bandes étroite et large bande)• Caractéristiques statistiques (moments statistiques, analyse de la corrélation etc.)• Caractéristiques basées sur un model (ex. modelés AR, ARMA etc.)• Caractéristiques data-driven (utiliser Phase Space Embedding pour les séries temporelles)Chaque approche a été testée et évaluée sur une base de données de signaux construite avec soin, capable de fournir la variabilité du monde réel, dans un cadre d'évaluation statistique qui permet de trouver des seuils appropriés et leurs plages associées. Il donne également des performances relatives, d'une fonctionnalité à l'autre, en fonction de la façon dont les plages de seuils couvrent tout l'espace des caractéristiques.Une approche prometteuse est montrée avec un résultat intermédiaire sur la figure 8. La configuration est plutôt courante, avec une charge résistive (R-Load) en fonctionnement normal, avec un gradateur allumé et ajouté dans la configuration et un arc persistant apparaissant dans le circuit.Il suffit d'analyser simplement la forme d'onde du courant 50 Hz, car même lors d'une simple inspection visuelle, il est difficile d'identifier l'origine du défaut d'arc et s'il est stable ou s'il s'éteint après (ou où). En mesurant correctement le bruit de défaut d'arc haute fréquence et en sélectionnant correctement la bande passante, nous parvenons à obtenir un signal beaucoup plus facile à traiter. L'arc est difficile à détecter en raison de la variation de l'intensité énergétique d'un réseau à l'autre (encore plus: pour un même réseau, ajouter / enlever des charges ou des rallonges modifie la distribution d'amplitude et de fréquence de l'arc). Par conséquent, nous exploitons le caractère aléatoire intrinsèque de l'arc, ce qui permet une variabilité suffisante d'une réalisation d'arc à une autre.En conclusion, nous proposons une nouvelle méthodologie de traitement du signal pour la détection des défauts d'arc, à mettre en œuvre dans un algorithme de produit AFDD. En outre, une autre approche est présentée, basée sur l'analyse de diagramme de phase, qui permet la séparation entre les arcs et les signaux de communication, ce qui est également un grand défi dans ce domaine. / Context & Motivation:Electrical installations in buildings deteriorate, over time and the severity and rate of deterioration depend on environmental factors (such as heat, humidity, corrosive chemical reactions and aging insulations) and unwanted external actions (such as human mishandling, that leads to damaged devices or cables/network).Caution is mandatory when handling electrical installations, seeing that potential hazards include electric shocks, burns, explosions and fire, if proper safety precautions are ignored or neglected. The European Fire Academy (EFA) and many property and casualty insurance companies report that 25% of building fires are electrical in origin. These fires can be triggered by overloaded circuits, short-circuits, earth leakage currents, overvoltage and/or electrical arc faults in connections and cables.Classical electrical protection such as circuit breakers and RCDs offer insufficient protection. For example, in case of series arc faults, the arc fault current value remains below the rated current value, since it is limited by the resistance of the carbon generated by the arc fault and by the load itself. In this case, no existing protection can detect such kind of fault.Arc Fault Detection: Signal Processing ApproachIn the context of this work, the objective has been to detect each instant of arcing, which for an AC network, would mean correctly identifying each arcing in each network half-cycle where it occurs.Depending on the numerical feature used for detection purposes, we introduced different classes of methods:• Energy-related features (narrow and wideband)• Statistical features (statistical moments, correlation analysis etc.)• Model-based features (using numerical models, such as AR, for example)• Data-driven features (using Phase Space Embedding for time series)Each approach has been tested & evaluated on a carefully constructed signal database, capable of supplying real-world variability, within a statistical evaluation framework which enables finding suitable thresholds and their appropriate ranges. It also gives relative performances, from one feature to another, based on how threshold ranges cover the entire feature space.A promising approach is shown with an intermediary result in Figure 9. The configuration is rather common, with a resistive load (R – Load) in normal operation, with a dimmer being turned on and added in the configuration and a persistent arc appearing in the circuit.Figure 9 Resistive load, dimmer and persistent arcing – processing result (example).Simply analyzing the 50Hz line current waveform is insufficient, as even at a simple visual inspection there is difficulty in identifying where the arc fault ignites and if it is a stable one, or if it extinguishes afterwards (or where). By correctly measuring the high frequency arc fault noise and with correct selection of the bandwidth, we manage to obtain a signal much easier to process further on. Arcing is inherently difficult to detect, due to high frequency energy intensity variation from one network to another (even more: for the same network, adding/removing loads or extension cords will change the amplitude and frequency distribution of the arc fault energy). Therefore, we exploit the intrinsic randomness of arcing, which enables sufficient variability from one arcing realization to another.To conclude, we propose a new signal processing methodology for arc fault detection, to be implemented in an AFDD product algorithm. Also, another approach is presented, based on phase diagram analysis, that allows the separation between the arcs and communication signals, which is also a great challenge in this field.

Le modèle du tas de sable est un modèle de diffusion discret et isotrope introduit par les physiciens Bak, Tang et Wiesenfeld comme illustration de la criticalité auto-organisée. Pour tout graphe, souvent supposé fini, Dhar a formalisé de nombreuses propriétés simplifiant son analyse. Cette thèse propose des études de ce modèle sur la grille bidimensionnelle usuelle et certains de ses sous-graphes également infinis que sont les bandes bi-infinies de hauteur finie. Des approximations du comportement de la pile de sable peuvent se rapprocher de certains modèles de bootstrap percolation avec un support de stabilisation rectangulaire. Les lois sur son demi-périmètre peuvent se décrire à l’aide de statistiques sur les permutations. Un sous-produit de ce travail fait apparaître une différence de deux séries génératrices comptant des permutations selon deux statistiques mahoniennes classiques dont est extrait un polynôme à coefficients entiers et surtout positifs. La suite de cette thèse revisite dans le cadre de ces graphes infinis, des structures jusque-là bien définies uniquement dans le cas des graphes finis, notamment la récurrence. Dans le modèle sur une bande de hauteur finie H, l’existence donnée par Járai et Lyons d’automates finis reconnaissant les configurations récurrentes lues colonne par colonne est étendu par une construction explicite d’automates avec un nombre moindre d’états, se rapprochant de la conjecture de Gamlin. Dans une seconde approche, l’étude se concentre sur les configurations sur la grille entière qui sont périodiques dans les deux directions. Le puits, un sommet du graphe garantissant la terminaison de la stabilisation, est placé à l’infini dans une direction de pente rationnelle. Ceci permet à la fois de préserver la bipériodicité et de proposer une forme affaiblie du critère de Dhar caractérisant ainsi par un algorithme effectif les configurations récurrentes. Ces configurations récurrentes bipériodiques sont des candidates naturelles pour être les éléments de sous-groupes finis de l’éventuel groupe du tas de sable sur la grille. Des éléments de construction de cette loi de groupe donnent expérimentalement quelques sous-groupes finis. / The sandpile model is a discrete model for diffusion of grains on a graph introduced by physicists Bak, Tang and Wiesenfeld as an illustration for self-organised criticality. For any finite graph, Dhar identified many of its numerous structures which simplify its analysis. This thesis focus on the usual square lattice and its subgraphs which are strips of height H, both notions of infinite graphs. Approximations on the behaviour of the stabilisation of a large stack of grains at the origin of the square lattice lead to some random distribution of grains, which stabilisation is connected to some models of bootstrap percolation where modified vertices by this stabilisation forms a rectangle. The laws of the half-perimeter of this rectangle are described by statistics on permutations. As a byproduct, the difference between the generating functions over some permutations of two classical mahonian statistics on permutations appears to mainly be a polynomial with coefficients which are integers and especially positive. Then, this thesis visits in the case of the studied infinite graphs some well-defined structures on finite graphs, in particular the recurrence. In the model on an horizontal strip of height H, we extend the existence of finite automata recognizing recurrent configurations read column by column presented by Járai and Lyons to new automata with significantly less states and these numbers are closer to a conjecture due to Gamlin. An implementation leads to explicit automata for heights 3 and 4 while up to now only the case 2 was obtained by hand. In a second approach, we consider the configurations on the twodimensional square lattice which are periodic in two directions. We suggest to place the sink ensuring that the stabilisation ends at infinity in a direction of rational slope which allows to preserve biperiodicity and a weaker form of Dhar criterion for recurrent configurations. Hence we obtain an effective algorithm defining recurrent configurations among the biperiodic and stable configurations. These biperiodic and recurrent configurations are natural candidates for being the elements of finite subgroups of the hypothetical group on configurations of the sandpile model on the square lattice. We discuss some notions allowing the definition of the law of such a group and experimentally provide some finite subgroups.

O câncer colorretal é a terceira neoplasia mais freqüente no mundo ocidental. No reto encontram-se 30 a 57% dos casos, sendo 80% destes passíveis de alcançar pelo exame digital do reto. Apesar de inúmeros avanços diagnósticos e terapêuticos, ainda são detectados tardiamente. A sobrevida em 5 anos mantém-se em torno de 50%, e os aspectos clínicopatológicos são os critérios prognósticos disponíveis. O presente estudo objetiva avaliar a sobrevida e os fatores clínico-patológicos relacionados aos tumores de reto em nosso meio. Os prontuários de 112 pacientes com adenocarcinoma primário e único de reto submetidos a cirurgia eletiva, num período de 7 anos (1988 a 1995), foram revisados retrospectivamente quanto a: idade, sexo, CEA pré-operatório, tipo e curabilidade da cirurgia, recidiva, seguimento, sobrevida e histopatologia do tumor. Para análise da sobrevida utilizou-se o método de Kaplan-Meyer. Nas análises bivariada e estratificada, um valor de P <0,05 foi considerado significativo. O coeficiente de correlação de Kendall foi utilizado para comparação dos sistemas de estadiamento. No modelo multivariado, utilizou-se um IC de 90%. A média de idade foi de 62,03±14,37 anos, com 52% dos casos do sexo feminino.Sessenta e dois por cento dos tumores encontravam-se no reto distal. Foram submetidos a amputação do reto 56,3% dos casos, e os demais, a retossigmoidectomia, sendo 25 procedimentos não-curativos. Sessenta e quatro pacientes (57%) apresentaram recidiva tumoral (14,3% locorregional, 32,1% à distância, e 10,7% ambas). Quarenta e cinco pacientes (40%) faleceram da neoplasia. O tempo mediano de seguimento foi de 35,27 meses (14,5 – 57,63). A sobrevida em 5 anos foi de 51%, tendo reduzido proporcionalmente ao avanço dos estágios histopatológicos da doença (P<0,001). O sistema de estadiamento de Dukes/Astler- Coller se correlacionou significativamente com o sistema TNM (τ=0,91). O adenocarcinoma 6 moderadamente diferenciado foi o mais freqüente (73,2%). A maioria dos tumores (68,2%) estendia-se até os tecidos perirretais; apenas 3,8% eram restritos à mucosa. Sessenta e sete pacientes tinham linfonodos positivos (29,4%, N1; e 30,3%, N2). Na classificação de Dukes/Astler-Coller, 14 eram estágio D; 55, C1 e C2; 15, B2; e 28, B1 e A. A localização do tumor no reto médio ou distal não influenciou a ocorrência dos óbitos, as recidivas, nem a curabilidade da cirurgia. Os tumores menos diferenciados (P=0,009), com maior penetração na parede (P=0,013), com envolvimento linfonodal (N2>N1, P<0,001), com cirurgia nãocurativa (P=0,002) e os que apresentaram recidiva (P<0,001) influenciaram significativamente a mortalidade. A classificação de Dukes/Astler-Coller e a diferenciação tumoral (P=0,089) foram fatores prognósticos independentes, bem como a penetração do tumor na parede retal (P=0,091) e o comprometimento linfonodal (P<0,001), quando excluída a classificação histopatológica. Os achados deste estudo estão de acordo com a literatura. Além da diferenciação tumoral, os fatores prognósticos identificados correspondem aos níveis de classificação dos sistemas de estadiamento vigentes, concordando com a observação de que estes ainda são os critérios prognósticos disponíveis. Este relato corrobora a predominância dos tumores em estágios avançados ao diagnóstico e a ausência de valor prognóstico da localização do tumor no reto. / Colorectal cancer is the third most common neoplasia in the Western world. About 30 to 57% of cases occur in the rectum, and, among these, 80% can be reached by digital rectal examination. However, most of them are diagnosed in advanced stages. Five-year survival rate is maintained at about 50% and clinicopathological features are still the prognostic criteria available. To evaluate survival and clinicopathological factors, the records of 112 patients submitted to elective resection of adenocarcinoma of the rectum from 1988 to 1995 were reviewed. Data were analysed as for: age, gender, preoperative serum level of CEA, type and curability of surgery, recurrence, follow-up, survival and tumor histopathology. Kaplan- Meyer method was used to analyse survival. Statistical significance on bivariate and stratified analysis was considered for a P value less than 0,05. Kendall correlation coefficient was used to compare histopathologic classifications. In the multivariate model, a 90% confidence interval was considered significant. Mean (SD) age was 62 (14) years and 52% of patients were female. Sixty-two percent of rectal tumors were distally located. Abdominoperineal resection was performed in 56% of cases and the remaining patients underwent low anterior resection. Twenty-five (22%) of these were non-curative procedures.Overall, recurrence was observed in 64 (57%) patients: local in 14%, distant in 32% and both in 10%. Forty-five patients (40%) died from disease. Median (range) follow-up was 35 (14 - 57) months. Fiveyear survival rate was 51% and reduced significantly by tumor progression (histopathologic stages; P<0,001). Both stage systems were well correlated (τ=0,91). Moderately differentiated adenocarcinoma was the most frequent tumor grade (73%). Most tumors (68%) extended to perirectal tissues, only 4% were confined to the mucosa. Positive lymph nodes were observed in 67 patients (30% each, N1 and N2). Fourteen patients were Dukes/Astler-Coller D stage; 55 were C1 or C2; 15 were B2; and 28 were B1 or A. Tumor location had no influence on deaths, recurrences or curability. On bivariate analysis, tumor grade (P=0,009), depth (P=0,013) or recurrence (P<0,001), lymph node involvement (N2>N1, P<0,001), noncurative procedure (P=0,002) related with poorer outcome. On multivariate analyses, Dukes/Astler-Coller stages and tumor grade (P=0,089) were found to be independent prognostic factors, as well as depth of invasion and lymph node involvement, when excluding Dukes staging (P=0,091 and <0,001, respectively). These findings are similar to those reported in the literature. Besides tumor grade, prognostic factors identified meet classification levels on current staging systems. Accordingly, these criteria are still the prognostic factors available. The present report corroborates the predominance of advanced stage tumours at diagnosis and the lack of prognostic value of tumor location in the rectum.