Search

Einleitung: Lymphome machen nur eine geringe Anzahl aller malignen Neoplasien mit ca. 5% im Hals-, Nasen- und Ohrenbereich aus. Als extranodale Manifestation von Lymphomen steht der Kopf-Halsbereich an zweiter Stelle hinter dem Gastrointestinaltrakt. Am häufigsten tritt das Non-Hodgkin-Lymphom auf, Hodgkin-Lymphome kommen sehr selten vor.Methoden: Es wurden 3 Fälle von Non-Hodgkin-Lymphomen im Kopf-Halsbereich untersucht. Ein Lymphom des Sinus maxillaris et ethmoidalis, der Trachea und des Nasopharynx.Anamnese, Therapie, Histologie und Ergebnisse der interdisziplinären Tumorkonferenz wurden analysiert und erörtert.Ergebnisse: Anamnese und klinische Beschwerden der Patienten weisen zunächst nicht in allen Fällen auf eine maligne Neoplasie hin. Die histologische Differenzierung benötigt weiterführende immunhistochemische Untersuchungen mit Expressionsmarkern. Die Therapie ist abhängig vom Allgemeinzustand des Patienten. Als adjuvante Therapie wird in Abhängigkeit von B-oder T-Zell Lymphom eine Polychemotherapie und zusätzlich eine Radiatio angewendet.Schlussfolgerung: Obwohl Lymphome im Kopf-Hals-Bereich zu den seltenen Tumorentitäten gehören, sollten diese differentialdiagnostisch immer in Betracht gezogen werden. Die Therapie unterscheidet sich erheblich von anderen Tumoren im HNO Bereich, sie ist nicht primär chirurgisch. In den meisten Fällen ist eine primäre Chemotherapie indiziert, Radiotherapie, Antikörpertherapie, Radioimmuntherapie, Zytokine oder Stammzelltherapie kommen zum Einsatz.Die histopathologische Diagnose erfordert immunhistochemische Zusatzuntersuchungen.Der Erstautor gibt keinen Interessenkonflikt an.

Einleitung: Maligne Tumoren der Nasenhaupthöhle und des Nasenneben-höhlensystems machen 12% aller Tumoren des Kopf-Hals-Bereiches aus. In ca. 60% der Fälle handelt es sich dabei um Plattenepithelkarzinome. 1% der malignen Nasennebenhöhlentumoren sind maligne Melanome und 3% Lymphome. Sarkome und kleinzellige Tumoren sind Raritäten und treten in der Literatur lediglich als Einzelfallbeschreibungen auf. Die histologische Diagnose kann durch schlechte Differenzierung der Tumoren erschwert sein, ist jedoch von entscheidender therapeutischer Bedeutung. Material und Methoden: In der vorliegenden Untersuchung werden 3 Fälle von malignen Nasennebenhöhlentumoren mit untypischer histologischer Differenzierung vorgestellt: Sarkom, Non-Hogdkin-Lymophom und kleinzelliges Karzinom. Es werden jeweils Anamnese, klinische Befunde, Operation, histologisches Ergebnis, Ergebnis der interdisziplinären Tumorkonferenz, das weitere Therapiekonzept und der klinische Verlauf vorgestellt. Ergebnis: Aufgrund fehlender Frühsymptome wurden alle Tumoren erst spät diagnostiziert. Sowohl die späte Diagnose, als auch die histologische Klassifizierung waren Ursachen für den schlechten Verlauf. Immunhistochemische Untersuchungen haben zur richtigen Diagnose wesentlich beigetragen. Fazit: Obwohl es sich bei der Mehrzahl der malignen Tumoren des Nasenneben-höhlensystems um Plattenepithelkarzinome handelt, muss differentialdiagnostisch auch an andere Tumorentitäten gedacht und auf diese untersucht werden, um Patienten mit seltenen Histologien ebenfalls eine adäquate Therapie zu ermöglichen.Der Erstautor gibt keinen Interessenkonflikt an.

Recently, a new auditory-based feature extraction algorithm for robust speech recognition in noisy environments was proposed. The new features are derived by mimicking closely the human peripheral auditory process and the filters in the outer ear, middle ear, and inner ear are obtained from psychoacoustics literature with some manual adjustments. In this paper, we extend the auditory-based feature extraction algorithm and propose to further train the auditory-based filters through discriminative training. Using the data-driven approach, we optimize the filters by minimizing the subsequent recognition errors on a task. One significant contribution over similar efforts in the past (generally under the name of

Recently, a new auditory-based feature extraction algorithm for robust speech recognition in noisy environments was proposed. The new features are derived by mimicking closely the human peripheral auditory process and the filters in the outer ear, middle ear, and inner ear are obtained from psychoacoustics literature with some manual adjustments. In this paper, we extend the auditory-based feature extraction algorithm and propose to further train the auditory-based filters through discriminative training. Using the data-driven approach, we optimize the filters by minimizing the subsequent recognition errors on a task. One significant contribution over similar efforts in the past (generally under the name of

Recently, a new auditory-based feature extraction algorithm for robust speech recognition in noisy environments was proposed. The new features are derived by mimicking closely the human peripheral auditory process and the filters in the outer ear, middle ear, and inner ear are obtained from psychoacoustics literature with some manual adjustments. In this paper, we extend the auditory-based feature extraction algorithm and propose to further train the auditory-based filters through discriminative training. Using the data-driven approach, we optimize the filters by minimizing the subsequent recognition errors on a task. One significant contribution over similar efforts in the past (generally under the name of

During minimum-classification-error (MCE) training, competing hypotheses against the correct one are commonly derived by the N-best algorithm. One problem with the N-best algorithm is that, in practice, some misclassified data can have very large misclassification distances from the N-best competitors and fall out of the steep/trainable region of the sigmoid function, and thus cannot be utilized effectively. Although one may alleviate the problem by adjusting the shape of the sigmoid and then using an appropriate learning rate, it requires careful tuning of these training parameters. In this paper, we propose using the nearest competing hypothesis instead of the traditional N-best hypotheses for MCE training. The aim is to keep the training data as close to the trainable region as possible. Consequently, the amount of "effective" training data is increased. Furthermore, by progressively beating the nearest competitors, the training seems to be more stable. We also design an approximation algorithm based on beam search to locate the nearest competing hypothesis efficiently. We compare the performance of MCE training using 1-nearest or 1-best competing hypotheses on the Aurora database and find that the new approach (using 1-nearest hypotheses) reduces the word error rates by 5.1% and 17.8% over the latter (of using the 1-best competing hypotheses) and the official Aurora baseline respectively.

During minimum-classification-error (MCE) training, competing hypotheses against the correct one are commonly derived by the N-best algorithm. One problem with the N-best algorithm is that, in practice, some misclassified data can have very large misclassification distances from the N-best competitors and fall out of the steep/trainable region of the sigmoid function, and thus cannot be utilized effectively. Although one may alleviate the problem by adjusting the shape of the sigmoid and then using an appropriate learning rate, it requires careful tuning of these training parameters. In this paper, we propose using the nearest competing hypothesis instead of the traditional N-best hypotheses for MCE training. The aim is to keep the training data as close to the trainable region as possible. Consequently, the amount of "effective" training data is increased. Furthermore, by progressively beating the nearest competitors, the training seems to be more stable. We also design an approximation algorithm based on beam search to locate the nearest competing hypothesis efficiently. We compare the performance of MCE training using 1-nearest or 1-best competing hypotheses on the Aurora database and find that the new approach (using 1-nearest hypotheses) reduces the word error rates by 5.1% and 17.8% over the latter (of using the 1-best competing hypotheses) and the official Aurora baseline respectively.

During minimum-classification-error (MCE) training, competing hypotheses against the correct one are commonly derived by the N-best algorithm. One problem with the N-best algorithm is that, in practice, some misclassified data can have very large misclassification distances from the N-best competitors and fall out of the steep/trainable region of the sigmoid function, and thus cannot be utilized effectively. Although one may alleviate the problem by adjusting the shape of the sigmoid and then using an appropriate learning rate, it requires careful tuning of these training parameters. In this paper, we propose using the nearest competing hypothesis instead of the traditional N-best hypotheses for MCE training. The aim is to keep the training data as close to the trainable region as possible. Consequently, the amount of "effective" training data is increased. Furthermore, by progressively beating the nearest competitors, the training seems to be more stable. We also design an approximation algorithm based on beam search to locate the nearest competing hypothesis efficiently. We compare the performance of MCE training using 1-nearest or 1-best competing hypotheses on the Aurora database and find that the new approach (using 1-nearest hypotheses) reduces the word error rates by 5.1% and 17.8% over the latter (of using the 1-best competing hypotheses) and the official Aurora baseline respectively.