[Shape analysis of retroperitoneal tumors in childhood in magnetic resonance imaging]. / Formanalyse in der Magnetresonanztomografie - Landmarken und Objektdifferenzierung bei retroperitonealen Tumoren im Kindesalter.

2D-shape analysis of biological objects is described first in 2007 with MRI-data (magnetic resonance imaging) of renal tumours of infancy. For shape analysis the evaluation of landmarks is necessary (n >2). In this study 24 landmarks are selected. Every object is described by these landmarks. The shape is the standardised and centred object. The procedure is applied on transversal as well as on frontal images. The results for frontal and transversal images are compared. Tumours of different origin, topography and size can be analysed. The differentiation of relevant landmarks is important for statistical and medical reason. In this study, evaluation of landmarks and their possibility for tumour differentiation is demonstrated.

[Statistical analysis of renal tumours with infants]. / Statistische Analyse der Form von Nierentumoren bei Kleinkindern.

By using medical imaging a diagnosis for a patient is to be given. The shape of three-dimensional medical relevant objects in the human body is the basis for an accurate diagnosis. Based on the correct diagnosis, the appropriate therapy can be decided. In the following study the shape of renal tumours is explored by the mathematical method of shape analysis and by the test of Ziezold (1994). There are n=24 patients whose frontal view records are available. The possibility of classification can be presented only in an exploratory manner, due to the low case numbers of patients.

[Explanation and forecast: relapse of juvenile offenders]. / Erklirung und Vorhersage von "Rückfälligkeit" bei jugendlichen Straftätern.

On the basis of n=82 juvenile offenders from a prison for juvenile offenders in Rheinland Pfalz the model of the logistic regression is compared with a procedure from the family of the neural nets in its efficiency to explain and predict "relapse" in form of a renewed imprisonment or prosecution /police search after dismissal. The group which can be examined is limited by the population of the prison for juvenile offenders and the explaining variables for "relapse" as "addicted to drugs" present non-metric scaling. For the explanation only probabilities for "relapse" can be indicated in this connection. By means of this probability it is possible to classify the individual case. The forecast is simulated by coincidental dividing of the data: the first part of the data is used for the explanation, the second for the forecast. With the comparison of the logistic regression with the neural nets, the superiority of neural nets in the explanation of "relapse" can be shown, since the neural nets are able to consider dependence between the explaining variables and according to that they offer a differentiated explanation. Their efficiency to predict "relapse" depends on the comparability of the distribution in the two coincidentally provided samples, the training data record for determining the explanation and the test case for the use of the explanation regarding the forecast. For optimal explanation and forecast neural nets are to be preferred to the logistic regression, since in the model with the better explanation also includes the potential for a usable better forecast. Moreover the model of the logistic regression is in fact a special case of the neural net, with a reduced complexity of the net.