Random forest of perfect trees: concept, performance, applications and perspectives.

MOTIVATION: The principle of Breiman's random forest (RF) is to build and assemble complementary classification trees in a way that maximizes their variability. We propose a new type of random forest that disobeys Breiman's principles and involves building trees with no classification errors in very large quantities. We used a new type of decision tree that uses a neuron at each node as well as an in-innovative half Christmas tree structure. With these new RFs, we developed a score, based on a family of ten new statistical information criteria, called Nguyen information criteria (NICs), to evaluate the predictive qualities of features in three dimensions. RESULTS: The first NIC allowed the Akaike information criterion to be minimized more quickly than data obtained with the Gini index when the features were introduced in a logistic regression model. The selected features based on the NICScore showed a slight advantage compared to the support vector machines-recursive feature elimination (SVM-RFE) method. We demonstrate that the inclusion of artificial neurons in tree nodes allows a large number of classifiers in the same node to be taken into account simultaneously and results in perfect trees without classification errors. AVAILABILITY AND IMPLEMENTATION: The methods used to build the perfect trees in this article were implemented in the 'ROP' R package, archived at https://cran.r-project.org/web/packages/ROP/index.html. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Beds Simulator 1.0: a software for the modelisation of the number of beds required for a hospital department.

The determination of the number of beds needed for a hospital department is a complex problem that try to take into account efficiency, forecasting of needs, appropriateness of stays. Health authority used methods based on ratios that do not take into account local specificities and use rather to support an economic decision. On the other side, the models developed are too specific to be applied to all type of hospital department. Moreover, all the solutions depend on the LoS (Length of Stay). We have developed a non parametric method to solve this problem. This modelisation was successfully tested in teaching and non teaching hospitals, for an Intensive Care Unit, two Internal Medicine and a surgical departments. A software easy to use was developed, working on Windows available on our website www.sante.univ-nantes.fr/med/stat/.

Development of an individual score for melanoma risk.

Melanoma is one of the fastest growing cancers worldwide. We need to have tools to identify patients with high risk of melanoma. We carried out a case-control study and tested three methods to develop an individual score of melanoma risk, usable in routine practice. All cases included newly diagnosed invasive cutaneous melanoma of stage I or II (6th American Joint Committee on Cancer) seen in 2007 at the Skin Cancer Unit of Nantes Hospital, France. Controls included 1500 consecutive patients consulting their general practitioners. A self-administrated questionnaire was used for assessment of melanoma risk factors. Three methods of scoring were used and compared: one with common relative risks reported in the literature, one with odds ratios estimated by logistic regression, and a combinatorial analysis. The method based on combinatorial analysis permitted one to obtain a simple rule to define individuals at risk: the association of the rule 'presence of at least three risk factors or presence of more than 20 naevi on the arms' for the patients aged under 60 years and 'presence of at least three risk factors or presence of freckles' for the patients aged 60 years and above (sensitivity: 63.2% and specificity: 68.8%). The tool we propose is easy to use every day in routine health care to select patients with high risk of melanoma. It can be assessed without any computer or calculator and is based on the self-assessment of the melanoma risk factors by the patient and thus is not medical time consuming.

Transtorno afetivo recorrente a ciclos rápidos em idosos: um subtipo clínico ou uma forma específica de transtorno bipolar? / Rapid cycling affective disorder in the elderly: a clinical subtype or a specific course of manic-depressive illness

Transtorno afetivo bipolar a ciclos rápidos é raramente encontrado em pacientes idosos. Quatro casos de transtorno afetivo recorrente a ciclos rápidos em mulheres idosas (idades entre 78 e 86 anos) säo apresentados. Duas delas iniciaram o transtorno bibolar quando ainda jovens (30 e 49 anos), mas estas passaram a apresentar ciclos rápidos mais tarde somente (76 e 82 anos). As duas outras pacientes começaram a apresentar transtorno bipolar a ciclos rápidos desde o início da doença (62 e 66 anos). Somente uma delas apresentava um transtorno mental orgânico associado ao transtorno afetivo com subsequente aparecimento de ciclos rápidos parece responder melhor ao tratamento profilático, respectivamente à carbamazepina e ao valproato