Data-driven clustering : new methods and applications

Abstract : This work is concerned with the development and application of novel unsupervised learning methods, having in mind two target applications: the analysis of forensic case data and the classification of remote sensing images. First, a method based on a symbolic optimization of the inter-sample distance measure is proposed to improve the flexibility of spectral clustering algorithms, and applied to the problem of forensic case data. This distance is optimized using a loss function related to the preservation of neighborhood structure between the input space and the space of principal components, and solutions are found using genetic programming. Results are compared to a variety of state-of--the-art clustering algorithms. Subsequently, a new large-scale clustering method based on a joint optimization of feature extraction and classification is proposed and applied to various databases, including two hyperspectral remote sensing images. The algorithm makes uses of a functional model (e.g., a neural network) for clustering which is trained by stochastic gradient descent. Results indicate that such a technique can easily scale to huge databases, can avoid the so-called out-of-sample problem, and can compete with or even outperform existing clustering algorithms on both artificial data and real remote sensing images. This is verified on small databases as well as very large problems. Résumé : Ce travail de recherche porte sur le développement et l'application de méthodes d'apprentissage dites non supervisées. Les applications visées par ces méthodes sont l'analyse de données forensiques et la classification d'images hyperspectrales en télédétection. Dans un premier temps, une méthodologie de classification non supervisée fondée sur l'optimisation symbolique d'une mesure de distance inter-échantillons est proposée. Cette mesure est obtenue en optimisant une fonction de coût reliée à la préservation de la structure de voisinage d'un point entre l'espace des variables initiales et l'espace des composantes principales. Cette méthode est appliquée à l'analyse de données forensiques et comparée à un éventail de méthodes déjà existantes. En second lieu, une méthode fondée sur une optimisation conjointe des tâches de sélection de variables et de classification est implémentée dans un réseau de neurones et appliquée à diverses bases de données, dont deux images hyperspectrales. Le réseau de neurones est entraîné à l'aide d'un algorithme de gradient stochastique, ce qui rend cette technique applicable à des images de très haute résolution. Les résultats de l'application de cette dernière montrent que l'utilisation d'une telle technique permet de classifier de très grandes bases de données sans difficulté et donne des résultats avantageusement comparables aux méthodes existantes.

Free-living energy expenditure measured by two independent techniques in pregnant and nonpregnant Gambian women.

Free-living energy expenditure (EE) was assessed in 37 young pregnant Gambian women at the 12th (n = 11, 53.5 +/- 1.7 kg), 24th (n = 14, 54.7 +/- 2.1 kg), and 36th (n = 12, 65.0 +/- 2.6 kg) wk of pregnancy and was compared with nonpregnant nonlactating (NPNL) control women (n = 12, 50.3 +/- 1.6 kg). The following two methods were used to assess EE: 1) the heart rate (HR) method using individual regression lines (HR vs EE) established at different activity levels in a respiration chamber and 2) the doubly labeled water (2H2(18)O) method in a subgroup of 25 pregnant and 7 control women. With the HR method the EE during the agricultural rainy season was found to be 2,408 +/- 87, 2,293 +/- 122, and 2,782 +/- 130 kcal/day at 12, 24, and 36 wk of gestation and were not significantly different from the control group (2,502 +/- 133 kcal/day). These findings were confirmed by the 2H2(18)O measurements, which failed to show any effect of pregnancy on EE. Expressed per unit body weight, the free-living EE was found to be lower (P less than 0.01 with 2H2(18)O method) at 36 wk of gestation than in the NPNL group. It is concluded that, in these Gambian women, energy-sparing mechanisms that contribute to meet the additional energy stress of gestation are operating during pregnancy (e.g., diminished spontaneous physical activity).

La validation des techniques en chirurgie laparoscopique [Validation of laparoscopic surgical techniques].

The use of laparoscopic surgery has increased rapidly. However, a technically feasible procedure is not automatically recommendable. Thus, if cholecystectomy and fundoplication are currently fully validated techniques, this does not hold true for gastroplasty and kidney harvesting for transplantation: these operations are feasible indeed but their efficacy remains to be proved. Laparoscopic oncology has been shown to be feasible too, but its efficacy has not been documented yet.

Traitement des tumeurs maligne du foie par radiologie interventionnelle: techniques intra-artérielles [Interventional radiology procedures for malignancies of the liver treatment: Intraarterial procedures].

Intraarterial procedures such as chemoembolization and radioembolization aim for the palliative treatment of advanced hepatocellular carcinoma (stage BCLC B and C with tumoral portal thrombosis). The combination of hepatic intraarterial chemotherapy and systemic chemotherapy can increase the probability of curing colorectal cancer with hepatic metastases not immediately accessible to surgical treatment or percutaneous ablation.

Percutaneous endoscopic gastrostomy in head and neck cancer patients: indications, techniques, complications and results.

The aim of this study was to review our experience in percutaneous endoscopic gastrostomy (PEG) performed in patients with cancer of the upper aerodigestive tract. Descriptive retrospective study of 142 patients (115 males, 27 females), mean age 62.4 years (25-84 years), with head and neck or esophageal cancer, who underwent PEG tube insertion between January 2006 and December 2008. The studied parameters were indications, success rate, rate and type of complications, and their management. Percutaneous endoscopic gastrostomy was inserted before chemoradiation therapy in 80% and during or after cancer treatment in 20% of the patients. PEG placement was possible in 137 patients (96%). Major complications were observed in 9 (7%) and minor complications in 22 (17%) of the 137 patients. Seven of the 9 patients with a major complication needed revision surgery. The mortality directly related to the procedure was 0.7%. Percutaneous endoscopic gastrostomy tube insertion has a high success rate. In patients with upper aerodigestive tract cancer, PEG should be the first choice for enteral nutrition when sufficient oral intake is not possible. Although apparently easy, the procedure may occasionally lead to severe complications. Therefore, a strict technique and knowledge of clinical signs of possible complications are mandatory.

Transobturator surgery for female stress incontinence: a comparative anatomical study of outside-in vs inside-out techniques.

OBJECTIVE To assess the specific risks of injury to neural and vascular structures inherent in two approaches to transobturator surgery for inserting a suburethral sling, i.e. the outside-in (standard technique) and inside-out approaches. MATERIALS AND METHODS The study comprised seven cadavers, providing 14 obturator regions. Five specimens had a tape inserted outside-in on one side, and inside-out on the other; of the remaining two cadavers, one had an inside-out tape and one an outside-in tape, bilaterally. After tape insertion, the cadavers were dissected. Particular attention was paid to the distances between the tape and the deep external pudendal vessels, and between the tape and the posterior branch of the obturator nerve. RESULTS With the inside-out technique, the safety margins were reduced, and the external pudendal vessels and the posterior branch of the obturator nerve were at greater risk of injury. CONCLUSION The two techniques are not equivalent, with a lower risk of injury to vascular and nerve structures with the outside-in technique.

Recent advances in remote sensing image processing

Remote sensing image processing is nowadays a mature research area. The techniques developed in the field allow many real-life applications with great societal value. For instance, urban monitoring, fire detection or flood prediction can have a great impact on economical and environmental issues. To attain such objectives, the remote sensing community has turned into a multidisciplinary field of science that embraces physics, signal theory, computer science, electronics, and communications. From a machine learning and signal/image processing point of view, all the applications are tackled under specific formalisms, such as classification and clustering, regression and function approximation, image coding, restoration and enhancement, source unmixing, data fusion or feature selection and extraction. This paper serves as a survey of methods and applications, and reviews the last methodological advances in remote sensing image processing.