Analysis, modelling and classification of geospatial data using machine learning

The research considers the problem of spatial data classification using machine learning algorithms: probabilistic neural networks (PNN) and support vector machines (SVM). As a benchmark model simple k-nearest neighbor algorithm is considered. PNN is a neural network reformulation of well known nonparametric principles of probability density modeling using kernel density estimator and Bayesian optimal or maximum a posteriori decision rules. PNN is well suited to problems where not only predictions but also quantification of accuracy and integration of prior information are necessary. An important property of PNN is that they can be easily used in decision support systems dealing with problems of automatic classification. Support vector machine is an implementation of the principles of statistical learning theory for the classification tasks. Recently they were successfully applied for different environmental topics: classification of soil types and hydro-geological units, optimization of monitoring networks, susceptibility mapping of natural hazards. In the present paper both simulated and real data case studies (low and high dimensional) are considered. The main attention is paid to the detection and learning of spatial patterns by the algorithms applied.

Soil spectral library and its use in soil classification

Soil science has sought to develop better techniques for the classification of soils, one of which is the use of remote sensing applications. The use of ground sensors to obtain soil spectral data has enabled the characterization of these data and the advancement of techniques for the quantification of soil attributes. In order to do this, the creation of a soil spectral library is necessary. A spectral library should be representative of the variability of the soils in a region. The objective of this study was to create a spectral library of distinct soils from several agricultural regions of Brazil. Spectral data were collected (using a Fieldspec sensor, 350-2,500 nm) for the horizons of 223 soil profiles from the regions of Matão, Paraguaçu Paulista, Andradina, Ipaussu, Mirandópolis, Piracicaba, São Carlos, Araraquara, Guararapes, Valparaíso (SP); Naviraí, Maracajú, Rio Brilhante, Três Lagoas (MS); Goianésia (GO); and Uberaba and Lagoa da Prata (MG). A Principal Component Analysis (PCA) of the data was then performed and a graphic representation of the spectral curve was created for each profile. The reflectance intensity of the curves was principally influenced by the levels of Fe2O3, clay, organic matter and the presence of opaque minerals. There was no change in the spectral curves in the horizons of the Latossolos, Nitossolos, and Neossolos Quartzarênicos. Argissolos had superficial horizon curves with the greatest intensity of reflection above 2,200 nm. Cambissolos and Neossolos Litólicos had curves with greater reflectance intensity in poorly developed horizons. Gleisols showed a convex curve in the region of 350-400 nm. The PCA was able to separate different data collection areas according to the region of source material. Principal component one (PC1) was correlated with the intensity of reflectance samples and PC2 with the slope between the visible and infrared samples. The use of the Spectral Library as an indicator of possible soil classes proved to be an important tool in profile classification.

A Computerized Method for the Hydrologic Design of Culverts, February 1974

Nationwide, about five cents of each highway construction dollar is spent on culverts. In Iowa, average annual construction costs on the interstate, primary, and federal-aid secondary systems are about $120,000,000. Assuming the national figure applies to Iowa, about $6,000,000 are spent on culvert construction annually. For each one percent reduction in overall culvert costs, annual construction costs would be reduced by $60,000. One area of potential cost reduction lies in the sizing of the culvert. Determining the flow area and hydraulic capacity is accomplished in the initial design of the culvert. The normal design sequence is accomplished in two parts. The hydrologic portion consists of the determination of a design discharge in cubic feet per second using one of several available methods. This discharge is then used directly in the hydraulic portion of the design to determine the proper type, size, and shape of culvert to be used, based on various site and design restrictions. More refined hydrologic analyses, including rainfall-runoff analysis, flood hydrograph development, and streamflow routing techniques, are not pursued in the existing design procedure used by most county and state highway engineers.

Highlands of the upper Jequitinhonha valley, Brazil: I - characterization and classification

In the upper Jequitinhonha valley, state of Minas Gerais, Brazi, there are large plane areas known as "chapadas", which are separated by areas dissected by tributaries of the Jequitinhonha and Araçuaí rivers. These dissected areas have a surface drainage system with tree, shrub, and grass vegetation, more commonly known as "veredas", i.e., palm swamps. The main purpose of this study was to characterize soil physical, chemical and morphological properties of a representative toposequence in the watershed of the Vereda Lagoa do Leandro, a swamp near Minas Novas, MG, on "chapadas", the highlands of the Alto Jequitinhonha region Different soil types are observed in the landscape: at the top - Typic Haplustox (LVA), in the middle slope - Xanthic Haplustox (LA), at the footslope - Xanthic Haplustox, gray color, here called "Gray Haplustox" ("LAC") and, at the bottom of the palm swamp - Typic Albaquult (GXbd). These soils were first morphologically described; samples of disturbed and undisturbed soils were collected from all horizons and subhorizons, to evaluate their essential physical and chemical properties, by means of standard determination of Fe, Al, Mn, Ti and Si oxides after sulfuric extraction. The contents of Fe, Al and Mn, extracted with dithionite-citrate-bicarbonate and oxalate treatments, were also determined. In the well-drained soils of the slope positions, the typical morphological, physical and chemical properties of Oxisols were found. The GXbd sample, from the bottom of the palm swamp, is grayish and has high texture gradient (B/A) and massive structure. The reduction of the proportion of crystalline iron compounds and the low crystallinity along the slope confirmed the loss of iron during pedogenesis, which is reflected in the current soil color. The Si and Al contents were lowest in the "LAC" soil. There was a decrease of the Fe2O3/TiO2 ratio downhill, indicating progressive drainage restriction along the toposequence. The genesis and all physical and chemical properties of the soils at the footslope and the bottom of the palm swamp of the "chapadas" of the Alto Jequitinhonha region are strongly influenced by the occurrence of ground water on the surface or near the surface all year long, at present and/or in the past. Total concentrations of iron oxides, Fe d and Fe o in soils of the toposequence studied are related to the past and/or present soil colors and drainage conditions.

Detection of colorectal carcinoma by emission-computerized tomography after injection of 123I-labeled Fab or F(ab')2 fragments from monoclonal anti-carcinoembryonic antigen antibodies.

This clinical study was based on experimental results obtained in nude mice grafted with human colon carcinoma, showing that injected 131I-labeled F(ab')2 and Fab fragments from high affinity anti-carcinoembryonic antigen (CEA) monoclonal antibodies (MAb) gave markedly higher ratios of tumor to normal tissue localization than intact MAb. 31 patients with known colorectal carcinoma, including 10 primary tumors, 13 local tumor recurrences, and 21 metastatic involvements, were injected with 123I-labeled F(ab')2 (n = 14) or Fab (n = 17) fragments from MAb anti-CEA. The patients were examined by emission-computerized tomography (ECT) at 6, 24, and sometimes 48 h after injection using a rotating dual head scintillation camera. All 23 primary tumors and local recurrences except one were clearly visualized on at least two sections of different tomographic planes. Interestingly, nine of these patients had almost normal circulating CEA levels, and three of the visualized tumors weighed only 3-5 g. Among 19 known metastatic tumor involvements, 14 were correctly localized by ECT. Two additional liver and several bone metastases were discovered by immunoscintigraphy. Altogether, 86% of the tumor sites were detected, 82% with F(ab')2 and 89% with Fab fragments. The contrast of the tumor images obtained with Fab fragments suggests that this improved method of immunoscintigraphy has the potential to detect early tumor recurrences and thus to increase the survival of patients. The results of this retrospective study, however, should be confirmed in a prospective study before this method can be recommended for the routine diagnosis of cancer.

Generalized Schmidt decomposition and classification of three-quantum-bit states

We prove for any pure three-quantum-bit state the existence of local bases which allow one to build a set of five orthogonal product states in terms of which the state can be written in a unique form. This leads to a canonical form which generalizes the two-quantum-bit Schmidt decomposition. It is uniquely characterized by the five entanglement parameters. It leads to a complete classification of the three-quantum-bit states. It shows that the right outcome of an adequate local measurement always erases all entanglement between the other two parties.

Computer-generated reminders delivered on paper to healthcare professionals; effects on professional practice and health care outcomes.

BACKGROUND: Clinical practice does not always reflect best practice and evidence, partly because of unconscious acts of omission, information overload, or inaccessible information. Reminders may help clinicians overcome these problems by prompting the doctor to recall information that they already know or would be expected to know and by providing information or guidance in a more accessible and relevant format, at a particularly appropriate time. OBJECTIVES: To evaluate the effects of reminders automatically generated through a computerized system and delivered on paper to healthcare professionals on processes of care (related to healthcare professionals' practice) and outcomes of care (related to patients' health condition). SEARCH METHODS: For this update the EPOC Trials Search Co-ordinator searched the following databases between June 11-19, 2012: The Cochrane Central Register of Controlled Trials (CENTRAL) and Cochrane Library (Economics, Methods, and Health Technology Assessment sections), Issue 6, 2012; MEDLINE, OVID (1946- ), Daily Update, and In-process; EMBASE, Ovid (1947- ); CINAHL, EbscoHost (1980- ); EPOC Specialised Register, Reference Manager, and INSPEC, Engineering Village. The authors reviewed reference lists of related reviews and studies.  SELECTION CRITERIA: We included individual or cluster-randomized controlled trials (RCTs) and non-randomized controlled trials (NRCTs) that evaluated the impact of computer-generated reminders delivered on paper to healthcare professionals on processes and/or outcomes of care. DATA COLLECTION AND ANALYSIS: Review authors working in pairs independently screened studies for eligibility and abstracted data. We contacted authors to obtain important missing information for studies that were published within the last 10 years. For each study, we extracted the primary outcome when it was defined or calculated the median effect size across all reported outcomes. We then calculated the median absolute improvement and interquartile range (IQR) in process adherence across included studies using the primary outcome or median outcome as representative outcome. MAIN RESULTS: In the 32 included studies, computer-generated reminders delivered on paper to healthcare professionals achieved moderate improvement in professional practices, with a median improvement of processes of care of 7.0% (IQR: 3.9% to 16.4%). Implementing reminders alone improved care by 11.2% (IQR 6.5% to 19.6%) compared with usual care, while implementing reminders in addition to another intervention improved care by 4.0% only (IQR 3.0% to 6.0%) compared with the other intervention. The quality of evidence for these comparisons was rated as moderate according to the GRADE approach. Two reminder features were associated with larger effect sizes: providing space on the reminder for provider to enter a response (median 13.7% versus 4.3% for no response, P value = 0.01) and providing an explanation of the content or advice on the reminder (median 12.0% versus 4.2% for no explanation, P value = 0.02). Median improvement in processes of care also differed according to the behaviour the reminder targeted: for instance, reminders to vaccinate improved processes of care by 13.1% (IQR 12.2% to 20.7%) compared with other targeted behaviours. In the only study that had sufficient power to detect a clinically significant effect on outcomes of care, reminders were not associated with significant improvements. AUTHORS' CONCLUSIONS: There is moderate quality evidence that computer-generated reminders delivered on paper to healthcare professionals achieve moderate improvement in process of care. Two characteristics emerged as significant predictors of improvement: providing space on the reminder for a response from the clinician and providing an explanation of the reminder's content or advice. The heterogeneity of the reminder interventions included in this review also suggests that reminders can improve care in various settings under various conditions.

Characterization and classification of soils in the Taquari river basin - Pantanal region, state of Mato Grosso do Sul, Brazil

Among the soils in the Mato Grosso do Sul, stand out in the Pantanal biome, the Spodosols. Despite being recorded in considerable extensions, few studies aiming to characterize and classify these soils were performed. The purpose of this study was to characterize and classify soils in three areas of two physiographic types in the Taquari river basin: bay and flooded fields. Two trenches were opened in the bay area (P1 and P2) and two in the flooded field (P3 and P4). The third area (saline) with high sodium levels was sampled for further studies. In the soils in both areas the sand fraction was predominant and the texture from sand to sandy loam, with the main constituent quartz. In the bay area, the soil organic carbon in the surface layer (P1) was (OC) > 80 g kg-1, being diagnosed as Histic epipedon. In the other profiles the surface horizons had low OC levels which, associated with other properties, classified them as Ochric epipedons. In the soils of the bay area (P1 and P2), the pH ranged from 5.0 to 7.5, associated with dominance of Ca2+ and Mg2+, with base saturation above 50 % in some horizons. In the flooded fields (P3 and P4) the soil pH ranged from 4.9 to 5.9, H+ contents were high in the surface horizons (0.8-10.5 cmol c kg-1 ), Ca2+ and Mg² contents ranged from 0.4 to 0.8 cmol c kg-1 and base saturation was < 50 %. In the soils of the bay area (P1 and P2) iron was accumulated (extracted by dithionite - Fed) and OC in the spodic horizon; in the P3 and P4 soils only Fed was accumulated (in the subsurface layers). According to the criteria adopted by the Brazilian System of Soil Classification (SiBCS) at the subgroup level, the soils were classified as: P1: Organic Hydromorphic Ferrohumiluvic Spodosol. P2: Typical Orthic Ferrohumiluvic Spodosol. P3: Typical Hydromorphic Ferroluvic Spodosol. P4: Arenic Orthic Ferroluvic Spodosol.

Automatic Mapping and Classification of Spatial Environmental Data

The paper deals with the development and application of the generic methodology for automatic processing (mapping and classification) of environmental data. General Regression Neural Network (GRNN) is considered in detail and is proposed as an efficient tool to solve the problem of spatial data mapping (regression). The Probabilistic Neural Network (PNN) is considered as an automatic tool for spatial classifications. The automatic tuning of isotropic and anisotropic GRNN/PNN models using cross-validation procedure is presented. Results are compared with the k-Nearest-Neighbours (k-NN) interpolation algorithm using independent validation data set. Real case studies are based on decision-oriented mapping and classification of radioactively contaminated territories.

A colorectal cancer classification system that associates cellular phenotype and responses to therapy.

Colorectal cancer (CRC) is a major cause of cancer mortality. Whereas some patients respond well to therapy, others do not, and thus more precise, individualized treatment strategies are needed. To that end, we analyzed gene expression profiles from 1,290 CRC tumors using consensus-based unsupervised clustering. The resultant clusters were then associated with therapeutic response data to the epidermal growth factor receptor-targeted drug cetuximab in 80 patients. The results of these studies define six clinically relevant CRC subtypes. Each subtype shares similarities to distinct cell types within the normal colon crypt and shows differing degrees of 'stemness' and Wnt signaling. Subtype-specific gene signatures are proposed to identify these subtypes. Three subtypes have markedly better disease-free survival (DFS) after surgical resection, suggesting these patients might be spared from the adverse effects of chemotherapy when they have localized disease. One of these three subtypes, identified by filamin A expression, does not respond to cetuximab but may respond to cMET receptor tyrosine kinase inhibitors in the metastatic setting. Two other subtypes, with poor and intermediate DFS, associate with improved response to the chemotherapy regimen FOLFIRI in adjuvant or metastatic settings. Development of clinically deployable assays for these subtypes and of subtype-specific therapies may contribute to more effective management of this challenging disease.

Comparison between artificial neural networks and maximum likelihood classification in digital soil mapping

Soil surveys are the main source of spatial information on soils and have a range of different applications, mainly in agriculture. The continuity of this activity has however been severely compromised, mainly due to a lack of governmental funding. The purpose of this study was to evaluate the feasibility of two different classifiers (artificial neural networks and a maximum likelihood algorithm) in the prediction of soil classes in the northwest of the state of Rio de Janeiro. Terrain attributes such as elevation, slope, aspect, plan curvature and compound topographic index (CTI) and indices of clay minerals, iron oxide and Normalized Difference Vegetation Index (NDVI), derived from Landsat 7 ETM+ sensor imagery, were used as discriminating variables. The two classifiers were trained and validated for each soil class using 300 and 150 samples respectively, representing the characteristics of these classes in terms of the discriminating variables. According to the statistical tests, the accuracy of the classifier based on artificial neural networks (ANNs) was greater than of the classic Maximum Likelihood Classifier (MLC). Comparing the results with 126 points of reference showed that the resulting ANN map (73.81 %) was superior to the MLC map (57.94 %). The main errors when using the two classifiers were caused by: a) the geological heterogeneity of the area coupled with problems related to the geological map; b) the depth of lithic contact and/or rock exposure, and c) problems with the environmental correlation model used due to the polygenetic nature of the soils. This study confirms that the use of terrain attributes together with remote sensing data by an ANN approach can be a tool to facilitate soil mapping in Brazil, primarily due to the availability of low-cost remote sensing data and the ease by which terrain attributes can be obtained.

Using numerical classification of profiles based on Vis-NIR spectra to distinguish soils from the Piracicaba Region, Brazil

Considering that information from soil reflectance spectra is underutilized in soil classification, this paper aimed to evaluate the relationship of soil physical, chemical properties and their spectra, to identify spectral patterns for soil classes, evaluate the use of numerical classification of profiles combined with spectral data for soil classification. We studied 20 soil profiles from the municipality of Piracicaba, State of São Paulo, Brazil, which were morphologically described and classified up to the 3rd category level of the Brazilian Soil Classification System (SiBCS). Subsequently, soil samples were collected from pedogenetic horizons and subjected to soil particle size and chemical analyses. Their Vis-NIR spectra were measured, followed by principal component analysis. Pearson's linear correlation coefficients were determined among the four principal components and the following soil properties: pH, organic matter, P, K, Ca, Mg, Al, CEC, base saturation, and Al saturation. We also carried out interpretation of the first three principal components and their relationships with soil classes defined by SiBCS. In addition, numerical classification of the profiles based on the OSACA algorithm was performed using spectral data as a basis. We determined the Normalized Mutual Information (NMI) and Uncertainty Coefficient (U). These coefficients represent the similarity between the numerical classification and the soil classes from SiBCS. Pearson's correlation coefficients were significant for the principal components when compared to sand, clay, Al content and soil color. Visual analysis of the principal component scores showed differences in the spectral behavior of the soil classes, mainly among Argissolos and the others soils. The NMI and U similarity coefficients showed values of 0.74 and 0.64, respectively, suggesting good similarity between the numerical and SiBCS classes. For example, numerical classification correctly distinguished Argissolos from Latossolos and Nitossolos. However, this mathematical technique was not able to distinguish Latossolos from Nitossolos Vermelho férricos, but the Cambissolos were well differentiated from other soil classes. The numerical technique proved to be effective and applicable to the soil classification process.

Advanced Kernel Methods For Remote Sensing Image Classification

Résumé Suite aux recentes avancées technologiques, les archives d'images digitales ont connu une croissance qualitative et quantitative sans précédent. Malgré les énormes possibilités qu'elles offrent, ces avancées posent de nouvelles questions quant au traitement des masses de données saisies. Cette question est à la base de cette Thèse: les problèmes de traitement d'information digitale à très haute résolution spatiale et/ou spectrale y sont considérés en recourant à des approches d'apprentissage statistique, les méthodes à noyau. Cette Thèse étudie des problèmes de classification d'images, c'est à dire de catégorisation de pixels en un nombre réduit de classes refletant les propriétés spectrales et contextuelles des objets qu'elles représentent. L'accent est mis sur l'efficience des algorithmes, ainsi que sur leur simplicité, de manière à augmenter leur potentiel d'implementation pour les utilisateurs. De plus, le défi de cette Thèse est de rester proche des problèmes concrets des utilisateurs d'images satellite sans pour autant perdre de vue l'intéret des méthodes proposées pour le milieu du machine learning dont elles sont issues. En ce sens, ce travail joue la carte de la transdisciplinarité en maintenant un lien fort entre les deux sciences dans tous les développements proposés. Quatre modèles sont proposés: le premier répond au problème de la haute dimensionalité et de la redondance des données par un modèle optimisant les performances en classification en s'adaptant aux particularités de l'image. Ceci est rendu possible par un système de ranking des variables (les bandes) qui est optimisé en même temps que le modèle de base: ce faisant, seules les variables importantes pour résoudre le problème sont utilisées par le classifieur. Le manque d'information étiquétée et l'incertitude quant à sa pertinence pour le problème sont à la source des deux modèles suivants, basés respectivement sur l'apprentissage actif et les méthodes semi-supervisées: le premier permet d'améliorer la qualité d'un ensemble d'entraînement par interaction directe entre l'utilisateur et la machine, alors que le deuxième utilise les pixels non étiquetés pour améliorer la description des données disponibles et la robustesse du modèle. Enfin, le dernier modèle proposé considère la question plus théorique de la structure entre les outputs: l'intègration de cette source d'information, jusqu'à présent jamais considérée en télédétection, ouvre des nouveaux défis de recherche. Advanced kernel methods for remote sensing image classification Devis Tuia Institut de Géomatique et d'Analyse du Risque September 2009 Abstract The technical developments in recent years have brought the quantity and quality of digital information to an unprecedented level, as enormous archives of satellite images are available to the users. However, even if these advances open more and more possibilities in the use of digital imagery, they also rise several problems of storage and treatment. The latter is considered in this Thesis: the processing of very high spatial and spectral resolution images is treated with approaches based on data-driven algorithms relying on kernel methods. In particular, the problem of image classification, i.e. the categorization of the image's pixels into a reduced number of classes reflecting spectral and contextual properties, is studied through the different models presented. The accent is put on algorithmic efficiency and the simplicity of the approaches proposed, to avoid too complex models that would not be used by users. The major challenge of the Thesis is to remain close to concrete remote sensing problems, without losing the methodological interest from the machine learning viewpoint: in this sense, this work aims at building a bridge between the machine learning and remote sensing communities and all the models proposed have been developed keeping in mind the need for such a synergy. Four models are proposed: first, an adaptive model learning the relevant image features has been proposed to solve the problem of high dimensionality and collinearity of the image features. This model provides automatically an accurate classifier and a ranking of the relevance of the single features. The scarcity and unreliability of labeled. information were the common root of the second and third models proposed: when confronted to such problems, the user can either construct the labeled set iteratively by direct interaction with the machine or use the unlabeled data to increase robustness and quality of the description of data. Both solutions have been explored resulting into two methodological contributions, based respectively on active learning and semisupervised learning. Finally, the more theoretical issue of structured outputs has been considered in the last model, which, by integrating outputs similarity into a model, opens new challenges and opportunities for remote sensing image processing.