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.

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.

Classification of urban multi-angular image sequences by aligning their manifolds

When dealing with multi-angular image sequences, problems of reflectance changes due either to illumination and acquisition geometry, or to interactions with the atmosphere, naturally arise. These phenomena interplay with the scene and lead to a modification of the measured radiance: for example, according to the angle of acquisition, tall objects may be seen from top or from the side and different light scatterings may affect the surfaces. This results in shifts in the acquired radiance, that make the problem of multi-angular classification harder and might lead to catastrophic results, since surfaces with the same reflectance return significantly different signals. In this paper, rather than performing atmospheric or bi-directional reflection distribution function (BRDF) correction, a non-linear manifold learning approach is used to align data structures. This method maximizes the similarity between the different acquisitions by deforming their manifold, thus enhancing the transferability of classification models among the images of the sequence.

Characterization and classification of matrix effects in biological samples analyses.

An exhaustive classification of matrix effects occurring when a sample preparation is performed prior to liquid-chromatography coupled to mass spectrometry (LC-MS) analyses was proposed. A total of eight different situations were identified allowing the recognition of the matrix effect typology via the calculation of four recovery values. A set of 198 compounds was used to evaluate matrix effects after solid phase extraction (SPE) from plasma or urine samples prior to LC-ESI-MS analysis. Matrix effect identification was achieved for all compounds and classified through an organization chart. Only 17% of the tested compounds did not present significant matrix effects.

Tire traces: discrimination and classification of pyrolysis-GC/MS profiles

Tire traces can be observed on several crime scenes as vehicles are often used by criminals. The tread abrasion on the road, while braking or skidding, leads to the production of small rubber particles which can be collected for comparison purposes. This research focused on the statistical comparison of Py-GC/MS profiles of tire traces and tire treads. The optimisation of the analytical method was carried out using experimental designs. The aim was to determine the best pyrolysis parameters regarding the repeatability of the results. Thus, the pyrolysis factor effect could also be calculated. The pyrolysis temperature was found to be five time more important than time. Finally, a pyrolysis at 650 °C during 15 s was selected. Ten tires of different manufacturers and models were used for this study. Several samples were collected on each tire, and several replicates were carried out to study the variability within each tire (intravariability). More than eighty compounds were integrated for each analysis and the variability study showed that more than 75% presented a relative standard deviation (RSD) below 5% for the ten tires, thus supporting a low intravariability. The variability between the ten tires (intervariability) presented higher values and the ten most variant compounds had a RSD value above 13%, supporting their high potential of discrimination between the tires tested. Principal Component Analysis (PCA) was able to fully discriminate the ten tires with the help of the first three principal components. The ten tires were finally used to perform braking tests on a racetrack with a vehicle equipped with an anti-lock braking system. The resulting tire traces were adequately collected using sheets of white gelatine. As for tires, the intravariability for the traces was found to be lower than the intervariability. Clustering methods were carried out and the Ward's method based on the squared Euclidean distance was able to correctly group all of the tire traces replicates in the same cluster than the replicates of their corresponding tire. Blind tests on traces were performed and were correctly assigned to their tire source. These results support the hypothesis that the tested tires, of different manufacturers and models, can be discriminated by a statistical comparison of their chemical profiles. The traces were found to be not differentiable from their source but differentiable from all the other tires present in the subset. The results are promising and will be extended on a larger sample set.

Paper 5: Surveillance of multiple congenital anomalies: implementation of a computer algorithm in European registers for classification of cases.

BACKGROUND: Surveillance of multiple congenital anomalies is considered to be more sensitive for the detection of new teratogens than surveillance of all or isolated congenital anomalies. Current literature proposes the manual review of all cases for classification into isolated or multiple congenital anomalies. METHODS: Multiple anomalies were defined as two or more major congenital anomalies, excluding sequences and syndromes. A computer algorithm for classification of major congenital anomaly cases in the EUROCAT database according to International Classification of Diseases (ICD)v10 codes was programmed, further developed, and implemented for 1 year's data (2004) from 25 registries. The group of cases classified with potential multiple congenital anomalies were manually reviewed by three geneticists to reach a final agreement of classification as "multiple congenital anomaly" cases. RESULTS: A total of 17,733 cases with major congenital anomalies were reported giving an overall prevalence of major congenital anomalies at 2.17%. The computer algorithm classified 10.5% of all cases as "potentially multiple congenital anomalies". After manual review of these cases, 7% were agreed to have true multiple congenital anomalies. Furthermore, the algorithm classified 15% of all cases as having chromosomal anomalies, 2% as monogenic syndromes, and 76% as isolated congenital anomalies. The proportion of multiple anomalies varies by congenital anomaly subgroup with up to 35% of cases with bilateral renal agenesis. CONCLUSIONS: The implementation of the EUROCAT computer algorithm is a feasible, efficient, and transparent way to improve classification of congenital anomalies for surveillance and research.

Classification of current anticancer immunotherapies.

During the past decades, anticancer immunotherapy has evolved from a promising therapeutic option to a robust clinical reality. Many immunotherapeutic regimens are now approved by the US Food and Drug Administration and the European Medicines Agency for use in cancer patients, and many others are being investigated as standalone therapeutic interventions or combined with conventional treatments in clinical studies. Immunotherapies may be subdivided into "passive" and "active" based on their ability to engage the host immune system against cancer. Since the anticancer activity of most passive immunotherapeutics (including tumor-targeting monoclonal antibodies) also relies on the host immune system, this classification does not properly reflect the complexity of the drug-host-tumor interaction. Alternatively, anticancer immunotherapeutics can be classified according to their antigen specificity. While some immunotherapies specifically target one (or a few) defined tumor-associated antigen(s), others operate in a relatively non-specific manner and boost natural or therapy-elicited anticancer immune responses of unknown and often broad specificity. Here, we propose a critical, integrated classification of anticancer immunotherapies and discuss the clinical relevance of these approaches.

A new classification of the Turkish terranes and sutures and its implication for the paleotectonic history of the region

The Turkish part of the Tethyan realm is represented by a series of terranes juxtaposed through Alpine convergent movements and separated by complex suture zones. Different terranes can be defined and characterized by their dominant geological background. The Pontides domain represents a segment of the former active margin of Eurasia, where back-arc basins opened in the Triassic and separated the Sakarya terrane from neighbouring regions. Sakarya was re-accreted to Laurasia through the Balkanic mid-Cretaceous orogenic event that also affected the Rhodope and Strandja zones. The whole region from the Balkans to the Caucasus was then affected by a reversal of subduction and creation of a Late Cretaceous arc before collision with the Anatolian domain in the Eocene. If the Anatolian terrane underwent an evolution similar to Sakarya during the Late Paleozoic and Early Triassic times, both terranes had a diverging history during and after the Eo-Cimmerian collision. North of Sakarya, the Küre back-arc was closed during the Jurassic, whereas north of the Anatolian domain, the back-arc type oceans did not close before the Late Cretaceous. During the Cretaceous, both domains were affected by ophiolite obduction, but in very different ways: north directed diachronous Middle to Late Cretaceous mélange obduction on the Jurassic Sakarya passive margin; Senonian synchronous southward obduction on the Triassic passive margin of Anatolia. From this, it appears that the Izmir-Ankara suture, currently separating both terranes, is composite, and that the passive margin of Sakarya is not the conjugate margin of Anatolia. To the south, the Cimmerian Taurus domain together with the Beydağları domain (part of the larger Greater Apulian terrane), were detached from north Gondwana in the Permian during the opening of the Neotethys (East-Mediterranean basin). The drifting Cimmerian blocks entered into a soft collision with the Anatolian and related terranes in the Eo-Cimmerian orogenic phase (Late Triassic), thus suturing the Paleotethys. At that time, the Taurus plate developed foreland-type basins, filled with flysch-molasse deposits that locally overstepped the lower plate Taurus terrane and were deposited in the opening Neotethys to the south. These olistostromal deposits are characterized by pelagic Carboniferous and Permian material from the Paleotethys suture zone found in the Mersin mélange. The latter, as well as the Antalya and Mamonia domains are represented by a series of exotic units now found south of the main Taurus range. Part of the Mersin exotic material was clearly derived from the former north Anatolian passive margin (Huğlu-type series) and re-displaced during the Paleogene. This led us to propose a plate tectonic model where the Anatolian ophiolitic front is linked up with the Samail/Baër-Bassit obduction front found along the Arabian margin. The obduction front was indented by the Anatolian promontory whose eastern end was partially subducted. Continued slab roll-back of the Neotethys allowed Anatolian exotics to continue their course southwestward until their emplacement along the Taurus southern margin (Mersin) and up to the Beydağları promontory (Antaya-Mamonia) in the latest Cretaceous-Paleocene. The supra-subduction ocean opening at the back of the obduction front (Troodos-type Ocean) was finally closed by Eocene north-south shortening between Africa and Eurasia. This brought close to each other Cretaceous ophiolites derived from the north of Anatolia and those obducted on the Arabian promontory. The latter were sealed by a Maastrichtian platform, and locally never affected by Alpine tectonism, whereas those located on the eastern Anatolian plate are strongly deformed and metamorphosed, and affected by Eocene arc magmatism. These observations help to reconstruct the larger frame of the central Tethyan realm geodynamic evolution.

Classification of very high spatial resolution imagery using mathematical morphology and support vector machines

We investigate the relevance of morphological operators for the classification of land use in urban scenes using submetric panchromatic imagery. A support vector machine is used for the classification. Six types of filters have been employed: opening and closing, opening and closing by reconstruction, and opening and closing top hat. The type and scale of the filters are discussed, and a feature selection algorithm called recursive feature elimination is applied to decrease the dimensionality of the input data. The analysis performed on two QuickBird panchromatic images showed that simple opening and closing operators are the most relevant for classification at such a high spatial resolution. Moreover, mixed sets combining simple and reconstruction filters provided the best performance. Tests performed on both images, having areas characterized by different architectural styles, yielded similar results for both feature selection and classification accuracy, suggesting the generalization of the feature sets highlighted.

Mixed spectral-structural classification of very high resolution images with summation kernels

In this paper, mixed spectral-structural kernel machines are proposed for the classification of very-high resolution images. The simultaneous use of multispectral and structural features (computed using morphological filters) allows a significant increase in classification accuracy of remote sensing images. Subsequently, weighted summation kernel support vector machines are proposed and applied in order to take into account the multiscale nature of the scene considered. Such classifiers use the Mercer property of kernel matrices to compute a new kernel matrix accounting simultaneously for two scale parameters. Tests on a Zurich QuickBird image show the relevance of the proposed method : using the mixed spectral-structural features, the classification accuracy increases of about 5%, achieving a Kappa index of 0.97. The multikernel approach proposed provide an overall accuracy of 98.90% with related Kappa index of 0.985.