Search

Background To initiate the development of a machine learning algorithm capable of comparing segments of pre and post pamidronate whole body MRI scans to assess treatment response and to compare the results of this algorithm with the analysis of a panel of paediatric radiologists. Methods Whole body MRI of patients under the age of 16 diagnosed with CNO and treated with pamidronate at a tertiary referral paediatric hospital in United Kingdom between 2005 and 2017 were reviewed. Pre and post pamidronate images of the commonest sites of involvement (distal femur and proximal tibia) were manually selected (n = 45). A machine learning algorithm was developed and tested to assess treatment effectiveness by comparing pre and post pamidronate scans. The results of this algorithm were compared with the results of a panel of radiologists (ground truth). Results When tested initially the machine algorithm predicted 4/7 (57.1%) examples correctly in the multi class model, and 5/7 (71.4%) correctly in the binary group. However when compared to the ground truth, the machine model was able to classify only 33.3% of the samples correctly but had a sensitivity of 100% in detecting improvement or worsening of disease. Conclusion The machine learning could detect new lesions or resolution of a lesion with good sensitivity but failed to classify stable disease accurately. However, further validation on larger datasets are required to improve the specificity and accuracy of the machine model.

Models are central for the development and management of complex systems. In order to be useful along the entire life cycle of software they must provide reliable support and enable automatic management. For that purpose, they must be precise, consistent, correct, and be subject to stringent quality verification and control criteria. Model automation calls for deep formal study of models, so that tools can provide inclusive support to users. This paper deals with optimization of multiplicity constraints in Class Models, i.e., models that provide abstraction on the static structure of software. The paper introduces an in-depth analysis of redundancy of multiplicity constraints, i.e., multiplicities that cannot be realized in any legal instance. The analysis includes: (1) a formal study of gaps of redundancies in multiplicity intervals; (2) algorithmic and rule-based methods for removing redundancies in multiplicity constraints; (3) a formal study of completeness of the algorithmic procedures, with respect to most UML class model constraints. The algorithmic procedures are implemented in our FiniteSatUSE tool. To the best of our knowledge, there is no previous study of properties of multiplicity redundancy, no completeness analysis of tightening methods, and no systematic study of these features with respect to most UML class model constraints.

In this paper we consider the exogenous indifference classes model of Barbera and Ehlers (2011) and Sato (2009) and analyze further the relationship between the structure of indifference classes across agents and dictatorship results. The key to our approach is the pairwise partition graph. We provide necessary conditions on these graphs for strategy-proofness and unanimity (or efficiency) to imply dictatorship. These conditions are not sufficient; we also provide separate stronger conditions that are sufficient. A full characterization is obtained in the case of two agents for domains where strategy-proofness and efficiency imply dictatorship.

We propose a method for localization and classification of brand logos in natural images. The system has to overcome multiple challenges such as perspective deformations, warping, variations of the shape and colors, occlusions, background variations. To deal with perspective variation, we rely on homography matching between the SIFT keypoints of logo instances of the same class. To address the changes in color, we construct a weighted graph of logo interconnections that is further analyzed to extract potentially multiple instances of the class. The main instance is built by grouping the keypoints of the graph connected logos onto the central image. The secondary instance is needed for color inverted logos and is obtained by inverting the orientation of the main instance. The constructed logo recognition system is tested on two databases (FlickrLogos-32 and BelgaLogos), outperforming state of the art with more than 10 % accuracy.

The number of common neighbor between nodes is applied to the modeling of resting-state brain function network in order to analyze the effect of anatomical distance on the modeling of resting-state brain function network. Three models based on anatomical distance, the number of common neighbor, or anatomical distance and the number of common neighbor are designed. Basing on residuals creates the evaluation criteria for selecting the optimal brain function model network in each class model. The model is selected to simulate the human real brain function network by comparison with real data functional magnetic resonance imaging (fMRI) network. Finally, the result shows that the best model only is based on anatomical distance.

In the present study, near-infrared spectroscopy (NIRS) was explored as a fast and reliable screening method for the detection of adulteration of skim milk powder (SMP). Sixty genuine SMP were adulterated with acid whey (1–25 % w/w), starch (2 and 5 %) and maltodextrin (2 and 5 %) for a total of 348 adulterated samples. Two chemometric approaches were employed. In the first approach, an untargeted one class model for genuine skim milk powder was developed by Soft Independent Modelling of Class Analogy. In the second approach, adulterant-specific regression models were developed to assess the amount of each adulterant by partial least square regression and principal component regression. The class modelling approach had the advantage that several adulterants could be detected with the same chemometric model, including situations where multiple adulterants are present in the test sample or where yet unknown adulterants are present. Regression models showed a better sensitivity with genuine SMP samples completely discriminated from samples adulterated with 5 % acid whey and 2 % of starch or maltodextrin. NIRS proved to be a useful tool for the rapid and cost-efficient untargeted and/or targeted detection of adulterations in SMP.

Classical location models fix an objective function and then attempt to find optimal points to this objective. In the last years a flexible approach, the ordered median problem, has been introduced. It handles a wide class of objectives, such as the median, the center and the centdian function. In this paper we present new properties of the ordered median problem such as solvability for the situation of attractive and repulsive locations. We also develop a new solution method that even yields local optimal points for non-convex objective functions. Furthermore, we discuss separability of ordered median problems without repulsion and derive a sufficient criterion. Finally, we introduce a useful model extension, the facility class model, which allows to deal with a wider range of real world problems in the ordered median setting.

We report preliminary results on the investigation of the hypothesis of the existence of latent classes in the total demand for days out in a vast section of the Italian Eastern Alps. Finite mixing is informed by socioeconomic variables and it is limited to the demand portion of a hurdle model of visitation. Gradually more flexible count models are estimated moving from the Poisson to the Negative Binomial -1 and -2, to the generalized negative binomial. The implications for expected consumers surplus, predicted visitation and choke prices are discussed for a plausible 2 class model where years of experience play an important role in class membership.

Some extensions of the existing determinantal methods for solving the accounting equations in latent class analysis are presented. These extensions cover more cases than previous methods, give rise to new sufficient conditions for identifiability of the latent class model, and give insight into the necessity of various sufficient conditions for identifiability. These implications to the identifiability problem are discussed.