A Fault Detection System for a Geothermal Heat Exchanger Sensor Based on Intelligent Techniques.

This paper proposes a methodology for dealing with an issue of crucial practical importance in real engineering systems such as fault detection and recovery of a sensor. The main goal is to define a strategy to identify a malfunctioning sensor and to establish the correct measurement value in those cases. As study case, we use the data collected from a geothermal heat exchanger installed as part of the heat pump installation in a bioclimatic house. The sensor behaviour is modeled by using six different machine learning techniques: Random decision forests, gradient boosting, extremely randomized trees, adaptive boosting, k-nearest neighbors, and shallow neural networks. The achieved results suggest that this methodology is a very satisfactory solution for this kind of systems.

Machine learning for medical images analysis.

This article discusses the application of machine learning for the analysis of medical images. Specifically: (i) We show how a special type of learning models can be thought of as automatically optimized, hierarchically-structured, rule-based algorithms, and (ii) We discuss how the issue of collecting large labelled datasets applies to both conventional algorithms as well as machine learning techniques. The size of the training database is a function of model complexity rather than a characteristic of machine learning methods.

Quantitative analysis of four types of primary glomeropathy by application of a decision forest to ultrasonic and laboratory characteristics.

The aim of this study was to apply a decision forest to analysis of the ultrasound characteristics and laboratory test indices of four types of primary glomerulopathy, and quantitative analysis of the four pathologic types using a combination of these two methods. The decision trees were derived from 41 clinical indices and 5 characteristic sonographic indices obtained for the left kidney. Fifty-six patients who had undergone ultrasound-guided renal biopsy were reviewed retrospectively, and on pathologic examination, the patients were diagnosed with primary glomerulopathy, which includes mesangial proliferative glomerulonephritis, membranous nephropathy, immunoglobulin A nephropathy and minimal change disease. In this study, eight characteristic indicators were correlated with pathologic type in the 56 cases of primary glomerulopathy. The order calculated by decision forests, from high to low, is proteinuria, length of kidney, serum creatinine, plasma albumin, area of kidney, total protein, thickness of renal parenchyma, 24-h urine protein. The glomerulopathy with the highest ++++ proteinuria is membranous nephropathy, which accounts for 39.2% (22/56) of the total sample; this was followed by minimal change disease, mesangial proliferative glomerulonephritis and immunoglobulin A nephropathy. On the basis of our analysis of 41 clinical indices, the key indices for quantitative analysis of primary glomerulonephritis are laboratory tests, and these include urine protein, serum creatinine, plasma albumin, total serum protein and 24-h urine protein. The three key sonographic features are measurement indices: renal length, renal area and renal parenchymal thickness. From the eight characteristic indicators, we observed that with respect to severity (from most severe to least severe), the four types of glomerulopathy are membranous nephropathy, minimal change disease, mesangial proliferative glomerulonephritis and immunoglobulin A nephropathy.

Neighbourhood approximation using randomized forests.

Leveraging available annotated data is an essential component of many modern methods for medical image analysis. In particular, approaches making use of the "neighbourhood" structure between images for this purpose have shown significant potential. Such techniques achieve high accuracy in analysing an image by propagating information from its immediate "neighbours" within an annotated database. Despite their success in certain applications, wide use of these methods is limited due to the challenging task of determining the neighbours for an out-of-sample image. This task is either computationally expensive due to large database sizes and costly distance evaluations, or infeasible due to distance definitions over semantic information, such as ground truth annotations, which is not available for out-of-sample images. This article introduces Neighbourhood Approximation Forests (NAFs), a supervised learning algorithm providing a general and efficient approach for the task of approximate nearest neighbour retrieval for arbitrary distances. Starting from an image training database and a user-defined distance between images, the algorithm learns to use appearance-based features to cluster images approximating the neighbourhood structured induced by the distance. NAF is able to efficiently infer nearest neighbours of an out-of-sample image, even when the original distance is based on semantic information. We perform experimental evaluation in two different scenarios: (i) age prediction from brain MRI and (ii) patch-based segmentation of unregistered, arbitrary field of view CT images. The results demonstrate the performance, computational benefits, and potential of NAF for different image analysis applications.