Catálogo de publicaciones - libros

Some foot plant diseases such as flat foot and cave foot are usually diagnosed by a human expert. In this paper we propose an original method to diagnose these diseases by using optical color foot plant images. A number of modern image processing and pattern recognition techniques have been employed to configure a system that can dramatically decrease the time in which such analysis are performed, besides delivering robust and reliable results to complement efficiently the specialist’s task. Our results demonstrate the feasibility of building such automatic diagnosis systems that can be used as massive first screening methods for detecting foot plant pathologies.

Based on Markov Random Fields (MRF) theory, Bayesian methods have been accepted as an effective solution to overcome the ill-posed problems of image restoration and reconstruction. Traditionally, the knowledge in most of prior models is from a simply weighted differences between the pixel intensities within a small local neighborhood, so it can only provide limited prior information for regularization. Exploring the ways of incorporating more large-scale knowledge into prior model, this paper proposes an effective approach to incorporate large-scale image knowledge into MRF prior model. And a novel nonlocal prior is put forward. Relevant experiments in emission tomography prove that the proposed MRF nonlocal prior is capable of imposing more effective regularization on original reconstructions.

We present a novel method for predictive modeling of human brain states from functional neuroimaging (fMRI) data. Extending the traditional canonical correlation analysis of discrete data to the domain of stochastic functional measurements, the method explores the functional canonical correlation between stimuli and fMRI training data. Via an incrementally steered pattern searching technique, subspaces of voxel time courses are explored to arrive at (spatially distributed) voxel clusters that optimize the relationship between stimuli and fMRI in terms of redundancy. Application of the method for prediction of naturalistic stimuli from unknown fMRI data shows that the method finds highly predictive brain areas, i.e. brain areas relevant in processing the stimuli.

Living guidelines are documents presenting up-to-date and state-of-the-art knowledge to practitioners. To have guidelines implemented by computer-support they firstly have to be formalized in a computer-interpretable form. Due to the complexity of such formats the formalization process is challenging, but burdensome and time-consuming.

The LASSIE methodology supports this task by formalizing guidelines in several steps from the textual form to the guideline representation language Asbru using a document-centric approach. LASSIE uses Information Extraction technique to semi-automatically accomplish these steps.

We apply LASSIE to support the implementation of living guidelines. Based on a living guideline published by the Scottish Intercollegiate Guidelines Network (SIGN) we show that adaptations of previously formalized guidelines can be accomplished easily and fast. By using this new approach only new and changed text parts have to be modeled. Furthermore, models can be inherited from previously modeled guideline versions that were added by domain experts.

Medical critiquing systems criticise clinical actions performed by a physician. In order to provide useful feedback, an important task is to find differences between the actual actions and a set of ‘ideal’ actions as described by a clinical guideline. In case differences exist, insight to which extent they are compatible is provided by the critiquing system. We propose a methodology for such critiquing, where the ideal actions are given by a formal model of a clinical guideline, and where the actual actions are derived from real world patient data. We employ model checking to investigate whether a part of the actual treatment is consistent with the guideline. Furthermore, it is shown how critiquing can be cast in terms of temporal logic, and what can be achieved by using model checking. The methodology has been applied to a clinical guideline of breast cancer in conjunction with breast cancer patient data.

Research in the computerization of Clinical Guidelines (CG) has often opposed document-based approaches to knowledge-based ones. In this paper, we suggest that both approaches can be used simultaneously to assess the contents of textual Clinical Guidelines. In this first experiment, we investigate the mapping between a document model, which has been marked-up to structure its recommendations, and a knowledge structure representing the management of specific disease. This knowledge representation is based on planning formalisms, more specifically Hierarchical Task Networks (HTN). Our system operates by first automatically encoding the textual guideline through the identification of specific expressions with surface natural language processing, as described in previous work. In a subsequent step, the HTN, constructed manually and independently, and represented as an explicit AND/OR graph, is searched for a solution sub-graph using an algorithm derived from AO*. Whilst the HTN is being traversed, corresponding information is accessed in the encoded textual CG, to guide the solution extraction process. We illustrate this through a case study developed around French guidelines for the management of hypertension. Recommendations included in the textual guideline provide complementary information for the instantiation of an HTN on specific patient data. The mapping takes place at different levels, from the pre-condition of operators to the rules playing a role as selection heuristics when extracting a solution sub-graph. Such a process, which explores the textual document from the prospective of a task model, can help analyzing the overall structure of clinical guidelines and ultimately improving its applicability.

A computer-interpretable guideline knowledge base can be a very large network whose information content is difficult for developers and clinicians to comprehend and review. We created a method to annotate a guideline model and use the annotations to export the guideline knowledge base in an XML format that can be transformed into a readable document. We applied this method to knowledge bases developed in three different guideline modeling projects to analyze uses and limitations of this approach. We demonstrate the promise of creating such document-oriented views, but conclude that guideline models and knowledge bases should be constructed with the goal of creating such human-comprehensible views from the beginning.

Translating clinical guidelines into formal models is beneficial in many ways, but expensive. The progress in medical knowledge requires clinical guidelines to be updated at relatively short intervals, leading to the term . This causes potentially expensive, frequent updates of the corresponding formal models.

When performing these updates, there are two goals: The modelling effort must be minimised and the links between the original document and the formal model must be maintained. In this paper, we describe our solution, using tools and techniques developed during the Protocure II project.

The practice of medicine is becoming increasingly evidence-based and clinical practice guidelines (CPGs) are necessary for advancing evidence-based medicine (EBM). We hypothesize that machine learning methods can play an important role in learning CPGs automatically from data . Automatically induced CPGs can then be used for further manual refinement and deployment, for automated guideline compliance checking, for better understanding of disease processes, and for improved physician education. We discuss why learning CPGs is a special form of computational causal discovery and why simply predictive (i.e., non-causal) methods may not be appropriate for this task.

Lately, there have been considerable efforts to computerize Clinical Practice Guidelines (CPG) so that they can be executed via Clinical Decision Support Systems (CDSS) at the point of care. We present a Semantic Web framework to both model and execute the knowledge within a CPG to develop knowledge-centric CDSS. Our approach entails knowledge modeling through a synergy between multiple ontologies–i.e. a domain ontology, CPG ontology and patient ontology. We develop decision-rules based on the ontologies, and execute them with a proof engine to derive CPG-based patient specific recommendations. We present a prototype of our CPG-based CDSS to execute the CPG for Follow-up after Treatment for Breast Cancer.