Luca Anselma

Applying Artificial Intelligence to Clinical Guidelines: The GLARE Approach.

In this paper, we present GLARE, a domain-independent system for acquiring, representing and executing clinical guidelines. GLARE is characterized by the adoption of Artificial Intelligence (AI) techniques at different levels in the definition and implementation of the system. First of all, a high-level and user-friendly knowledge representation language has been designed, providing a set of representation primitives. Second, a user-friendly acquisition tool has been designed and implemented, on the basis of the knowledge representation formalism. The acquisition tool provides various forms of help for the expert physicians, including different levels of syntactic and semantic tests in order to check the “well-formedness” of the guidelines being acquired. Third, a tool for executing guidelines on a specific patient has been made available. The execution module provides a hypothetical reasoning facility, to support physicians in the comparison of alternative diagnostic and/or therapeutic strategies. Moreover, advanced and extended AI techniques for temporal reasoning and temporal consistency checking are used both in the acquisition and in the execution phase. The GLARE approach has been successfully tested on clinical guidelines in different domains, including bladder cancer, reflux esophagitis, and heart failure.

Supporting Knowledge-Based Decision Making in the Medical Context: The GLARE Approach

Knowledge-based clinical decision making is one of the most challenging activities of physicians. Clinical Practice Guidelines are commonly recognized as a useful tool to help physicians in such activities by encoding the indications provided by evidence-based medicine. Computer-based approaches can provide useful facilities to put guidelines into practice and to support physicians in decision-making. Specifically, GLARE (GuideLine Acquisition, Representation and Execution) is a domain-independent prototypical tool providing advanced Artificial Intelligence techniques to support medical decision making, including what-if analysis, temporal reasoning, and decision theory analysis. The paper describes such facilities considering a real-world running example and focusing on the treatment of therapeutic decisions.

Recursive representation of periodicity and temporal reasoning

Representing and reasoning with repeated and periodic events is important in many real-world domains, such as protocol and guideline management. In this set, it is important to give support to complex periodicities, that can involve non-symmetric repetitions, imprecision, variability, pauses between repetitions, and nested time intervals. Also, in these domains it can be useful to give support to composite events, as well as classes of events (i.e. types of actions) and instances of events (i.e. specific actions). In this paper, we propose a general-purpose domain-independent knowledge server dealing with all these issues. In particular, we describe a compact and (hopefully) user-friendly formalism for representing repetition/periodicity temporal constraints that supports arbitrarily nested repetitions as well as possibly imprecise and variable delays between repetitions. Moreover, we define two algorithms for performing consistency checking on knowledge bases of (possibly repeated/periodic) classes and instances of events retaining the efficiency of less expressive approaches.

Extending BCDM to Cope with Proposals and Evaluations of Updates

The cooperative construction of data/knowledge bases has recently had a significant impulse (see, e.g., Wikipedia [1]). In cases in which data/knowledge quality and reliability are crucial, proposals of update/insertion/deletion need to be evaluated by experts. To the best of our knowledge, no theoretical framework has been devised to model the semantics of update proposal/ evaluation in the relational context. Since time is an intrinsic part of most domains (as well as of the proposal/evaluation process itself), semantic approaches to temporal relational databases (specifically, Bitemporal Conceptual Data Model (henceforth, BCDM) [2]) are the starting point of our approach. In this paper, we propose BCDMPV, a semantic temporal relational model that extends BCDM to deal with multiple update/insertion/deletion proposals and with acceptances/rejections of proposals themselves. We propose a theoretical framework, defining the new data structures, manipulation operations and temporal relational algebra and proving some basic properties, namely that BCDMPV is a consistent extension of BCDM and that it is reducible to BCDM. These properties ensure consistency with most relational temporal database frameworks, facilitating implementations.

Temporal detection and analysis of guideline interactions

BACKGROUND Clinical practice guidelines (CPGs) are assuming a major role in the medical area, to grant the quality of medical assistance, supporting physicians with evidence-based information of interventions in the treatment of single pathologies. The treatment of patients affected by multiple diseases (comorbid patients) is one of the main challenges for the modern healthcare. It requires the development of new methodologies, supporting physicians in the treatment of interactions between CPGs. Several approaches have started to face such a challenging problem. However, they suffer from a substantial limitation: they do not take into account the temporal dimension. Indeed, practically speaking, interactions occur in time. For instance, the effects of two actions taken from different guidelines may potentially conflict, but practical conflicts happen only if the times of execution of such actions are such that their effects overlap in time. OBJECTIVES We aim at devising a methodology to detect and analyse interactions between CPGs that considers the temporal dimension. METHODS In this paper, we first extend our previous ontological model to deal with the fact that actions, goals, effects and interactions occur in time, and to model both qualitative and quantitative temporal constraints between them. Then, we identify different application scenarios, and, for each of them, we propose different types of facilities for user physicians, useful to support the temporal detection of interactions. RESULTS We provide a modular approach in which different Artificial Intelligence temporal reasoning techniques, based on temporal constraint propagation, are widely exploited to provide users with such facilities. We applied our methodology to two cases of comorbidities, using simplified versions of CPGs. CONCLUSION We propose an innovative approach to the detection and analysis of interactions between CPGs considering different sources of temporal information (CPGs, ontological knowledge and execution logs), which is the first one in the literature that takes into account the temporal issues, and accounts for different application scenarios.

Querying now-relative data

Now-relative temporal data play an important role in most temporal applications, and their management has been proved to impact in a crucial way the efficiency of temporal databases. Though several temporal relational approaches have been developed to deal with now-relative data, none of them has provided a whole temporal algebra to query them. In this paper we overcome such a limitation, by proposing a general algebra which is parametrically adapted to cope with the relational approaches to now-relative data in the literature, i.e., MIN, MAX, NULL and POINT approaches. Besides being general enough to provide a query language for several approaches in the literature, our algebra has been designed in such a way to satisfy several theoretical and practical desiderata: closure with respect to representation languages, correctness with respect to the “consensus” BCDM semantics, reducibility to the standard non-temporal algebra (which involves interoperability with non-temporal relational databases), implementability and efficiency. Indeed, the experimental evaluation we have drawn on our implementation has shown that only a slight overhead is added by our treatment of now-relative data (with respect to an approach in which such data are not present).

Valid-Time Indeterminacy in Temporal Relational Databases: A Family of Data Models

Valid-time indeterminacy concerns not knowing exactly when a fact holds in the modeled reality. In this paper, we first propose a reference approach (data model and algebra) in which all possible temporal scenarios induced by valid-time indeterminacy can be extensionally modeled. We then specify a family of sixteen more compact representational data models. We demonstrate their correctness with respect to the reference approach and analyze several properties, including their data expressiveness and correctness with respect to the reference approach. Finally, we compare these compact models along several relevant dimensions.

Temporal reasoning about composite and/or periodic events

In many application areas, including planning, workflow, guideline and protocol management, the description of the domain involves composite and/or periodic events, mutually related by temporal constraints on the execution order. Such events represent ‘classes’, since they can be instantiated to specific executions of the plan, guideline etc., and each execution must ‘respect’ the temporal constraints imposed on the corresponding classes. The main goal of our work is to propose an approach dealing with the above-mentioned temporal phenomena. To achieve such an objective, the authors propose a tractable domain-independent temporal reasoner. This enhances the generality of our approach, which provides a domain-independent module that can be integrated with other software tools to solve temporal problems in specific domains. From the methodological point of view, the authors first devise a representation formalism coping with the aforesaid phenomena, and then they describe temporal constraint propagation algorithms to deal with constraint inheritance and to perform temporal consistency checking. The representation formalism has been designed carefully, to obtain algorithms that are both complete and tractable. Finally, the paper also shows experimental results, including an application of the authors’ approach to clinical guidelines, evaluating their impact on future applications and research activities.

Automatically Decomposing Configuration Problems

Configuration was one of the first tasks successfully approached via AI techniques. However, solving configuration problems can be computationally expensive. In this work, we show that the decomposition of a configuration problem into a set of simpler and independent subproblems can decrease the computational cost of solving it. In particular, we describe a novel decomposition technique exploiting the compositional structure of complex objects and we show experimentally that such a decomposition can improve the efficiency of configurators.

A 1NF temporal relational model and algebra coping with valid-time temporal indeterminacy

In the real world, many phenomena are time related and in the last three decades the database community has devoted much work in dealing with “time of facts” in databases. While many approaches incorporating time in the relational model have been already devised, most of them assume that the exact time of facts is known. However, this assumption does not hold in many practical domains, in which temporal indeterminacy of facts occurs. The treatment of valid-time indeterminacy requires in-depth extensions to the current relational approaches. In this paper, we propose a theoretically grounded approach to cope with this issue, overcoming the limitations of related approaches in the literature. In particular, we present a 1NF temporal relational model and propose a new temporal relational algebra to query it. We also formally study the properties of the new data model and algebra, thus granting that our approach is interoperable with pre-existent temporal and non-temporal relational approaches, and is implementable on top of them. Finally, we consider computational complexity, showing that only a limited overhead is added when moving from determinate to indeterminate time.