Traumatic brain injury (TBI) is a leading cause of disability worldwide. It is the most common cause of death and disability during the first three decades of life and accounts for more productive years of life lost than cancer, cardiovascular disease and HIV/AIDS combined. Cognitive Rehabilitation (CR), as part of Neurorehabilitation, aims to reduce the cognitive deficits caused by TBI. CR treatment consists of sequentially organized tasks that require repetitive use of impaired cognitive functions. While task repetition is not the only important feature, it is becoming clear that neuroplastic change and functional improvement only occur after a number of specific tasks are performed in a certain order and repetitions and does not occur otherwise. Until now, there has been an important lack of well-established criteria and on-field experience by which to identify the right number and order of tasks to propose to each individual patient. This thesis proposes the CMIS methodology to support health professionals to compose CR programs by selecting the most promising tasks in the right order. Two contributions to this topic were developed for specific steps of CMIS through innovative data mining techniques SAIMAP and NRRMR methodologies. SAIMAP (Sequence of Activities Improving Multi-Area Performance) proposes an innovative combination of data mining techniques in a hybrid generic methodological framework to find sequential patterns of a predefined set of activities and to associate them with multi-criteria improvement indicators regarding a predefined set of areas targeted by the activities. It combines data and prior knowledge with preprocessing, clustering, motif discovery and classes` post-processing to understand the effects of a sequence of activities on targeted areas, provided that these activities have high interactions and cumulative effects. Furthermore, this work introduces and defines the Neurorehabilitation Range (NRR) concept to determine the degree of performance expected for a CR task and the number of repetitions required to produce maximum rehabilitation effects on the individual. An operationalization of NRR is proposed by means of a visualization tool called SAP. SAP (Sectorized and Annotated Plane) is introduced to identify areas where there is a high probability of a target event occurring. Three approaches to SAP are defined, implemented, applied, and validated to a real case: Vis-SAP, DT-SAP and FT-SAP. Finally, the NRRMR (Neurorehabilitation Range Maximal Regions) problem is introduced as a generalization of the Maximal Empty Rectangle problem (MER) to identify maximal NRR over a FT-SAP. These contributions combined together in the CMIS methodology permit to identify a convenient pattern for a CR program (by means of a regular expression) and to instantiate by a real sequence of tasks in NRR by maximizing expected improvement of patients, thus provide support for the creation of CR plans. First of all, SAIMAP provides the general structure of successful CR sequences providing the length of the sequence and the kind of task recommended at every position (attention tasks, memory task or executive function task). Next, NRRMR provides specific tasks information to help decide which particular task is placed at each position in the sequence, the number of repetitions, and the expected range of results to maximize improvement after treatment. From the Artificial Intelligence point of view the proposed methodologies are general enough to be applied in similar problems where a sequence of interconnected activities with cumulative effects are used to impact on a set of areas of interest, for example spinal cord injury patients following physical rehabilitation program or elderly patients facing cognitive decline due to aging by cognitive stimulation programs or on educational settings to find the best way to combine mathematical drills in a program for a specific Mathematics course.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados