International journal of bio-medical computing最新文献

The Computer Controlled Video Perimetry (CCVP) is a computer screening test for detecting visual function loss caused by onchocerciasis, glaucoma, etc. Installed on portable computers, the CCVP has been shown to be high acceptability in field community investigation. However, it is regarded to be difficult in obtaining reliable results from portable computer screening tests because of human behavioural variants and the lack of standard testing environment. In this paper, we propose an architecture for implementing a more reliable CCVP system. In particular, a self-organising neural network is applied to manage measurement noise caused by behavioural factors. A control unit is introduced to manage the overall behaviour of the system. The integrated test system has been used to screen optic nerve disease in onchocercal communities of rural Nigeria and the experimental results obtained from a large number of test records are very encouraging: reliable results from volatile test environments may be obtained using the proposed method.

Diagnostic hypercalcaemia discriminant functions, discriminating between clinically significant and non-significant hypercalcaemia, were tested 5 years after their development in order to evaluate the impact of time on their diagnostic capacity. Two populations, consisting of 257 and 129 patients with hypercalcaemia, were consecutively recorded, during six and three months respectively, 5 years apart under similar circumstances. The prevalence of hypercalcaemia was comparable in both populations, being 2.57 and 2.38% respectively (non-significant) (NS). The female/male ratio was 1.9 and 1.7 (NS). The discriminant functions correctly classified 81 and 80% of the women, respectively (NS) and respectively 75% and 64% of the men (NS) in the first and second recorded populations.

Four different neural network algorithms, binary adaptive resonance theory (ART1), self-organizing map, learning vector quantization and back-propagation, were compared in the diagnosis of acute appendicitis with different parameter groups. The results show that supervised learning algorithms learning vector quantization and back-propagation were better than unsupervised algorithms in this medical decision making problem. The best results were obtained with the learning vector quantization. The self-organizing map algorithm showed good specificity, but this was in conjunction with lower sensitivity. The best parameter group was found to be the clinical signs. It seems beneficial to design a decision support system which uses these methods in the decision making process.

Searching for a method to objectively detect the cognitive activity of the brain, the variability of visual evoked responses (ER) was analysed in 75 human subjects and 10 animals. The individual ERs of a normal subject were found typically very scattered in the first approx. 120 ms after stimulation, converging at 160–220 ms and then diverging again progressively. This variability pattern (VP) is event-related and is not attributable to background noise. On the other hand, statistically significant correlation showed that in most patients, with anatomically intact visual structures but with pronounced mental troubles, the VP is absent and the ERs are randomly scattered. Based on these results we consider that the event-related variability reflects the cognition function of the subjects and that it is instrumental in evidencing the distinction between cognitive and perceptive processes. The results are further consistent with the idea that cognition implies the chaotic activity of certain neural populations and that the VP reflects this chaotic, non-repetitive, non-linear and impredictable but effective neural activity.

This paper discusses a new framework and definition for the development of nursing information systems. Graves and Corcoran presented a framework for nursing informatics and defined the science. In this article their framework is expanded to include nursing practice as domain of study and to better integrate the fact that nursing informatics finds its foundation in both the nursing and the informatics disciplines. The expansion is based on both the review of definitions of medical and nursing informatics and of models and descriptions from the informatics discipline. For research purposes, the domain of nursing informatics is defined, and methods for conducting investigations in the area of analysis, modeling and development of nursing information systems are described.

Traditional approaches to the generation of finite element meshes are well suited for modeling the homogeneous or mildly heterogeneous domains presented by man-made objects, but are difficult to apply to the complex 3-D domains encountered in some biomedical applications. In this paper, we describe an adaptive algorithm that automates the modeling of these domains. The method differs from traditional approaches in that no explicit description is required of the boundaries between objects with dissimilar material properties. The algorithm uses images of the tissue class to build irregular meshes, and continuity is enforced by constraining the solution at irregular nodes. Local estimates of the error in the flux solution are used to refine the mesh. For an analytic problem with a rapid change along a spherical boundary, the adaptive method converges to a 1% voltage error using 25% of the degrees of freedom required by a uniform refinement, and to a 5% voltage gradient error using 11% of the degrees of freedom. For a defibrillation model in a pig thorax, the voltage gradient solution in the ventricles of the heart converges to within 5% of a uniform mesh solution using less than 8% of the memory and processing resources required by a uniform mesh, which has been the only practical alternative for subject-specific modeling.