Color image retrieval based on hidden Markov models

In this correspondence, a new approach to retrieving images from a color image database is proposed. Each image in the database is represented by a two-dimensional pseudo-hidden Markov model (2-D PHMM), which characterizes the chromatic and spatial information about the image. In addition, a flexible pictorial querying method is used, by which users can paint the rough content of the desired images in a query picture. Image matching is achieved by comparing the query picture with each 2-D PHMM in the database. Experimental results show that the proposed approach is indeed effective.     

Cathepsin E expression by normal and premalignant cervical epithelium

We have investigated the expression of the aspartic proteinase cathepsin E and HLA-DR and the presence of HPV16 in normal squamous epithelium (n = 8) and low-grade (n = 21) and high-grade (n = 14) intraepithelial squamous lesions of the uterine cervix. Immunohistochemistry of cervical biopsies revealed that up-regulation of cathepsin E expression was related to increasing severity of the cervical intraepithelial neoplasia (CIN). Up-regulation of protein was associated with increased message as assessed by in situ hybridization. Langerhans cells and the majority of koilocytes did not express detectable cathepsin E levels. Although there was also an up-regulation of HLA-DR expression by cervical keratinocytes in cervical intraepithelial neoplasia lesions, as determined by immunohistochemistry, no significant correlation was found between HLA-DR and cathepsin E expression in these lesions; neither was expression of cathepsin E correlated to the presence of HPV16, detected by polymerase chain reaction. The expression of cathepsin E, an aspartic proteinase that is reported to play a role in antigen processing for presentation by class II major histocompatibility complex molecules, is associated with cellular dedifferentiation in cervical intraepithelial neoplasia.     

Evaluating the impact of CASE: an empirical comparison of retrospective and cross-sectional survey approaches

The purpose of this paper is to evaluate two methods of assessing the productivity and quality impact of Computer Aided Software Engineering (CASE) and Fourth Generation Language (4GL) technologies: (1) by the retrospective method; and (2) the cross-sectional method. Both methods involve the use of questionnaire surveys. Developers' perceptions depend on the context in which they are expressed and this includes expectations about the effectiveness of a given software product. Consequently, it is generally not reliable to base inferences about the relative merits of CASE and 4GLs on a cross-sectional comparison of two separate samples of users. The retrospective method that requires each respondent to directly compare different products is shown to be more reliable. However, there may be scope to employ cross-sectional comparisons of the findings from different samples where both sets of respondents use the same reference point for their judgements, and where numerical rather than verbal rating scales are used to measure perceptions.     

Longitudinal weight changes, length of survival, and energy requirements of long-term care residents with dementia

OBJECTIVE: We hypothesized that institutionalized patients with dementia, who frequently have feeding problems and require supervised and assisted feeding, would lose more weight during their residency than nondemented, independently functioning residents and have compromised survival. To test this hypothesis, we examined the survival and longitudinal changes in weight of two cohorts of institutionalized residents with dementia and compared these cohorts with a cohort of nondemented residents. We also measured the resting energy expenditures of a subset of the subjects with dementia as an indicator of their energy needs. DESIGN: A longitudinal cohort study with retrospective baseline chart review and subsequent follow-up of monthly weights and mortality over 4 years. SETTING: A 725-bed long-term care institution with specified levels of care. SUBJECTS: Two cohorts of residents with dementia, one consisting of subjects who required total care throughout their institutional stay (n = 31) and another group who did not initially require total care (n = 48); these were compared with a cohort with normal mentation who were functionally independent in their daily activities (n = 26). The total number of subjects was 105. MEASUREMENTS: Demographics, medical problems, and medications by chart review; functional and mental status evaluations; longitudinal monthly weights and mortality for the 48-month study period; and resting energy expenditures by indirect calorimetry. MAIN RESULTS: Residents with dementia had lower weights on admission and throughout their stay than nondemented, independently functioning residents, and they were more likely to have a weight loss of 10 lbs or more at some point during the 4-year study period. However, their mean weights did not change during the study period. The mean survival from admission of those demented residents who died was more than 3 years. Resting energy expenditures of women residents with advanced dementia were 12% lower than predicted from the Harris Benedict equations. CONCLUSION: Dementia is not necessarily associated with unremitting weight loss during institutionalization despite the frequent occurrence of feeding difficulties and temporary weight loss. This may be caused partly by the lower than expected resting energy expenditures and, hence, energy needs of affected residents as their dementia progresses. Demented residents weighed significantly less than nondemented, independently functioning residents throughout their institutional stay. Nevertheless, nursing staff are able to maintain weight and survival for extended periods even in very impaired residents.     

Use of an intravascular T1 contrast agent to improve MR cine myocardial-blood pool definition in man

The feasibility of improving myocardial/blood pool contrast in MR cine images through use of an intravascular contrast agent (Ferumoxtran, Advanced Magnetics, Inc., Cambridge, MA) was tested in four subjects. The contrast-to-noise ratio (CNR) demonstrated a trend toward improvement in the short axis and improved significantly in the long axis cine by an average of 128% (P < .05). Image intensity gradients at the myocardial/blood pool interface increased significantly in both the short and long axis (P < .01). It is expected that larger image intensity gradients at the endocardial border should improve the capabilities of automated segmentation algorithms, reducing the uncertainty and need for manual editing.     

Electrical linewidth test structures fabricated in monocrystallinefilms for reference-material applications

The physical widths of reference features incorporated into electrical linewidth test structures patterned in films of monocrystalline silicon have been determined from Kelvin voltage measurements. The films in which the test structures are patterned are electrically insulated from the bulk-silicon substrate by a layer of silicon dioxide provided by SIMOX (Separation by the IMplantation of OXygen) processing. The motivation is to facilitate the development of linewidth reference materials for critical-dimension (CD) metrology-instrument calibration. The selection of the (110) orientation of the starting silicon and the orientation of the structures' features relative to the crystal lattice enable a lattice-plane-selective etch to generate reference-feature properties of rectangular cross section and atomically planar sidewalls. These properties are highly desirable for CD applications in which feature widths are certified with nanometer-level uncertainty for use by a diverse range of CD instruments. End applications include the development and calibration of new generations of CD instruments directed at controlling processes for manufacturing devices having sub-quarter-micrometer features     

Biodegradation of indole at high concentration by persolvent fermentation with Pseudomonas sp. ST-200

The value of flexible videoendoscopy as a biofeedback tool in conservative swallowing rehabilitation of patients suffering from aspiration after head and neck surgery was examined. For this purpose, the outcomes of conventional swallowing therapy and of swallowing therapy with support of videoendoscopic biofeedback were compared. In the first 40 days of swallowing therapy, videoendoscopic biofeedback significantly increased the chance of therapeutic success, shortening the period of functional rehabilitation in comparison to conventional swallowing therapy. Limitations of this visual biofeedback procedure are poor cognitive skills of the patient, rejection of the procedure by the patient and local factors that make the positioning of the endoscope difficult.     

Partitioned encoding schemes for algorithm-based fault tolerance inmassively parallel systems

Considers the applicability of algorithm based fault tolerance (ABET) to massively parallel scientific computation. Existing ABET schemes can provide effective fault tolerance at a low cost For computation on matrices of moderate size; however, the methods do not scale well to floating-point operations on large systems. This short note proposes the use of a partitioned linear encoding scheme to provide scalability. Matrix algorithms employing this scheme are presented and compared to current ABET schemes. It is shown that the partitioned scheme provides scalable linear codes with improved numerical properties with only a small increase in hardware and time overhead     

Weight smoothing to improve network generalization

A weight smoothing algorithm is proposed in this paper to improve a neural network's generalization capability. The algorithm can be used when the data patterns to be classified are presented on an n-dimensional grid (n>/=1) and there exists some correlations among neighboring data points within a pattern. For a fully-interconnected feedforward net, no such correlation information is embedded into the architecture. Consequently, the correlations can only be extracted through sufficient amount of network training. With the proposed algorithm, a smoothing constraint is incorporated into the objective function of backpropagation to reflect the neighborhood correlations and to seek those solutions that have smooth connection weights. Experiments were performed on problems of waveform classification, multifont alphanumeric character recognition, and handwritten numeral recognition. The results indicate that (1) networks trained with the algorithm do have smooth connection weights, and (2) they generalize better.     

Network information criterion-determining the number of hiddenunits for an artificial neural network model

The problem of model selection, or determination of the number of hidden units, can be approached statistically, by generalizing Akaike's information criterion (AIC) to be applicable to unfaithful (i.e., unrealizable) models with general loss criteria including regularization terms. The relation between the training error and the generalization error is studied in terms of the number of the training examples and the complexity of a network which reduces to the number of parameters in the ordinary statistical theory of AIC. This relation leads to a new network information criterion which is useful for selecting the optimal network model based on a given training set.