A disease category feature database construction method of brain image based on deep convolutional neural network.

BACKGROUND: Constructing a medical image feature database according to the category of disease can achieve a quick retrieval of images with similar pathological features. Therefore, this approach has important application values in the fields such as auxiliary diagnosis, teaching, research, and telemedicine. METHODS: Based on the deep convolutional neural network, an image classifier applicable to brain disease was designed to distinguish between the image features of the different brain diseases with similar anatomical structures. Through the extraction and analysis of visual features, the images were labelled with the corresponding semantic features of a specific disease category, which can establish an association between the visual features of brain images and the semantic features of the category of disease which will permit to construct a disease category feature database of brain images. RESULTS: Based on the similarity measurement and the matching strategy of high-dimensional visual feature, a high-precision retrieval of brain image with semantics category was achieved, and the constructed disease category feature database of brain image was tested and evaluated through large numbers of pathological image retrieval experiments, the accuracy and the effectiveness of the proposed approach was verified. CONCLUSION: The disease category feature database of brain image constructed by the proposed approach achieved a quick and effective retrieval of images with similar pathological features, which is beneficial to the categorization and analysis of intractable brain diseases. This provides an effective application tool such as case-based image data management, evidence-based medicine and clinical decision support.

Application of Ontology Technology in Health Statistic Data Analysis.

Research Purpose: establish health management ontology for analysis of health statistic data. Proposed Methods: this paper established health management ontology based on the analysis of the concepts in China Health Statistics Yearbook, and used protégé to define the syntactic and semantic structure of health statistical data. RESULTS: six classes of top-level ontology concepts and their subclasses had been extracted and the object properties and data properties were defined to establish the construction of these classes. By ontology instantiation, we can integrate multi-source heterogeneous data and enable administrators to have an overall understanding and analysis of the health statistic data. CONCLUSION: ontology technology provides a comprehensive and unified information integration structure of the health management domain and lays a foundation for the efficient analysis of multi-source and heterogeneous health system management data and enhancement of the management efficiency.

Prospective study of colonization and infection because of Pseudomonas aeruginosa in mechanically ventilated patients at a neonatal intensive care unit in China.

BACKGROUND: Ventilator-associated pneumonia (VAP) is an important nosocomial infection at neonatal intensive care units (NICU), frequently caused by Pseudomonas aeruginosa. A 6-month prospective study from January 2009 through June 2009 was performed to investigate the respective contribution of endogenous and exogenous transmission of P aeruginosa in the respiratory colonization or/and infection in the mechanically ventilated patients at a NICU to identify routes of lung infection with P aeruginosa and to assess risk factors for colonization or respiratory infection with P aeruginosa. METHODS: Samples from oropharyngeal swab, tracheobronchial aspirates, gastric aspirate, and rectal swab were obtained in each patient after intubation and then twice a week. Surveillance cultures for the presence of P aeruginosa from environmental surfaces of the ICU were taken once every 5 days during the study period. Pulsed-field gel electrophoresis was used to characterize the clonal relatedness of the strains by SpeI-digested genomic DNA. RESULTS: Eighteen patients (78.3%) had colonization of the upper respiratory tract. Sixteen (69.6%) patients with colonization of the respiratory tract were infected from other patients or environmental surfaces, which was considered exogenous, and, among strains causing pulmonary infection, there were 4 (50%) patients with exogenous infection. Eight of these developed VAP after a mean of 9 ± 3.4 days. The incidence of P aeruginosa VAP on the unit was 6.2%. The respiratory tract was the earliest site of colonization in all patients of VAP. Low birth weight, duration of mechanical ventilation, previous ampicillin group use, and previous second-generation cephalosporins use were independently associated with patient-related acquisition of P aeruginosa. CONCLUSION: Our results confirm that the upper respiratory tract acts as an important reservoir of P aeruginosa colonization and infection in the mechanically ventilated patients and emphasize the importance of exogenous acquisition of P aeruginosa. A combination of early identification and eradication of airways colonization by P aeruginosa plus infection control measures may be the basis to prevent pulmonary infection.