Evaluation of Machine Translation Accuracy Focused on the Adverse Event Terminology for Medical Devices.

The purpose of this study was to evaluate the accuracy of deep neural machine translation focused on medical device adverse event terminology. 10 models were obtained, and their English-to-Japanese translation accuracy was evaluated using quantitative and qualitative measures. No significant difference was found in the quantitative index except for a few pairs. In the qualitative evaluation, there was a significant difference and googletrans and GPT-3 were regarded as useful models.

[Inter-rater Reliability and Cost in Pricing for Creating Dataset Focused on Mediolateral Oblique View in Mammography].

PURPOSE: The purpose of this study was to assess inter-rater reliability and workload for creating accurate training data in the clinical evaluation of mammographic positioning for deep learning. METHODS: A total of 107 mammographic images without lesions were labeled by two certified radiologic technologists in seven items: six clinical image evaluation criteria in positioning and breast tissue density. The kappa coefficient was calculated as an indicator of interrater reliability. Furthermore, the labeling cost per image was calculated based on labeling time and salary for the technologists. RESULTS: The kappa coefficients were 0.71 for inframammary fold, 0.43 for nipple in profile, 0.45 for great pectoral muscle, 0.10 for symmetrical images, and 0.61 for retromammary fat. No significant difference was found in the coefficients of spread of breast tissue. The cost per image was calculated at 11.0 yen. CONCLUSION: The inter-rater reliability for the inframammary fold, nipple in profile, great pectoral muscle, and retromammary fat ranged from "moderate" to "substantial." The reliability for symmetrical images was "slight," indicating the need for a consensus among evaluators during labeling. The labeling cost was equivalent to or higher than that of existing services.

t-SNE Visualization of Vector Pairs of Similar and Dissimilar Definition Sentences Created by Word2vec and Doc2vec in Japanese Medical Device Adverse Event Terminology.

The purpose of our study is to identify the patterns of vectors in similar/dissimilar pairs of definition sentence created by Word2vec and doec2vec for elaboration of the terminology for Japanese Medical Device Adverse Events. 2-dimension vector space created by t-SNE showed that the pair with true positive located closer in a vector space, especially Doc2vec had a strong tendency. Comparing with Word2vec, Similar vectors in Doc2vec were close and tended to form clusters.

[Variation of the Implementation of PACS Functions and Image Data Management: Questionnaire Survey in Hokkaido Prefecture].

The image management function of picture archiving and communication system (PACS) plays a very important role while electronically preserving medical records. However, details of these mechanisms vary from manufacturer to manufacturer, which could be a problem if system specifications or image management functions are not confirmed by a user during system installation or operation. The purpose of this research was to record the differences in PACS settings and image data management in hospitals. A questionnaire survey was conducted in 261 hospitals having the PACS system in Hokkaido prefecture, Japan. Questionnaire contents consisted of a face sheet and the following two items: 1) PACS image deletion method, 2) rules regarding the PACS receiving an image of data that was already registered. Results of the questionnaire survey showed that the image deletion method in the PACS server, and the overwriting rule in case of existent images being re-transmitted to the server were different between different hospitals. Furthermore, these variations were not dependent on the vendors or manufacturers, but on the settings of each hospital. Since the image management method of PACS is not standardized, we were considered to be needed more attention and appropriate regulation for safe management of electronic medical records as per the national guideline.

Evaluation of Similar Term Definitions in Medical Device Adverse Event Terminology.

The purpose of this study is to extract similar term definitions used in the terminology of Japanese medical device adverse events. We employed Levenshtein and Jaro-Winkler distances as edit distances and Skip-gram, continuous-bag of words, and fast text to produce distributed representations in Word2Vec. A comparison of the accuracies of the models showed that Levenshtein distance had higher specificity whereas Skip-gram had higher sensitivity as compared to the other models.

Natural Language Processing Based Approach for Identification of Problems in Medical Image Management Using PACS.

We investigated problems concerning medical imaging management using PACS in medical settings through text analysis. We conducted a questionnaire survey in Hokkaido, Japan, where PACS related problems were described by radiological technologists. After analyzing the descriptions in response to the questionnaire, we detected four main problems in PACS using a word co-occurrence network; i) image management, ii) image fetching error, iii) DICOM violation and iv) name notation.

Should the Picture Archiving and Communication System (PACS) Settings Be Standardized? Questionnaire Survey for Safe Medical Image Management.

The purpose of this research was to record the differences in picture archiving and communication system (PACS) settings in hospitals. A questionnaire survey was conducted among 261 hospitals that used the PACS system in Japan. As a result, the image deletion method in the PACS server, and other rules were different among hospitals. Furthermore, these variations were dependent on settings of each medical institution. It seemed necessary to define standardized rules for safe image management.

An ontology design for visualizing implicit knowledge of the radiotherapy risk.

Knowledge sharing of radiotherapy risk contributes to keep the safety practice between the oncologists, medical physicists, and therapists. But the mechanisms of risks tend to complex because of its diversities. Recently, an ontological analysis is tried to share knowledge in the various domains. Therefore, the aim of this study is to construct Radiotherapy Risk Ontology (RRO) and clarify the implicit structures of risks towards knowledge sharing. The method of this study is below. First, the classes and relations were extracted from the risk categorization of the publications. Second, these classes and relationships were connecting and mapping by the ontology editor. Third, the total relationships of classes were verified using paired comparison chart. This paper distinguishes between publication-oriented relationships and the other relationships. And these relationships was defined the explicit and implicit relationships. RRO was constructed 789 classes and 14 types of the explicit relationships. The 22 types of the implicit relationships were clarified using paired comparison. RRO provided the conceptual mapping and the implicit knowledge. The result of this study assisted the knowledge sharing.

Development of terminology for mammographic techniques for radiological technologists.

We are developing a mammographic ontology to share knowledge of the mammographic domain for radiologic technologists, with the aim of improving mammographic techniques. As a first step in constructing the ontology, we used mammography reference books to establish mammographic terminology for identifying currently available knowledge. This study proceeded in three steps: (1) determination of the domain and scope of the terminology, (2) lexical extraction, and (3) construction of hierarchical structures. We extracted terms mainly from three reference books and constructed the hierarchical structures manually. We compared features of the terms extracted from the three reference books. We constructed a terminology consisting of 440 subclasses grouped into 19 top-level classes: anatomic entity, image quality factor, findings, material, risk, breast, histological classification of breast tumors, role, foreign body, mammographic technique, physics, purpose of mammography examination, explanation of mammography examination, image development, abbreviation, quality control, equipment, interpretation, and evaluation of clinical imaging. The number of terms that occurred in the subclasses varied depending on which reference book was used. We developed a terminology of mammographic techniques for radiologic technologists consisting of 440 terms.

Quantitative evaluation of expression difference in report assignments between nursing and radiologic technology departments.

Our purpose in this study was to investigate the expression differences in report assignments between students in nursing and radiologic technology departments. We have known that faculties could identify differences, such as word usage, through grading their students' assignments. However, there are no reports in the literature dealing with expression differences in vocabulary usage in medical informatics education based on statistical techniques or other quantitative measures. The report assignment asked for students' opinions in the event that they found a rare case of a disease in a hospital after they graduated from professional school. We processed student report data automatically, and we applied the space vector model and TF/IDF (term frequency/inverse document frequency) scoring to 129 report assignments. The similarity-score distributions among the assignments for these two departments were close to normal. We focused on the sets of terms that occurred exclusively in either department. For terms such as "radiation therapy" or "communication skills" that occurred in the radiologic technology department, the TF/IDF score was 8.01. The same score was obtained for terms such as "privacy guidelines" or "consent of patients" that occurred in the nursing department. These results will help faculties to provide a better education based on identified expression differences from students' background knowledge.

Quantitative analysis of ontology research articles in the radiologic domain.

To investigate the most advanced ontology research in health care and its impact on the radiologic domain, we proposed a concept identification and abstraction technique called "Concept Step". This technique identifies a MeSH term, medical subject headings used in PubMed, in a sentence and climbs up through its hierarchy to reach an abstract concept. We developed original Java software to implement this technique. We tested it on 2,774 abstracts in health-care ontology research retrieved from MEDLINE on 23 October 2008. The total number of MeSH terms was 112,690. We counted a total of 33 MeSH terms (0.029%) in the radiologic domain. The most frequently occurring term was "radiology", which occurred 21 times in the article set. Other frequent terms were "magnetic resonance imaging" and "tomography", the counts of which were 4 and 3, respectively. A pair plot showed no correlation among the MeSH categories "Analytical Diagnostic and Therapeutic Techniques and Equipment", "Anatomy", "Biological Sciences", and "Chemicals and Drugs". We conclude that ontology research is well established in the biomedical domain, and that further study is required in the radiologic domain.

[Developing and evaluating an auto-retrieval algorithm for abbreviations in academic articles].

The purpose of this study was to develop an automated acquisition support for the semantics of abbreviations and evaluate its performance with respect to academic articles. Our goal was to support the maintenance of an institution-specific semantics inventory for abbreviations on a continuous basis. We retrieved articles from MEDLINE with the keyword "Liver [MeSH]," and 100 abstracts were randomly selected. Abbreviations and their full forms were retrieved using original Java software based on the following rules. (1) Searching the parentheses in the abstracts, the words inside the parentheses were retrieved as "INNER" and the words in front of the parentheses were retrieved as "OUTER." (2) Matching rules, such as whether the first characters of INNER and OUTER were the same. (3) If the words satisfied the conditions stated at (2), INNER was saved as the abbreviation and OUTER as the full form. Performance was manually evaluated by two graduate students and a radiologist. Of the 165 pairs of abbreviations and full forms that were obtained, 145 (87.9%) constituted correct matches.

Parsing error correction of medical phrases for semantic annotation of clinical radiology reports.

The purpose of this study is to develop a module for correcting errors in the product of a natural language parser. When tested with 300 CT reports, a total of 604 patterns were generated. The recall and precision was improved to 90.7% and 74.1% after processed by the module from initial 80.5% and 42.8% respectively. This rule-based module will help health care personnel reduce the cost of manual tagging correction for corpus building.

Realization of reliable cerebral-blood-flow maps from low-dose CT perfusion images by statistical noise reduction using nonlinear diffusion filtering.

X-ray computed tomographic perfusion (CTP) imaging, a rapid method for measuring cerebral blood flow (CBF), is an effective modality for assessment of the severity and extent of brain tissue ischemia. Low-dose scanning has been required for CTP imaging for reducing the radiation exposure to patients, because the same plane is scanned repeatedly. Low-dose CTP imaging, however, results in substantial statistical noise in the images, which may negatively impact the accuracy of CBF values. Because CBF values are calculated from the set of CTP images, it is important to reduce the statistical noise in raw CTP images to make the values reliable. Noise reduction must be performed without blurring of vessel structures, because such blurring will overestimate CBF values. For this purpose, two-dimensional nonlinear diffusion filtering (NLDF) was introduced. It was applied to CTP images of a CTP phantom for evaluating the accuracy of CBF values in low-dose CTP and to clinical low-dose CTP images for determining its effectiveness in actual CTP examinations. NLDF successfully reduced the statistical noise in the CTP images while preserving the sharp edges. This feature generated CBF values close to the reference value, producing reliable CBF maps from low-dose CT perfusion images. The CBF maps obtained with NLDF were comparable to or better than those obtained by other, commercial CTP software programs. The use of NLDF was thus effective for manipulation of low-dose CT perfusion images.

New technique for visualizing cerebral vessels in MR angiographic images using three-dimensional discrete wavelet transform.

Visualization techniques for magnetic resonance (MR) angiographic images are generally based on the projection ray concept; for example, maximum intensity projection (MIP) and volume rendering (VR). A new technique based on a different concept from projection rays is explored in this study: three-dimensional (3-D) discrete wavelet transforms are used for visualizing cerebral vessels in the MR angiographic image. This technique successfully visualizes cerebral vessels and represents the spatial relationship between the cerebral vessels, as in the case of VR. The proposed technique is, thus, indicated to be promising for visualizing cerebral vessels in 3-D MR angiographic images.

Semantic distribution study of noun*noun compounds in the Japanese CT clinical reports.

The purpose of this study is to investigate the distribution of the semantic relationships between the two atomic medical terms of the noun-noun compounds extracted from the clinical documents in Japanese. Only 29.9% of the compounds had the UMLS semantic relation defined (mainly including "location_of" and "adjacent_to"). The results indicated that the semantic relations defined in the UMLS semantic network are not enough for describing the noun-noun medical compounds extracted from the Japanese clinical reports.

Model analysis for optimal allocation of pediatric emergency center.

Optimal allocation of Pediatric Emergency Centers was simulated using both Min-Sum model and Min-Max model, based on the travel distance and child population of Hokkaido prefecture, Japan. Our result should be helpful to select the optimal number and allocation of PECs in the actual scene of decision making, where the financial restriction exists.