Program Information
Development of New Biological Fingerprints for Patient Recognition to Identify Misfiled Images in a PACS Server
Y Shimizu1*, J Morishita2 , Y Yoon1 , K Iwase1 , S Yasumatsu1 , Y Matsunobu1 , (1) Department of Health Sciences, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan (2)Department of Health Sciences, Faculty of Medical Sciences, Kyushu University, Fukuoka, Japan
Presentations
SU-E-I-75 (Sunday, July 12, 2015) 3:00 PM - 6:00 PM Room: Exhibit Hall
Purpose: We are trying to develop an image-searching technique to identify misfiled images in a picture archiving and communication system (PACS) server by using five biological fingerprints: the whole lung field, cardiac shadow, superior mediastinum, lung apex, and right lower lung. Each biological fingerprint in a chest radiograph includes distinctive anatomical structures to identify misfiled images. The whole lung field was less effective for evaluating the similarity between two images than the other biological fingerprints. This was mainly due to the variation in the positioning for chest radiographs. The purpose of this study is to develop new biological fingerprints that could reduce influence of differences in the positioning for chest radiography.
Methods: Two hundred patients were selected randomly from our database (36,212 patients). These patients had two images each (current and previous images). Current images were used as the misfiled images in this study. A circumscribed rectangular area of the lung and the upper half of the rectangle were selected automatically as new biological fingerprints. These biological fingerprints were matched to all previous images in the database. The degrees of similarity between the two images were calculated for the same and different patients. The usefulness of new the biological fingerprints for automated patient recognition was examined in terms of receiver operating characteristic (ROC) analysis.
Results: Area under the ROC curves (AUCs) for the circumscribed rectangle of the lung, upper half of the rectangle, and whole lung field were 0.980, 0.994, and 0.950, respectively. The new biological fingerprints showed better performance in identifying the patients correctly than the whole lung field.
Conclusion: We have developed new biological fingerprints: circumscribed rectangle of the lung and upper half of the rectangle. These new biological fingerprints would be useful for automated patient identification system because they are less affected by positioning differences during imaging.
Contact Email: