Profile
International Journal of Radiology & Medical Imaging Volume 1 (2015), Article ID 1:IJRMI-105, 7 pages
https://doi.org/10.15344/2456-446X/2015/105
Research Article
Development of Novel Nuclear Medicine Image Filter to Improve Image Matrix size Through Multidivisional Short-time Data Acquisition

Yoshiyuki Hosokai1,*, Shigehisa Tanaka2, Akihito Usui1, Yuuji Kaga1 and Haruo Saito1

1Department of Diagnostic Image Analysis, Tohoku University Graduate School of Medicine, 2-1 Seiryo-machi, Sendai, Miyagi, 980-8575, Japan
2Sendai Kousei Hospital, 4-15 Hirose-chou, Aoba-ku, Sendai, Miyagi 980-0873, Japan
Dr. Yoshiyuki Hosokai, Department of Diagnostic Image Analysis, Tohoku University Graduate School of Medicine, 2-1 Seiryomachi, Sendai, Miyagi, 980-8575, Japan; E-mail: hosokai@med.tohoku.ac.jp
03 August 2015; 03 October 2015; 06 October 2015
Hosokai Y, Tanaka S, Usui A, Kaga Y, Saito H (2015) Development of Novel Nuclear Medicine Image Filter to Improve Image Matrix size Through Multidivisional Short-time Data Acquisition. Int J Radiol Med Imag 1: 105. doi: https://doi.org/10.15344/2456-446X/2015/105

Abstract

Rationale: Total count data of a nuclear medicine image are calculated using a definite integral over total acquisition time (T), although γ-rays are completely random and irregular irrespective of their direction and gamma camera detector incidence location. However, random events could be used to create images by rearranging the time series and changing the data observation direction. We examined a filtering method wherein a time series of the count data from four directions is observed, four approximation equations are created, and the matrix size is enlarged by complementing it two-fold.
Methods: Two phantoms and patient data were used here. Static and dynamic acquisition (T: 60 s or 300 s, time units: 1 s) were performed. The image data were a 3-dimensional data array within the timeline for each pixel, and the data analysis was performed for all pixels by fixing and varying the timeline in the data array. The set acquisition time data was rearranged into four patterns, a linear approximation equation was determined for each pattern (the least squares method), and then replotted in a 2-fold larger matrix area.
Results: All approximation equations exhibited a correlation of R2 0.79. Four patterns: Normal, Inverse, Forward, and Backward, were constructed from the time series of acquisitions for a set time, and approximation equation data for four pixels were created from the data of one pixel. The apparent resolution increases when an image is partially enlarged, but this fell well short of the data that were actually acquired, although the apparent resolution was increased by implementing the proposed technique.
Discussion: We proposed a novel filtering method that changes the matrix size without extending the imaging time by utilizing the acquisition events’ irregularity. Although the proposed technique requires changes in data processing software, it is extremely cost-effective because equipment modifications are unnecessary.