[OPTICAL REVIEW Vol. 7, No. 1 (2000) 44-53]
Long-Term Interval Change Detection from a Sequence of Personal Images
Yucel UGURLU,1 Takashi OBI,1 Masahiro YAMAGUCHI,1 Nagaaki OHYAMA,1,* Koji YOSHIZUMI2 and Toshiaki HIGASHI2
1Imaging Science and Engineering Laboratory, Tokyo Institute of Technology, 4259, Nagatsuta, Midori-ku, Yokohama, 226-8503 Japan, 2University of Occupational and Environmental Health, 1-1, Iseigaoka, Yahatanishi-ku, Kitakyushu, 807-8555 Japan
(Received June 24, 1999; Accepted October 23, 1999)
Temporal series of personal images may provide us valuable information regarding progression of a disease over time. A new method, which is based on a pattern histogram approach and time series analysis, is presented to detect long-term interval changes. Highly correlated and increasing pattern characteristics are considered as real interval changes, while background and noisy patterns exhibit more chaotic variations over time. The method has been tested on a sequence of chest radiographs using corresponding regions of interest, which include an increasing number of small opacities. The experimental results show that the method is promising and can be used in clinical applications over time. A key aspect of the proposed method is its independence of feature locations; therefore, it does not require any image registration technique to extract feature characteristics of the interval changes.
Key words : interval change, image sequence, pattern histogram, autoregressive, chest radiographs