An Automated Delineation Method of Tumor Biological Target Volume Based on PET/CT Image
DAI Yingxin1，LUO Shiyuan2，WU Ruixian1，WANG Zhiguo1
1. Department of Nuclear Medicine, The General Hospital of Northern Theater Command, Shenyang Liaoning 110016, China丨
2. College of Medicine and Biological Information Engineering, Northeastern University, Shenyang Liaoning 110169, China
Abstract：Objective Improve the traditional random walk algorithm to improve the accuracy of tumor biological target area
delineation. Methods In this paper, the texture feature information of PET images is first extracted by gray level co-occurrence
matrix (GLCM) in order to get a more accurate biological target volume. Next, this paper use fuzzy c-means clustering algorithm
(FCM) to divide the image into high uptake area, low uptake area and medium uptake area that containing tumor infiltration
boundary. Finally, The high uptake area is marked as the foreground seed point of the random walk algorithm, and the low uptake
area is all marked as the background seed point and the weight function of the edge of the random walk algorithm is calculated
by combining the variance texture feature of the image. Results This paper evaluates the experimental results by similarity and
sensitivity. the similarity index of the method is about 14.2% higher than the threshold method, which is about 11.2% higher than
the traditional random walk algorithm丨 the sensitivity index is about 19.8% higher than the threshold method, which is about 16.4%
higher than the traditional random walk algorithm. The similarity and sensitivity differences obtained by the three methods were
statistically significant (F=82.06, F=214.12, all P<0.01). Conclusion The experimental results show that the proposed method has a
greater improvement in similarity and sensitivity than the threshold method and the traditional random walk algorithm.
戴盈欣1，雒士源2，吴锐先1，王治国1. 基于PET/CT图像肿瘤生物靶区自动勾画算法的研究[J]. 中国医疗设备, 2021, 36(3): 85-89.
DAI Yingxin1，LUO Shiyuan2，WU Ruixian1，WANG Zhiguo1. An Automated Delineation Method of Tumor Biological Target Volume Based on PET/CT Image. China Medical Devices, 2021, 36(3): 85-89.
