講演抄録/キーワード |
講演名 |
2009-01-21 09:50
[ポスター講演]Computer-Assisted Texture Analysis of DMSA Scintigraphy in Diagnosis of Pyelonephritis ○Duck-Joo Choi・Kyung Hoon Hwang・Minkyung Lee・Wonsick Choe(Gachon Univ. Gil Hospital) MI2008-168 |
抄録 |
(和) |
DMSA renal scan has been used to evaluate renal parenchymal lesions such as pyelonephritis. However, it is frequently difficult to differentiate pelvocalyectatic defects from true cortical defects. We investigated whether DMSA renal scan images could be effectively evaluated by computer-assisted texture analysis such as fractal texture analysis. We retrospectively analyzed 35 DMSA renal scan images - 12 normal kidneys, 11 with pelvocalyectatic cortical defects and 12 with true cortical defects. The kidneys on DMSA scan images were segmented by the binary threshold and subtraction techniques, and the texture analysis was done using a fractal texture analysis tool (Image-J). The values from the texture analysis were compared between the three groups using student's t-test and ROC curve analysis. In addition, patients。ッ urinanalysis laborotary data was added to the texture analysis values and analyzed using discriminant analysis. The kidney images with true cortical defects showed higher fractal texture values than those of normal or pelvocalyectatic kidneys, but was not statistically significant. There was a significant difference in fractal texture values only between normal and pelvocalyectatic kidneys (p<0.05). ROC curve analysis in fractal texture values between the groups of true cortical defects and pelvocalyectases demonstrated the best cut-off value of 1.703 (Sensitivity of 91.7% and specificity of 72.7%) and area under curve of 0.833. Discriminant analysis combining the fractal texture values and patients。ッ urinanalysis laborotary data showed no further improvement in differential diagnostic accuracy. Although the number of images is limited, computer-assisted fractal texture analysis may be helpful to differentiate pelvocalyectatic defects from true defects. However, further studies including more cases and clinical parameters are warranted. |
(英) |
DMSA renal scan has been used to evaluate renal parenchymal lesions such as pyelonephritis. However, it is frequently difficult to differentiate pelvocalyectatic defects from true cortical defects. We investigated whether DMSA renal scan images could be effectively evaluated by computer-assisted texture analysis such as fractal texture analysis. We retrospectively analyzed 35 DMSA renal scan images - 12 normal kidneys, 11 with pelvocalyectatic cortical defects and 12 with true cortical defects. The kidneys on DMSA scan images were segmented by the binary threshold and subtraction techniques, and the texture analysis was done using a fractal texture analysis tool (Image-J). The values from the texture analysis were compared between the three groups using student's t-test and ROC curve analysis. In addition, patients。ッ urinanalysis laborotary data was added to the texture analysis values and analyzed using discriminant analysis. The kidney images with true cortical defects showed higher fractal texture values than those of normal or pelvocalyectatic kidneys, but was not statistically significant. There was a significant difference in fractal texture values only between normal and pelvocalyectatic kidneys (p<0.05). ROC curve analysis in fractal texture values between the groups of true cortical defects and pelvocalyectases demonstrated the best cut-off value of 1.703 (Sensitivity of 91.7% and specificity of 72.7%) and area under curve of 0.833. Discriminant analysis combining the fractal texture values and patients。ッ urinanalysis laborotary data showed no further improvement in differential diagnostic accuracy. Although the number of images is limited, computer-assisted fractal texture analysis may be helpful to differentiate pelvocalyectatic defects from true defects. However, further studies including more cases and clinical parameters are warranted. |
キーワード |
(和) |
Computer-assisted / Texture analysis / DMSA / Renal scan / Pyelonephritis / / / |
(英) |
Computer-assisted / Texture analysis / DMSA / Renal scan / Pyelonephritis / / / |
文献情報 |
信学技報, vol. 108, no. 385, MI2008-168, pp. 493-494, 2009年1月. |
資料番号 |
MI2008-168 |
発行日 |
2009-01-12 (MI) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
MI2008-168 |