Automatic Generation of Region of Interest for Kidney Ultrasound Images Using Texture Analysis Wan M. Hafizah, Eko Supriyanto Abstract— Kidney ultrasound imaging can be used to estimate kidney size and position, and help to diagnose structural abnormalities as well as the presence of cysts and stones. However, due to the presence of speckle noise in ultrasound images, performing the segmentation methods for the kidney images were very challenging and therefore, deleting and removing the complicated background will speeds up and increases the accuracy of the segmentation process. However, in previous studies, the ROI of the kidney is manually cropped. Therefore, this study proposed an automatic region of interest (ROI) generation for kidney ultrasound images. Firstly, some techniques of speckle noise reduction were implemented consist of median filter, Wiener filter and Gaussian low-pass filter. Then texture analysis was performed by calculating the local entropy of the image, continued with the threshold selection, morphological operations, object windowing, determination of seed point and last but not least the ROI generation. This method was performed to several kidney ultrasound images with different speckle noise reduction techniques and different threshold value selection. Based on the result, it shows that for median filter, threshold value of 0.6 gave the highest TRUE ROIs which were 70%. For Wiener filter, threshold value of 0.8 gave highest TRUE ROIs which were 80% and for Gaussian low-pass filter, threshold value of 0.7 gave highest TRUE ROIs which were 100%. By using the previous result, this method has been tested also to more than 200 kidney ultrasound images. As the result, for longitudinal kidney images, out of 120 images, 109 images generate true ROI (91%) and another 11 images generate false ROI (9%). For transverse kidney images, out of 100 images, 89 images generate true ROI (89%) and 11 images generate false ROI (11%). To conclude, the method in this study can be practically used for automatic generation of US kidney ROI. Keywords—kidney, region of interest, speckle noise reduction, texture filters, ultrasound I. INTRODUCTION HE kidneys are retroperitoneal organs, attached to the posterior abdominal wall, covered with the peritoneum and protected by the lower ribs where the kidney can be found just below the liver on the right, and just below the spleen on Manuscript received October 14, 2011. W. M. Hafizah is with the Advanced Diagnostics and Progressive Human Care Research Group, UTM Skudai, 81310 Johor, Malaysia (phone: +607- 553-5273; e-mail: wmhafizah@gmail.com). E. Supriyanto is with the Clinical Science and Engineering Department, FKBSK, UTM Skudai, Johor 81310 Malaysia (e-mail: eko@utm.my). the left. During the scanning session, if the kidney were scanned in longitudinal view, the kidney will appear as football-shaped, and in transverse view, the kidney will appear as C-shaped. The normal kidney has a bright area around it, made up of perinephric fat and Gerota’s fascia. The kidney periphery part appear grainy gray, consists of renal cortex and pyramids while the central area of the kidney, the renal sinus, will appear bright (echogenic), and consists of renal sinus fat, calyces, as well as renal pelvis. Since some other organs lie close to the kidney which may give effect to the performance of other image processing methods, finding a region of interest (ROI) for kidney is quite helpful. Besides, this ROI will improve the speed and accuracy of further segmentation process. Furthermore, many existing kidney ultrasound image processing methods, including enhancement and segmentation techniques have been developed based on a manually selected ROI, not on the whole image [1-4]. Some researchers define ROI as the rough contour or initial contour of the interest object, while the other defines ROI as a rectangular region containing both the interest object kidney and some background information. For this study, a rectangular ROI will be automatically generated and any further operation will be conducted only in that rectangular ROI. For further segmentation purpose, deleting and removing the complicated background not only speeds up the segmentation process, but also increases accuracy. Therefore, this ROI generation method can be used by any other segmentation method as a preprocessing step since it only cuts the background outside the rectangular ROI while keeping the interest object (kidney) and nearby surrounding tissues untouched. Automatic ROI has been proposed by other researchers but they were not using the kidney ultrasound images. Yap et al. for example has successfully developed an algorithm to automate the manual process of region of interest (ROI) labeling in computer-aided diagnosis (CAD) for breast lesions [5, 6]. There were other researchers who also focusing on automatically detection lesions in breast ultrasound images [7- 9]. However, ROI generation of lesion in breast ultrasound T INTERNATIONAL JOURNAL OF BIOLOGY AND BIOMEDICAL ENGINEERING Issue 1, Volume 6, 2012 26