Development of a Simple Image Processing Application that Makes Abdominopelvic Tumor Visible on Positron Emission Tomography/Computed Tomography Image

Introduction

Abdominopelvic tumors on positron emission tomography/computed tomography (PET/CT) images are hard to visualize due to the presence of radioactive urine (¹⁸F-Fluorodeoxyglucose) in the urinary bladder. Currently, diuretics are administered to flush radioactive urine that makes tumor visible on postdiuretic PET/CT image.[1234] It is possible to protect patients from side effect of diuretic administration and an extra radiation burden due to CT scan of PET/CT study. Saroha et al.[5] have claimed 88% probability of detection of abdominopelvic tumor using suprathreshold stochastic resonance (SSR)-based image processing method using a sample of thirty prediuretic PET/CT images. However, to utilize the advantage of this technique, the users need to have a tool that is handy, once the images are exported from the workstation. In this study, we have developed and validated an application that processes PET/CT images based on an SSR image processing method to reveal the presence of tumor.

Methods

Requirement analysis for the application program was made by conducting a brainstorming session among authors. It was decided that the program should be developed for personal computer, Microsoft Windows, and provide option to load PET/CT images (jpg, bmp, tiff, and dcm format) and display output images in a window with a slider control for selecting best visualization. It was implemented using MATLAB image processing toolbox as it has many verified and tested built-in functions. The flowchart in Figure 1 shows the steps of the application, including the computation performed for generating SSR.

Close

Figure 1 Flow chart of the program. The steps pertaining to the technique of suprathreshold stochastic resonance are indicated such. Dgdfm is the degree of freedom, an input parameter of Chi-square probability distribution

Export to PPT

Results and Discussion

The program automatically detects the format of the input file from its file extension. It can process the image files written in dcm, jpeg, jpg, png, and bmp format. If the input file is not from among these formats, the program exits with a descriptive message to the operator specifying correct usage.

The program has been tested on Microsoft Windows™ to display the output images both on the monochrome as well as on color display devices. The interface of the application is intuitively simple and uses familiar idiom of a graphical user interface experience. The menu-driven interface offers an on-screen list of choices, in which selections can be accomplished by one or more keystrokes [Figure 2] or point-and-click using a mouse. This eases the operator's memorization burden.

Close

Figure 2 Step-wise program flows. User types “SsrNewScript” on MATLAB command prompt. Program provides option for the selection of input image, displays selected image, user places rectangular region of interest on image, output images with slider

Export to PPT

To begin user types “SsrNewScript” on MATLAB command prompt as shown in Figure 2. The program provides option to the user for selection of input image. It displays image in a new window, and when user moves mouse pointer on the image, mouse pointer changes in shape. Holding mouse button down and dragging it, a rectangular region of interest (ROI) can be placed. At a double click on ROI, the program takes the image included within the ROI for processing using SSR image processing method. After processing the input image, the program displays 18 output images to be reviewed in a separate window having slider option so that user can view the processed images one by one.

The application was used to process the same image set as in[5] and also ten additional images; making a total set of forty prediuretic PET/CT images. It was found that the results reported in[5] were not only preserved but also were viewable with better ease. It is envisaged that this application should help in familiarization and deciding the usefulness of the SSR method when it is applied on a large number of prediuretic PET/CT images. The only requirement is that the user screen captures the input PET/CT slice image on which he is suspecting the presence of tumor that is not clearly visible on inspection.

The presence of abdominopelvic tumors can be detected even on a prediuretic PET/CT image using the developed application.

Conclusion

The developed application enables the user to select, process, and view output images in the process of utilizing SSR to detect the presence of abdominopelvic tumor on prediuretic PET/CT image.