Deep learning study on the mechanism of edge artifacts in point spread function reconstruction for numerical brain images.

OBJECTIVE: Non-blinded image deblurring with deep learning was performed on blurred numerical brain images without point spread function (PSF) reconstruction to obtain edge artifacts (EA)-free images. This study uses numerical simulation to investigate the mechanism of EA in PSF reconstruction based on the spatial frequency characteristics of EA-free images. METHODS: In 256 × 256 matrix brain images, the signal values of gray matter (GM), white matter, and cerebrospinal fluid were set to 1, 0.25, and 0.05, respectively. We assumed ideal projection data of a two-dimensional (2D) parallel beam with no degradation factors other than detector response blur to precisely grasp EA using the PSF reconstruction algorithm from blurred projection data. The detector response was assumed to be a shift-invariant and one-dimensional (1D) Gaussian function with 2-5 mm full width at half maximum (FWHM). Images without PSF reconstruction (non-PSF), PSF reconstruction without regularization (PSF) and with regularization of relative difference function (PSF-RD) were generated by ordered subset expectation maximization (OSEM). For non-PSF, the image deblurring with a deep image prior (DIP) was applied using a 2D Gaussian function with 2-5 mm FWHM. The 1D object-specific modulation transfer function (1D-OMTF), which is the ratio of 1D amplitude spectrum of the original and reconstructed images, was used as the index of spatial frequency characteristics. RESULTS: When the detector response was greater than 3 mm FWHM, EA in PSF was observed in GM borders and narrow GM. No remarkable EA was observed in the DIP, and the FWHM estimated from the recovery coefficient for the deblurred image of non-PSF at 5 mm FWHM was reduced to 3 mm or less. PSF of 5 mm FWHM showed higher spatial frequency characteristics than that of DIP up to around 2.2 cycles/cm but was lower than the latter after 3 cycles/cm. PSF-RD showed almost the same spatial frequency characteristics as that of DIP above 3 cycles/cm but was inferior below 3 cycles/cm. PSF-RD has a lower spatial resolution than DIP. CONCLUSIONS: Unlike DIP, PSF lacks high-frequency components around the Nyquist frequency, generating EA. PSF-RD mitigates EA while simultaneously suppressing the signal, diminishing spatial resolution.

Compressed sensing reconstruction shortens the acquisition time for myocardial perfusion imaging: a simulation study.

Compressed sensing (CS) has been used to improve image quality in single-photon emission tomography (SPECT) imaging. However, the effects of CS on image quality parameters in myocardial perfusion imaging (MPI) have not been investigated in detail. This preliminary study aimed to compare the performance of CS-iterative reconstruction (CS-IR) with filtered back-projection (FBP) and maximum likelihood expectation maximization (ML-EM) on their ability to reduce the acquisition time of MPI. A digital phantom that mimicked the left ventricular myocardium was created. Projection images with 120 and 30 directions (360°), and with 60 and 15 directions (180°) were generated. The SPECT images were reconstructed using FBP, ML-EM, and CS-IR. The coefficient of variation (CV) for the uniformity of myocardial accumulation, septal wall thickness, and contrast ratio (Contrast) of the defect/normal lateral wall were calculated for evaluation. The simulation was performed ten times. The CV of CS-IR was lower than that of FBP and ML-EM in both 360° and 180° acquisitions. The septal wall thickness of CS-IR at the 360° acquisition was inferior to that of ML-EM, with a difference of 2.5 mm. Contrast did not differ between ML-EM and CS-IR for the 360° and 180° acquisitions. The CV for the quarter-acquisition time in CS-IR was lower than that for the full-acquisition time in the other reconstruction methods. CS-IR has the potential to reduce the acquisition time of MPI.

Combination of compressed sensing-based iterative reconstruction and offset acquisition for I-123 FP-CIT SPECT: a simulation study.

Objectives: The purpose of this study was to validate undersampled single-photon emission computed tomography (SPECT) imaging using a combination of compressed sensing (CS) iterative reconstruction (CS-IR) and offset acquisition. Methods: Three types of numerical phantoms were used to evaluate image quality and quantification derived from CS with offset acquisition. SPECT images were reconstructed using filtered back-projection (FBP), maximum likelihood-expectation maximization (ML-EM), CS-IR, and CS-IR with offset acquisition. The efficacy of CS-IR with offset acquisition was examined in terms of spatial resolution, aspect ratio (ASR), activity concentration linearity, contrast, percent coefficient of variation (%CV), and specific binding ratio (SBR). Results: The full widths at half maximum remained unchanged as the number of projections decreased in CS-IR with offset acquisition. Changes in ASRs and linearities of count density were observed for ML-EM and CS-IR from undersampled projections. The %CV obtained by CS-IR with offset acquisition was substantially lower than that obtained by ML-EM and CS-IR. There were no significant differences between the %CVs obtained from 60 projections by CS-IR with offset acquisition and from 120 projections by FBP. Although the SBRs for CS-IR with offset acquisition tended to be slightly lower than for FBP, the SBRs for CS-IR with offset acquisition did not change with the number of projections. Conclusions: CS-IR with offset acquisition can provide good image quality and quantification compared with a commonly used SPECT reconstruction method, especially from undersampled projection data. Our proposed method could shorten overall SPECT acquisition times, which would benefit patients and enable quantification with dynamic SPECT acquisitions.

A novel few-views arrangement of the fixed X-ray tubes for tomosynthesis.

PURPOSE: Tomosynthesis is a technique that reconstructs a volume image from limited-angle projection data. In conventional tomosynthesis, the examination time is long, so it can be difficult for patients to hold their breath during certain examinations, such as chest imaging. Few-views tomosynthesis, which uses a linear arrangement of fixed X-ray tubes and enables an image to be obtained within 1 s, was found to be useful in the clinical setting in our previous study. In the present study, we attempted to develop a novel few-views tomosynthesis system that can obtain images with an improved image quality. METHODS: A novel few-views arrangement of X-ray tubes was proposed and the image reconstruction method with regularization term was applied. The linear arrangement was used for the X-ray tube arrangement in our previous few-views tomosynthesis, in contrast, a circular arrangement was proposed in this study. The validation of this system was conducted with a numerical simulation and a real data experiment. RESULTS: The wider the scan angle, the more the object shadow spreads from "in-plane", allowing for artifact suppression. In the circular arrangement, the constant scan angle of θ is used, but in the linear arrangement the scan angle is set from 0 to θ. The artifacts in "out-of-plane" were more strongly suppressed in the circular arrangement than in the linear arrangement. CONCLUSIONS: Artifacts spreading in the z-direction were more strongly suppressed using the circular arrangement than the linear arrangement. Therefore, the circular arrangement was deemed appropriate for few-views tomosynthesis.

[Incident Photon Number and Reconstructed Linear Attenuation Coefficients in Iterative CT Image Reconstruction].

[Purpose] The iterative CT image reconstruction (IR) method has been successfully incorporated into commercial CT scanners as a means to promote low-dose CT with high image quality. However, the algorithm of the IR method has not been made publicly available by scanner manufacturers. Kudo reviewed the fundamentals of IR methods on the basis of the articles published by the joint research group of each manufacture that were released before and during product development (Med Imag Tech 32: 239-248, 2014). According to this review, the object function of the IR method consists of the data fidelity term (likelihood) and the regularization term. The regularization term plays a significant role in the IR method; however, it has not been clarified whether or not the variance of projection data should be included into the likelihood to act the regularization term effectively. Our purpose in this study was to investigate the relationship of the incident photon number and the reconstructed linear attenuation coefficients of the IR method by numerical experiments.[Methods] We assumed the X-ray beam was a pencil beam, and the system matrix was given by the line integral of linear attenuation coefficients because we focused on the accuracy of the reconstructed linear attenuation coefficients in the ideal photon detection system equations given by Kudo. Total variation (TV) and the relative difference function were used for regularization of the IR method. Three kinds of numerical phantoms with 256×256 pixels were used as test images. Poisson noise was added to the projection data with 256 linear sampling and 256 views over 180°. The accuracy of reconstructed linear attenuation coefficients was evaluated by the mean reconstructed value within a region of interest (ROI) and the relative root mean square errors (%RMSEs) to the object image.[Results] The linear attenuation coefficients were reconstructed accurately by the IR method including the variance of projection data into the likelihood in comparison with the IR method without including the variance. When the incident photon number ranged from 100 to 2000 for the object having a mean linear attenuation coefficient of 0.067 to 0.087 cm-1, the reconstructed linear attenuation coefficients in ROI were close to the true values. However, when the incident photon number was 50, both the accuracy and the uniformity of reconstructed images decreased.[Discussion] From the viewpoint of the visual observation, the image quality of the IR method was superior to that of the filtered back-projection (FBP) image processed with the Gaussian filter of FWHM equal to 3 pixels. For the object with a high absorber, the FBP gives linear attenuation coefficients that were lower than the true values. This phenomenon was also observed in the IR method. The projection data of CT were given by the logarithm operation of the ratio between the incident photon and the transmitted photon numbers. If the transmitted photon number happened to be equal to 0 owing to the influence of noise, it was held to a value of 1 to avoid the logarithm of zero. This process caused an error of the linear attenuation coefficients.[Conclusion] The variance of projection data should be included into the likelihood to act the regularization term effectively in the IR method.

Performance of compressed sensing-based iterative reconstruction for single-photon emission computed tomography from undersampled projection data: a simulation study in 123I-N-ω-fluoropropyl-2ß-carbomethoxy-3ß-(4-iodophenyl)nortropane imaging.

OBJECTIVE: The aim of this study was to investigate the efficacy of compressed sensing (CS)-based iterative reconstruction (CS-IR) from undersampled projection data in I-N-ω-fluoropropyl-2ß-carbomethoxy-3ß-(4-iodophenyl)nortropane single-photon emission computed tomography (SPECT). MATERIALS AND METHODS: We used the cylinder/sphere and the striatal digital phantom models. The number of projections was set at 120, 90, 60, 40, and 30 projections. SPECT images were reconstructed using filtered back-projection (FBP), maximum likelihood-expectation maximization (ML-EM), and CS-IR. The total-variation transform with local image gradient in L1-norm was adopted in our CS algorithm. The efficacy of CS-IR was examined in terms of the spatial resolution, recovery coefficient, aspect ratio (ASR), activity concentration linearity, percent coefficient of variation (%CV), and specific binding ratio. RESULTS: As the number of projections decreased, the following results were observed. No differences of the spatial resolution and activity concentration linearity were observed between reconstruction methods. However, ASR for FBP slightly increased in contrast to ML-EM and CS-IR for which ASR remained constant. There were not any clear differences between recovery coefficients obtained from each reconstruction. The %CV obtained by CS-IR was significantly superior to that obtained by other reconstructions at all number of projections: for example, the %CV obtained by 60 projection CS-IR was equivalent to that obtained by 120 projection FBP and ML-EM. The specific binding ratio did not change with the number of projections, and there were no significant differences between FBP, ML-EM, and CS-IR. CONCLUSION: We have demonstrated that CS-IR with decreased number of projections can provide a good image quality compared with commonly used SPECT reconstruction methods. This CS could help to reduce overall acquisition time in I-N-ω-fluoropropyl-2ß-carbomethoxy-3ß-(4-iodophenyl)nortropane SPECT, particularly.

[An Error Evaluation of Iterative Image Reconstruction Methods Using Chi-Square (χ2) Statistic Minimization for Poisson-Distributed Projection Data].

[Purpose] Iterative image reconstruction (IR) methods using Neyman's chi-square statistic (χ2N) or Pearson's chi-square statistic (χ2P) have been investigated in nuclear medicine. However, these chi-square statistic-based image reconstructions have never been installed on clinical nuclear medicine instruments. Mighell developed another chi-square statistic (χ2M). Recently, Mighell's chi-square statistic has been incorporated into commercial SPECT instrument aiming at high accuracy in the iterative image reconstruction from low count projection data. However, the error evaluation for χ2M was not reported by the instrument manufacturer or the joint research group involved in the product development. Therefore, it is not certain to what extent χ2M is superior to χ2N or χ2P. In this study we investigated the accuracy of the chi-square statistic-based IR methods by computer simulation.[Methods] We used two kinds of numerical phantoms (256×256 pixels) for testing root mean square error (RMSE). Phantom A was a disk that was 18.4 cm in diameter and the count density was varied from 1 count/pixel to 10 counts/pixel at intervals of 1 count/pixel in each trial. Phantom B was a disk that was 18.4 cm in diameter and the count densities for the seven disk inserts (diameter 3 cm) which were investigated were 1, 2, 3, 4, 5, 6, and 7 counts/pixel. Poisson noise was added to the projection data with 256 linear samplings and 256 views over 180°. Projection data were assumed to be without attenuation and scatter effects, because we focused our evaluation on the noise propagation from projection data to the reconstructed image that was attributable to the mathematical equations of the different types of chi-square statistic. Minimization of the chi-square statistic-based IR methods was performed by conjugate gradient method.[Results] We found the noise was suppressed by including the variance of projection data in each chi-square statistic; however, it was not suppressed sufficiently by χ2P in comparison with χ2N and χ2M. For 1000 iterations, the RMSEs of Phantom A having the count density of 1 count/pixel were 21.46±2.75, 39.21±0.71, and 12.29±0.63, obtained by χ2N, χ2P, and χ2M in 20 trials, respectively. For 2 counts/pixel, RMSEs were 5.26±0.32, 19.89±1.29, and 4.23±0.08; and for 3 counts/pixel, they were 5.34±0.56, 10.27±0.38, and 4.03±0.07. With Phantom B, RMSEs of the 3 cm disk insert having the count density of 2 counts/pixel were 7.36±0.56, 21.21±1.52, and 6.79±0.54; for 3 counts/pixel it was 5.46±0.34, 14.43±1.08, and 4.84±0.32, for χ2N, χ2P, and χ2M, respectively.(View PDF for the rest of the abstract.).

[Implementation of Statistically-Based Image Reconstruction Algorithms for CT and Numerical Evaluation of Image Quality].

[Purpose] Statistically-based image reconstruction (SIR) methods that have been incorporated into commercial CT scanners have succeeded in promoting low-dose CT with high image quality in comparison with scanners using the filtered back-projection (FBP) method. Not only researchers but also medical doctors and technologists engaged in CT studies have an interest in the algorithms of the SIR methods, however, the algorithms have not been made available to users by the CT manufacturers. Kudo reviewed the fundamentals of SIR methods on the basis of the articles published by the joint research group of each manufacturer released before product development (Med Imag Tech 32: 239-248, 2014). He classified the SIR methods into true iterative reconstruction (true IR), hybrid IR, and image space denoising (ISD) methods. His review article has made a significant contribution to the CT community of users. However, the reconstructed images obtained by those methods have not been presented. Our purpose in this study is to implement the mathematical equations of three IR methods, one each of the true IR, hybrid IR and ISD methods, and evaluate their image quality.[Methods] The system matrix of IR methods used in commercial CT scanners uses a physical photon detection process based on the finite size of an X-ray focal spot, the beam width, and the X-ray detector. However, we assumed the X-ray beam was a pencil beam and the system matrix was then given by the line integral of linear attenuation coefficients because we focus on the image quality in the ideal photon detection system equations given by Kudo. Total variation (TV) was used for regularization of the true IR, hybrid IR and ISD methods. Four kinds of numerical phantoms with 256×256 pixels were used as test images. Gaussian noise of 15, 20, 25, and 30 dB was added to the projection data with 256 linear samplings and 256 views over 180°.[Results] Root mean square errors (RMSEs) of the true IR, hybrid IR, and ISD methods were 4.28-5.70, 15.87-16.47, and 16.94-17.17, respectively. RMSE of the FBP method ranged from 27.64-33.02 and that of the FBP method processed with a Gaussian filter of FWHM (full width at half maximum) of 3 pixels ranged from 8.14-17.28. The image quality of the true IR method was superior to that of the hybrid IR and ISD methods and the FBP method.[Discussion] The noise was slightly suppressed by including the variance of projection data; however, the regularization was inevitable even if the noise levels were in the range of 25-30 dB. The noise was not suppressed sufficiently by the hybrid IR and ISD methods because the noise due to the FBP image used as the initial image for these IR methods has a dominating effect in successive reconstruction or denoise processing. Mathematical equations of each IR method were easily realized by observing the intermediate images such as the regularization term of the iteration process. In addition to these equations, the reconstructed images by the SIR methods and their RMSEs presented in this study are useful in CT research.[Conclusions] The fundamental point of SIR methods is the regularization term used in minimizing the object function.

Undersampling patterns in k-space for compressed sensing MRI using two-dimensional Cartesian sampling.

In compressed sensing magnetic resonance imaging (CS-MRI), undersampling of k-space is performed to achieve faster imaging. For this process, it is important to acquire data randomly, and an optimal random undersampling pattern is required. However, random undersampling is difficult in two-dimensional (2D) Cartesian sampling. In this study, the effect of random undersampling patterns on image reconstruction was clarified using phantom and in vivo MRI, and a sampling pattern relevant for 2D Cartesian sampling in CS-MRI is suggested. The precision of image restoration was estimated with various acceleration factors and extents for the fully sampled central region of k-space. The root-mean-square error, structural similarity index, and modulation transfer function were measured, and visual assessments were also performed. The undersampling pattern was shown to influence the precision of image restoration, and an optimal undersampling pattern should be used to improve image quality; therefore, we suggest that the ideal undersampling pattern in CS-MRI for 2D Cartesian sampling is one with a high extent for the fully sampled central region of k-space.

[Evaluation of the Image Registration Program for Portal Images Using CR and DRR Images in Radiation Therapy].

In this study, computer simulations and experiments were used to verify the accuracy of a two-dimensional image registration program (program) for portal images that we previously developed. The program used a computed radiography cassette system and digitally reconstructed radiography images as planning images for external beam radiation therapy. Using this program, we also investigated the reason two-dimensional automatic image registration images experienced large misregistration in clinical practice using commercial image registration systems. Mutual information and normalized mutual information were used as the registration criteria. To investigate the influence of image background with or without a region of interest (ROI), results of image registrations were compared. Parameters of image registration were defined as translation in the horizontal and vertical directions (x and y, respectively) and rotation (θ) around the axis perpendicular to the x-y plane. There was no significant difference in image registration arising from the difference between mutual information and normalized mutual information. Image registration was improved with a ROI. Regardless of the registration criteria, errors in image registration with a ROI in the experimental study were ≤1.2 mm in directions x and y and ≤1.0 degree in rotation θ. We found that image registration required setting up as close to the planned position as possible.

[Numerical and Visual Evaluation of Compressed Sensing MRI Using 3D Cartesian Sampling].

We performed numerical and visual evaluation of compressed sensing MRI (CS-MRI) using 3D Cartesian sampling by numerical simulation. Three brain anatomical ROIs (white matter, gray matter, cerebrospinal fluid) of a T1-weighted image (T1WI), a T2-weighted image (T2WI) and a proton density-weighted image (PDWI) were used for numerical evaluation. Sampling ratio of the Cartesian grid was 30%. Reconstruction was performed by the projection onto convex sets (POCS) method with soft thresholding, subject to data fidelity constraints. In the absence of noise, RMSE of 3D-T1WI was 1.50, ant that of the 2D-T1WI of the transverse plane was in the range of 1.06 to 1.54; anatomical ROIs was in the range of 0.75 to 2.80; those of T2WI were 3.20, 2.77 to 3.06, and 1.81 to 4.51; those of PDWI were 1.69, 1.33 to 1.49, and 1.08 to 1.86. Visual evaluation was performed by three radiologists on the basis of three categories: artifact, anatomical structure, and tissue contrast. Average score of the visual evaluation indicated that anatomical structure and tissue contrast of CS images were equal to those of the original image, although a few artifacts were visible. If noise level was assumed to be 20 dB or less, anatomical structure and tissue contrast were not significantly degraded compared to noise-free CS images.

[Numerical and Visual Evaluations of Compressed Sensing MRI Using 2D Radial Sampling].

Two-dimensional radial MRI using compressed sensing (2D radial CS) enables incoherence sampling in k space unlike conventional Cartesian MRI, however 2D radial CS has not been sufficiently investigated. Numerical and visual evaluations of 2D radial CS were performed in this paper. Three brain anatomical ROIs (white matter, gray matter, cerebrospinal fluid) of a T1-weigthted image (T1WI), a T2-weighted image (T2WI) and a proton density-weighted image (PDWI) were used for the numerical evaluation. The Brainweb MRI Data Base was used for test images. Projection of 80 spokes with linear sampling of 256 pixels was used. Reconstruction was performed by minimizing the L1 norm of a transformed image using wavelet transform and spatial finite-differences (total variation), subject to data fidelity constraint. In the absence of noise, the root mean square error (RMSE) of T1WI was in the range of 3.75 to 5.05; that of the anatomical region of interests (ROIs) was in the range of 1.54 to 10.24; those of T2WI were 8.75 to 11.65 and 4.31 to 6.99; and those of PDWI were 3.44 to 4.46 and 1.34 to 3.09. Visual evaluation was performed by three radiologists on the basis of three categories: artifact, anatomical structure, and tissue contrast. Average percent scores of the visual evaluation were 96% for T1WI, 74-81% for T2WI, and 81-89% for PDWI.

[Numerical and Visual Evaluations of Compressed Sensing MRI Using 2D Cartesian Sampling].

This paper describes numerical and visual evaluations of compressed sensing MRI (CS-MRI) using 2D Cartesian sampling by numerical simulation. The BrainWeb MRI Data Base was used for test images. Three brain anatomical ROIs (white matter, gray matter, cerebrospinal fluid) of a T1-weighted image (T1WI), a T2-weighted image (T2WI) and a proton density-weighted image (PDWI) were used for the numerical evaluation. Sampling ratio was 50%. Reconstruction was performed by minimizing the L1 norm of a transformed image using wavelet transform and total variation, subject to data fidelity constraints. The conjugate gradient method was used in the minimization of the object function. In the absence of noise, the root mean square error (RMSE) of T1WI was in the range of 2.99 to 3.57; that of the anatomical region of interests (ROIs) was in the range of 1.77 to 8.53; those of T2WI were 4.72 to 5.65 and 3.28 to 5.54; and those of PDWI were 1.91 to 2.36 and 1.32 to 2.09. Visual evaluation was performed by three radiologists on the basis of three categories: artifact, anatomical structure, tissue contrast. CS image quality was nearly equal to that of the original image, although a few artifacts were visible. If the noise level was assumed to be 30 dB or less, T1-CS image and PD-CS images were not significantly degraded compared to noise-free images.

Iterative image reconstruction that includes a total variation regularization for radial MRI.

This paper presents an iterative image reconstruction method for radial encodings in MRI based on a total variation (TV) regularization. The algebraic reconstruction method combined with total variation regularization (ART_TV) is implemented with a regularization parameter specifying the weight of the TV term in the optimization process. We used numerical simulations of a Shepp-Logan phantom, as well as experimental imaging of a phantom that included a rectangular-wave chart, to evaluate the performance of ART_TV, and to compare it with that of the Fourier transform (FT) method. The trade-off between spatial resolution and signal-to-noise ratio (SNR) was investigated for different values of the regularization parameter by experiments on a phantom and a commercially available MRI system. ART_TV was inferior to the FT with respect to the evaluation of the modulation transfer function (MTF), especially at high frequencies; however, it outperformed the FT with regard to the SNR. In accordance with the results of SNR measurement, visual impression suggested that the image quality of ART_TV was better than that of the FT for reconstruction of a noisy image of a kiwi fruit. In conclusion, ART_TV provides radial MRI with improved image quality for low-SNR data; however, the regularization parameter in ART_TV is a critical factor for obtaining improvement over the FT.

[Development of a Text-Data Based Learning Tool That Integrates Image Processing and Displaying].

We developed a text-data based learning tool that integrates image processing and displaying by Excel. Knowledge required for programing this tool is limited to using absolute, relative, and composite cell references and learning approximately 20 mathematical functions available in Excel. The new tool is capable of resolution translation, geometric transformation, spatial-filter processing, Radon transform, Fourier transform, convolutions, correlations, deconvolutions, wavelet transform, mutual information, and simulation of proton density-, T1-, and T2-weighted MR images. The processed images of 128 x 128 pixels or 256 x 256 pixels are observed directly within Excel worksheets without using any particular image display software. The results of image processing using this tool were compared with those using C language and the new tool was judged to have sufficient accuracy to be practically useful. The images displayed on Excel worksheets were compared with images using binary-data display software. This comparison indicated that the image quality of the Excel worksheets was nearly equal to the latter in visual impressions. Since image processing is performed by using text-data, the process is visible and facilitates making contrasts by using mathematical equations within the program. We concluded that the newly developed tool is adequate as a computer-assisted learning tool for use in medical image processing.

[Development of the image registration program for portal and DRR images in radiation therapy].

In this article, the authors propose an image registration program of portal images and digitally reconstructed radiography (DRR) images used as simulation images for external beam radiation therapy planning. First, the center of the radiation field in a portal image taken using a computed radiograhy cassette is matched to the center of the portal image. Then scale points projected on a DRR image and the portal image are deleted, and the portal image with the radiation field is extracted. Registration of the DRR and portal images is performed using mutual information as the registration criterion. It was found that the absolute displacement misregistrations in two directions (x, y) were 1.2 +/- 0.7mm and 0.5 +/- 0.3 mm, respectively, and rotation disagreement about the z axis 0.3 +/- 0.3 degrees. It was concluded the proposed method was applicable to image registration of portal and DRR images in radiation therapy.