Usability of unbiased nonlocal means for de-noising intraoperative magnetic resonance images in neurosurgery.

PURPOSE: Intraoperative magnetic resonance imaging (iMRI) is a powerful tool that allows real-time image-guided excision of brain tumors. However, low magnetic field iMRI devices may produce low-quality images due to nonideal imaging conditions in the operating room and additional noise of unknown origin. The purpose of this study was to evaluate a three-dimensional unbiased nonlocal means filter for iMRI (UNLM-i) that we developed in order to enhance image quality and increase the diagnostic value of iMRI. METHODS: We first evaluated the effect of UNLM by assessing the modulation transfer function (MTF) and Weiner spectrum (WS) of UNLM in simulated imaging. We then tested the diagnostic value of UNLM-i de-noising by applying it to a series of randomly chosen iMR images that were assessed by 4 neurosurgeons and 4 radiological technologists using a 5-point rating scale to compare 13 parameters, including tumor visibility, edema, and sulci, before and after de-noising. RESULTS: Unbiased nonlocal means provided better MTF in comparison with other filters, and the WS for UNLM de-noising was reduced for all spatial frequencies. Postprocessing UNLM-i allowed de-noising with preserved edges and >twofold improvement in the signal-to-noise ratio without extending the MRI scanning time (p< 0.001) . The diagnostic value of UNLM-i de-noising was rated as "superior" or "better" in >80 % of cases in terms of contrast between white and gray matter and visibility of sulci, tumor, and edema (p< 0.001). CONCLUSIONS: Unbiased nonlocal means filter for iMRI de-noising proved very useful for image quality enhancement and assistance in the interpretation of iMR images.

Adapting non-local means of de-noising in intraoperative magnetic resonance imaging for brain tumor surgery.

In image-guided brain tumor surgery, intraoperative magnetic resonance imaging (iMRI) is a powerful tool for updating navigational information after brain shift, controlling the resection of brain tumors, and evaluating intraoperative complications. Low-field iMRI scans occasionally generate a lot of noise, the reason for which is yet to be determined. This noise adversely affects the neurosurgeons' interpretations. In this study, in order to improve the image quality of iMR images, we optimized and adapted an unbiased non-local means (UNLM) filter to iMR images. This noise appears to occur at a specific frequency-encoding band. In order to adapt the UNLM filter to the noise, we improved the UNLM, so that de-noising can be performed at different noise levels that occur at different frequency-encoding bands. As a result, clinical iMR images can be de-noised adequately while preserving crucial information, such as edges. The UNLM filter preserved the edges more clearly than did other classical filters attached to an anisotropic diffusion filter. In addition, UNLM de-noising can improve the signal-to-noise ratio of clinical iMR images by more than 2 times (p < 0.01). Although the computational time of the UNLM processing is very long, post-processing of UNLM filter images, for which the parameters were optimized, can be performed during other MRI scans. Therefore, The UNLM filter was more effective than increasing the number of signal averages. The iMR image quality was improved without extension of the MR scanning time. UNLM de-noising in post-processing is expected to improve the diagnosability of low-field iMR images.

[Development of a cone-beam CT system for radiological technologist education].

For radiological technologists, it is very important to understand the principle of computed tomography (CT) and CT artifacts derived from mechanical and electrical failure. In this study, a CT system for educating radiological technologists was developed. The system consisted of a cone-beam CT scanner and educational software. The cone-beam CT scanner has a simple structure, using a micro-focus X-ray tube and an indirect-conversion flat panel detector. For the educational software, we developed various educational functions of image reconstruction and reconstruction parameters as well as CT artifacts. In the experiments, the capabilities of the system were evaluated using an acrylic phantom. We verified that the system produced the expected results.

Development of quality control system for flat-panel detectors.

The characteristics of flat-panel detectors (FPD) are degraded by exposure to radiation. Degradation in a FPD progresses locally and has a nonlinear relationship to the radiation dose. In order to manage FPD systems properly, one must perform quality control (QC) such as evaluation of image degradation. However, no evaluation method for degradation has been established. In this paper, we first review the structure and degradation mechanism of FPDs, and then we propose a daily QC system for FPDs. To evaluate the degradation of FPDs, we investigated the number of defective pixels and lines, as well as the offset level of the pixel output. Furthermore, we developed daily QC software for FPD that can evaluate the image quality and is operationally simple. In the experiments, an indirect-conversion type FPD was evaluated by our proposed system. The offset level of FPD increased exponentially with X-ray exposure; no trends were seen for the number of defective pixels and defective lines. The required time for the evaluation of an FPD was about 1 min, and no special skills were needed for the analysis. These results indicate that our system may be useful for daily QC of FPDs.

Skin shift and its effect on navigation accuracy in image-guided neurosurgery.

Neuronavigation systems have been developed for image-guided neurosurgery to aid in the accurate resection of malignant brain tumors. Therefore, the accuracy of the neuronavigation is important. However, many factors can reduce the navigation accuracy during surgery. Before craniotomy, the patient's head is secured to a head frame with head pins; this fixation may cause displacement of fiducial markers and reduce the accuracy. We term this phenomenon skin shift. In this study, the extent of skin shift and its effect on navigation accuracy were determined by use of both preoperative magnetic resonance imaging (MRI) scans acquired before fixation and intraoperative MRI scans acquired after fixation. We measured the displacement of the fiducial markers by using fusion images obtained by integrating preoperative and intraoperative MRI scans. We also evaluated the navigation accuracy of registration based on preoperative and on intraoperative MRI. The mean (± SD) extent of skin shift was 5.34 (± 2.65) mm. The navigation accuracy of registration based on preoperative MRI was 4.06 (± 2.25) mm, and that of registration based on intraoperative MRI was 2.51 (± 1.32) mm. No significant correlation was observed between the extent of skin shift and the distance between the head pins and fiducial markers (p > 0.05). The navigation accuracy of registration based on intraoperative MRI was significantly higher than that of registration based on preoperative MRI (p < 0.001). The results indicated that skin shift was caused by the fixation, and that this shift reduced the navigation accuracy. Intraoperative MRI can correct the effect of skin shift.

[Development of automatic navigation measuring system using template-matching software in image guided neurosurgery].

An image-guided neurosurgery and neuronavigation system based on magnetic resonance imaging has been used as an indispensable tool for resection of brain tumors. Therefore, accuracy of the neuronavigation system, provided by periodic quality assurance (QA), is essential for image-guided neurosurgery. Two types of accuracy index, fiducial registration error (FRE) and target registration error (TRE), have been used to evaluate navigation accuracy. FRE shows navigation accuracy on points that have been registered. On the other hand, TRE shows navigation accuracy on points such as tumor, skin, and fiducial markers. This study shows that TRE is more reliable than FRE. However, calculation of TRE is a time-consuming, subjective task. Software for QA was developed to compute TRE. This software calculates TRE automatically by an image processing technique, such as automatic template matching. TRE was calculated by the software and compared with the results obtained by manual calculation. Using the software made it possible to achieve a reliable QA system.

Evaluation of errors influencing accuracy in image-guided neurosurgery.

Neurosurgeons sometimes find it difficult to locate tumors precisely during microsurgery, particularly tumors located in the brain parenchyma because of the absence of boundaries in this region. Image-guided neurosurgical techniques conducted with the help of neuronavigation systems have been developed and have gained importance recently. Accuracy is vital during image-guided neurosurgery. We used a phantom to evaluate the errors introduced during navigation. The three errors evaluated were skin-shift, marker-gap, and table-rotation errors. The skin-shift error occurs if the fiducial markers positioned on the scalp move when the head is fixed to a head holder with head pins. The marker-gap error occurs when the marker ball is positioned incorrectly in the marker socket. The table-rotation error occurs when the operating table is rotated for obtaining an intraoperative MR image and then returned to its original position. Our results indicated that skin shift decreased the navigation accuracy by an error of more than 4 mm, and the gap between the marker ball and the socket resulted in a decrease in navigation accuracy by an error of more than 5 mm. The table-rotation error was found to be negligible. The errors can be avoided by ensuring that the fiducial markers are positioned appropriately on the scalp and the marker ball is fitted well in the marker socket. A phantom is useful for evaluating accuracy, particularly for evaluating errors intrinsic to different operating rooms. Periodic quality assurance by use of a phantom in each operating room might aid in maintaining the accuracy of neuronavigation.

[Characteristic of evaluation of new image filter devised for improvement of workflow in CT examination].

In this study, we processed reconstructed images with a new image filter (Siemens Medical Systems, Adaptive Image Filter: AIF). As one of its characteristics, the filter uses low-pass filtering. When an image that emphasizes a high-frequency element is changed to one with a reduced high-frequency element, an image suitable for clinical use can be obtained. For the resolving characteristic and the noise characteristic, we evaluated the degree of transition, using the modulation transfer factor (MTF) and Wiener spectrum (WS). Moreover, we used the signal-to-noise ratio (SNR) to examine the total loss of signal detection capability after use of the AIF. The results showed that, when we changed to images using the AIF and made it the same level as B30 and U40, we had to hold down the kernel level to at least B60 and U80. The use of an image filter did not recognize less of an SNR in comparison with the reconstruction image. In this study, changes in detailed characteristics of the image and SNR could be evaluated objectively using the AIF. As for the effective method by AIF that raw data isn't used for is available for the control of an image (times) using reconstruction and the change of an image on database. Therefore, we consider the AIF useful to improve workflow in CT examinations.

[Complexity evaluation of data transfer in the Z-axis direction in 4-row multislice spiral CT using fractal dimension analysis].

Image reconstruction in multislice spiral/helical computed tomography (MSCT) consists of a package of data on the arbitrary direction of the Z-axis that can be collected by active detector arrays. Thus the recombined data vary with each spiral pitch. In certain cases of spiral pitch, data compression can occur, and the spiral artifacts that are characteristic of MSCT would change. In our study, we evaluated image complications by fractal dimensions, because the geometrical patterns from a conic phantom are closely related to data transfer in the direction of the Z-axis in spiral pitches. We hoped to establish useful spiral pitches and slice collimation for clinical use in a 4-row MSCT scanner. By employing a conic phantom of 120 mm in diameter and a cone angle of 100 degrees, we measured the fractal dimension of the conic phantom image by making a binary to outline from 2.0 to 8.0 of various slice collimations. Moreover, in order to evaluate the correlation between fractal dimensions and image artifacts, we confirmed the influence of spiral pitch and reconstruction slice thickness for clinical use. We found that, when the reconstruction slice thickness was the same, the cross section of the conic phantom that was from thin-slice collimations was more similar to an actual circle than that of wide-slice collimations. The former deserved a low value and showed slight changes, and, therefore, its fractal dimensions were fixed. As a phenomenon worthy of attention, when we employed wide-slice collimations (4x5.0 mm) during peculiar low spiral pitches of 2.5 to 3.0 fractal dimensions remained low and similar to an actual circle. By these analyses of the influence of data transfer in the direction of the Z-axis, we found that spiral pitch influenced the rate of slice collimation used for data acquisition closely to the reconstruction slice thickness. Based on these findings, when slice collimations and reconstruction slice thickness should be made equal, we estimated that spiral pitches of low image artifacts in 4row MSCT ranged from 2.5 to 3.0 using fractal dimensions. We consider that a new adaptation of fractal dimension analysis is possible when it is used as an index in determining protocols.

[Measurement of modulation transfer function in Z-axis for multi-slice spiral CT using the micro-disk method: comparison with the bead method and examination of geometric influence].

Many methods of measuring the section-sensitive profile (SSP) of computed tomography (CT) by the input of a delta function have been reported. In Japan, the bead method is used as a common measurement because of the high flexibility of the multi-purpose method. However, the intensity of the response of the bead method tends to decline, creating a relatively large error in the base of section-sensitive profiles. A problem is considered to be the accuracy of measurement in evaluating spatial resolution along the z-axis in multi-slice spiral/helical CT (MSCT). We therefore evaluated the modulation transfer function (MTF) by conducting research with the micro-disk method (100 micro m thickness and 1.0 mmphi diameter) and the bead method (1.0 mmphi diameter) for the same input width. Moreover, in the micro-disk method, we also examined alignment, circular region of interest (ROI), and the energy characteristic. Our comparison of MTFs obtained by the micro-disk method and the bead method showed that the former resulted in a higher value and lower standard deviation. The difference was significant at p<0.01. Measurement using the micro-disk method did not show significant differences in terms of alignment and ROIs. Moreover, the energy characteristic was not indicated. This research demonstrated that the accuracy of MTF measurement with the micro-disk method was greater than that with the bead method, and it was found that there was no influence on the actual measurement level of the geometric structure with the micro-disk method.

Evaluation of Linearity for the Radiophotoluminescence Glass Dosimeter Based on Monochromatic X-rays.

Although low energy X-rays have been utilized for mammography, their safety in medical use is a matter of concern. Characteristics of the radiophotoluminescence glass dosimeter, GD-403, consisting of a glass element and filters, were investigated with respect to monochromatic X-rays obtained from a synchrotron radiation for personal monitoring of low energy photons. We focused on low energy X-rays ranging from 8 to 20 keV to study the linearity of the GD-403 response between photon fluence and dose equivalent. The GD-403 was placed on a tough water phantom and irradiated using an 11-15 mm x 0.1-7 mm beam for modulation of the photon fluence. The tough water phantom could be moved through a distance of 110-150 mm with a stepping motor. For the dose equivalent at 1cm depth (H1), 3mm (H3) and 70 &mgr;m (H70), the GD-403 showed sufficient linearities against the photon fluences in the energy regions of 8 to 20 keV, 13 to 20 keV and 13 to 20 keV, respectively. However, H3 and H70 did not provide sufficient linearities in the energy region of 8 to 12 keV. Moreover, we compared the result in this experiment with the value calculated from the absorbed dose of air using the mass absorption coefficient for the X-ray energy ranging from 10 to 20 keV.