Kenta Miwa

FDG uptake heterogeneity evaluated by fractal analysis improves the differential diagnosis of pulmonary nodules

PURPOSE The present study aimed to determine whether fractal analysis of morphological complexity and intratumoral heterogeneity of FDG uptake can help to differentiate malignant from benign pulmonary nodules. MATERIALS AND METHODS We retrospectively analyzed data from 54 patients with suspected non-small cell lung cancer (NSCLC) who were examined by FDG PET/CT. Pathological assessments of biopsy specimens confirmed 35 and 19 nodules as NSCLC and inflammatory lesions, respectively. The morphological fractal dimension (m-FD), maximum standardized uptake value (SUV(max)) and density fractal dimension (d-FD) of target nodules were calculated from CT and PET images. Fractal dimension is a quantitative index of morphological complexity and tracer uptake heterogeneity; higher values indicate increased complexity and heterogeneity. RESULTS The m-FD, SUV(max) and d-FD significantly differed between malignant and benign pulmonary nodules (p<0.05). Although the diagnostic ability was better for d-FD than m-FD and SUV(max), the difference did not reach statistical significance. Tumor size correlated significantly with SUV(max) (r=0.51, p<0.05), but not with either m-FD or d-FD. Furthermore, m-FD combined with either SUV(max) or d-FD improved diagnostic accuracy to 92.6% and 94.4%, respectively. CONCLUSION The d-FD of intratumoral heterogeneity of FDG uptake can help to differentially diagnose malignant and benign pulmonary nodules. The SUV(max) and d-FD obtained from FDG-PET images provide different types of information that are equally useful for differential diagnoses. Furthermore, the morphological complexity determined by CT combined with heterogeneous FDG uptake determined by PET improved diagnostic accuracy.

A Monte Carlo study on 223Ra imaging for unsealed radionuclide therapy

PURPOSE Radium-223 ((223)Ra), an α-emitting radionuclide, is used in unsealed radionuclide therapy for metastatic bone tumors. The demand for qualitative (223)Ra imaging is growing to optimize dosimetry. The authors simulated (223)Ra imaging using an in-house Monte Carlo simulation code and investigated the feasibility and utility of (223)Ra imaging. METHODS The Monte Carlo code comprises two modules, hexagon and nai. The hexagon code simulates the photon and electron interactions in the tissues and collimator, and the nai code simulates the response of the NaI detector system. A 3D numeric phantom created using computed tomography images of a chest phantom was installed in the hexagon code. (223)Ra accumulated in a part of the spine, and three x-rays and 19 γ rays between 80 and 450 keV were selected as the emitted photons. To evaluate the quality of the (223)Ra imaging, the authors also simulated technetium-99m ((99m)Tc) imaging under the same conditions and compared the results. RESULTS The sensitivities of the three photopeaks were 147 counts per unit of source activity (cps MBq(-1); photopeak: 84 keV, full width of energy window: 20%), 166 cps MBq(-1) (154 keV, 15%), and 158 cps MBq(-1) (270 keV, 10%) for a low-energy general-purpose (LEGP) collimator, and those for the medium-energy general-purpose (MEGP) collimator were 33, 13, and 8.0 cps MBq(-1), respectively. In the case of (99m)Tc, the sensitivity was 55 cps MBq(-1) (141 keV, 20%) for LEGP and 52 cps MBq(-1) for MEGP. The fractions of unscattered photons of the total photons reflecting the image quality were 0.09 (84 keV), 0.03 (154 keV), and 0.02 (270 keV) for the LEGP collimator and 0.41, 0.25, and 0.50 for the MEGP collimator, respectively. Conversely, this fraction was approximately 0.65 for the simulated (99m)Tc imaging. The sensitivity with the LEGP collimator appeared very high. However, almost all of the counts were because of photons that penetrated or were scattered in the collimator; therefore, the proportions of unscattered photons were small. CONCLUSIONS Their simulation study revealed that the most promising scheme for (223)Ra imaging is an 84-keV window using an MEGP collimator. The sensitivity of the photopeaks above 100 keV is too low for (223)Ra imaging. A comparison of the fractions of unscattered photons reveals that the sensitivity and image quality are approximately two-thirds of those for (99m)Tc imaging.

Evaluation of a revised version of computer-assisted diagnosis system, BONENAVI version 2.1.7, for bone scintigraphy in cancer patients

ObjectiveBONENAVI is a computer-assisted diagnosis system that analyzes bone scintigraphy automatically. We experienced more than a few segmentation errors with the previous BONENAVI version (2.0.5). We have since obtained a revised version (2.1.7) and evaluate it.MethodsBone scans of patients were analyzed by BONENAVI version 2.0.5 and a revised version 2.1.7 with regard to segmentation errors, sensitivity, and specificity. Patients with skeletal metastases from prostate cancer, lung cancer, breast cancer, and other cancers were included in the study as true-positive cases. Patients with no skeletal metastasis (regardless of hot spots), and patients with abnormal bone scans but no skeletal metastasis were included as negative cases. Bone-scan patients were subjected to artificial neural network (ANN) evaluation. Values equal to or above 0.5 were regarded as positive, and those below 0.5 as negative. The patients whose clinical status did not correspond to their ANN scores were assessed for any similarities.ResultsThe frequency of segmentation errors was statistically significantly reduced when using BONENAVI version 2.1.7. The differences in sensitivity and specificity for the results of version 2.0.5 versus version 2.1.7 were not different, giving a high Cohen’s kappa coefficient. In the patients who showed an increased ANN value with version 2.1.7, a few false-positive thoracic lesions were identified. Patients whose ANN value was significantly high with version 2.0.5 showed no tendencies.ConclusionRevised BONENAVI version 2.1.7 for bone scintigraphy was superior with regard to segmentation errors. However, its sensitivity and specificity were similar to those of version 2.0.5. The false-positive identification of thoracic lesions in revised version 2.1.7 might be subject to remedy.

Evaluation of scatter limitation correction: a new method of correcting photopenic artifacts caused by patient motion during whole-body PET/CT imaging.

ObjectiveOvercorrection of scatter caused by patient motion during whole-body PET/computed tomography (CT) imaging can induce the appearance of photopenic artifacts in the PET images. The present study aimed to quantify the accuracy of scatter limitation correction (SLC) for eliminating photopenic artifacts. MethodsThis study analyzed photopenic artifacts in 18F-fluorodeoxyglucose (18F-FDG) PET/CT images acquired from 12 patients and from a National Electrical Manufacturers Association phantom with two peripheral plastic bottles that simulated the human body and arms, respectively. The phantom comprised a sphere (diameter, 10 or 37 mm) containing fluorine-18 solutions with target-to-background ratios of 2, 4, and 8. The plastic bottles were moved 10 cm posteriorly between CT and PET acquisitions. All PET data were reconstructed using model-based scatter correction (SC), no scatter correction (NSC), and SLC, and the presence or absence of artifacts on the PET images was visually evaluated. The SC and SLC images were also semiquantitatively evaluated using standardized uptake values (SUVs). ResultsPhotopenic artifacts were not recognizable in any NSC and SLC image from all 12 patients in the clinical study. The SUVmax of mismatched SLC PET/CT images were almost equal to those of matched SC and SLC PET/CT images. Applying NSC and SLC substantially eliminated the photopenic artifacts on SC PET images in the phantom study. SLC improved the activity concentration of the sphere for all target-to-background ratios. The highest %errors of the 10 and 37-mm spheres were 93.3 and 58.3%, respectively, for mismatched SC, and 73.2 and 22.0%, respectively, for mismatched SLC. ConclusionPhotopenic artifacts caused by SC error induced by CT and PET image misalignment were corrected using SLC, indicating that this method is useful and practical for clinical qualitative and quantitative PET/CT assessment.

Evaluation of Iterative Reconstruction Method and Attenuation Correction in Brain Dopamine Transporter SPECT Using an Anthropomorphic Striatal Phantom

Objective(s): The aim of this study was to determine the optimal reconstruction parameters for iterative reconstruction in different devices and collimators for dopamine transporter (DaT) single-photon emission computed tomography (SPECT). The results were compared between filtered back projection (FBP) and different attenuation correction (AC) methods. Methods: An anthropomorphic striatal phantom was filled with 123I solutions at different striatum-to-background radioactivity ratios. Data were acquired using two SPECT/CT devices, equipped with a low-to-medium-energy general-purpose collimator (cameras A-1 and B-1) and a low-energy high-resolution (LEHR) collimator (cameras A-2 and B-2). The SPECT images were once reconstructed by FBP using Chang’s AC and once by ordered subset expectation maximization (OSEM) using both CTAC and Chang’s AC; moreover, scatter correction was performed. OSEM on cameras A-1 and A-2 included resolution recovery (RR). The images were analyzed, using the specific binding ratio (SBR). Regions of interest for the background were placed on both frontal and occipital regions. Results: The optimal number of iterations and subsets was 10i10s on camera A-1, 10i5s on camera A-2, and 7i6s on cameras B-1 and B-2. The optimal full width at half maximum of the Gaussian filter was 2.5 times the pixel size. In the comparison between FBP and OSEM, the quality was superior on OSEM-reconstructed images, although edge artifacts were observed in cameras A-1 and A-2. The SBR recovery of OSEM was higher than that of FBP on cameras A-1 and A-2, while no significant difference was detected on cameras B-1 and B-2. Good linearity of SBR was observed in all cameras. In the comparison between Chang’s AC and CTAC, a significant correlation was observed on all cameras. The difference in the background region influenced SBR differently in Chang’s AC and CTAC on cameras A-1 and B-1. Conclusion: Iterative reconstruction improved image quality on all cameras, although edge artifacts were observed in images captured by cameras with RR. The SBR of OSEM with RR was higher than that of FBP, while the SBR of OSEM without RR was equal to that of FBP. Also, the SBR of Chang’s AC varied with different background regions in cameras A-1 and B-1.

Validation of Cross-calibration Schemes for Quantitative Bone SPECT/CT Using Different Sources under Various Geometric Conditions

PURPOSE Several cross-calibration schemes have been proposed to produce quantitative values in bone SPECT imaging. Differences in the radionuclide sources and geometric conditions can decrease the accuracy of cross-calibration factor (CCF). The present study aimed to validate the effects of calibration schemes using different sources under various geometric conditions. METHODS Temporal variations as well as variations in acquisition counts and the shapes of 57Co standard and 99mTc point sources and a 99mTc disk source were determined. The effects of the geometric conditions of the source-to-camera distance (SCD) and lateral distance on the CCF were investigated by moving the camera or source away from the origin. The system planar sensitivity of NEMA incorporated into a Symbia Intevo SPECT/CT device (Siemens®) was defined as reference values. RESULTS The temporal variation in CCF using the 57Co source was relatively stable within the range of 0.7% to 2.3%, whereas the 99mTc source ranged from 2.7% to 7.3%. In terms of source shape, the 57Co standard point source was the most stable. Both SCD and lateral distance decreased as a function of distance from the origin. Errors in the geometric condition were higher for the 57Co standard point source than the 99mTc disk source. CONCLUSIONS Different calibration schemes influenced the reliability of quantitative values. The 57Co standard point source was stable over a long period, and this helped to maintain the quality of quantitative SPECT/CT imaging data. The CCF accuracy of the 99mTc source decreased depending on the preparative method. The method of calibration for quantitative SPECT should be immediately standardized to eliminate uncertainty.

A Comparison of Planar Sensitivity and Spatial Resolution among Different Collimators and Energy Windows on 223 Ra Imaging

OBJECTIVE The present study aimed to reveal the influence of combination of different collimators and energy windows on the planar sensitivity and the spatial resolution during experimental 223Ra imaging, and to determine optimal imaging parameters. METHODS A vial type source containing 223Ra solution (4.55 MBq / 5.6 ml) was placed in the air at 100 mm away from the collimator surface. Planar images were acquired with LEHR, LMEGP, ELEGP and MEGP collimators on two dual-head gamma cameras (Symbia intevo (Siemens) and Infinia 3 (GE)). We compared three energy window combinations: 1) single window at 82 keV, 2) double window at 82+154 keV, 3) triple window at 82+154+270 keV. The energy spectrum, the sensitivity and the spatial resolution, such as full-width at half-maximum (FWHM) and full-width at tenth-maximum (FWTM), of each collimator were assessed. RESULTS Five energy spectra (at around 82, 154, 270, 351 and 405 keV) were essentially observed among four collimators. The sensitivity was high for LEHR collimator, then ELEGP and LMEGP collimator was 3-4 fold, which is greater than MEGP collimator. The 82 keV energy window of four collimators has best spatial resolution. Moreover, the spatial resolution of the 82 keV energy window with LMEGP and ELEGP collimator was almost equal to that of the triple window with MEGP collimator. CONCLUSIONS Optimal imaging parameters were single energy window using LMEGP or ELEGP, and then triple energy window using MEGP collimator.

Development of a Novel Body Phantom with Bone Equivalent Density for Evaluation of Bone SPECT

We developed a custom-designed phantom for bone single photon emission computed tomography (SPECT)-specific radioactivity distribution and linear attenuation coefficient. The aim of this study was to evaluate the accuracy of the phantom. The lumbar phantom consisted of the trunk of a body phantom (background) containing a cylinder (vertebral body), a sphere (tumor), and a T-shaped container (processus). The vertebral body, tumor, and processus phantoms contained a K(2)HPO(4) solution of bone equivalent density and 50, 300 and 50 kBq/mL of (99m)Tc, respectively. The body phantom contained 8 kBq/mL of (99m)Tc solution. SPECT images were acquired using low-energy high-resolution collimation, a 128 × 128 matrix and 120 projections over 360° with a dwell time of 15 sec/view × 4 times. Thereafter, CT images were acquired at 130 kV and 70 ref mAs using adaptive dose modulation. The SPECT data were reconstructed with ordered subset expectation maximization with three-dimensional, scatter, and CT-based attenuation correction. Count ratio, linear attenuation coefficient (LAC), and full-width at half-maximum (FWHM) were measured. Count ratios between the background, the vertebral body, and the tumor in SPECT images were 463.8: 2888.0: 15150.3 (1: 6.23: 32.7). The LAC of the background and vertebral body in the CT-derived attenuation map were 0.155 cm⁻¹ and 0.284 cm⁻¹, respectively, and the FWHM measured from the processus was 15.27 mm. The precise counts and LAC indicated that the phantom was accurate and could serve as a tool for evaluating acquisition, reconstruction parameters, and quantitation in bone SPECT images.

Application of novel calibration scheme based on traceable point-like 22 Na sources to various types of PET scanners

Purpose: We have been developing a practical and reliable calibration scheme based on the use of traceable pointlike sources. In using 22Na sources, special care should be taken to avoid the effects of 1.275-MeV γ rays accompanying β+ decays. The purpose of this study is to validate this calibration method with traceable point-like 22Na sources on various types of PET scanners. Method: The traceable point-like 22Na sources with a spherical absorber design used in this study were of the same type as those used in a previous study. The tested PET scanners included one clinical whole-body PET scanner, four types of clinical PET/CT scanners from different manufacturers, and one small-animal PET scanner. The ROI (region of interest) diameter dependence of ROI values were represented with a fitting function, which was assumed to consist of a recovery part due to spatial resolution and a quadratic background part originating from the scattered γ rays. Results: The calibration factors determined using the point-like source were consistent with those by the standard cross-calibration method within ±4%, which was comparable to the uncertainty of the standard cross-calibration method. Conclusion: The novel calibration scheme based on the use of traceable 22Na point-like sources was validated for six types of commercial PET scanners.