Deep learning enhanced ultra-fast SPECT/CT bone scan: quantitative assessment and clinical performance

Deep learning enhanced ultra-fast SPECT/CT bone scan: quantitative assessment and clinical performance | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Deep learning enhanced ultra-fast SPECT/CT bone scan: quantitative assessment and clinical performance Na Qi, Boyang Pan, Qingyuan Meng, Yihong Yang, Huiqian Chen, Weilun Wang, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-2190739/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background To evaluate clinical performance of deep learning enhanced ultra-fast SPECT/CT bone scan. Methods One hundred and two patients were enrolled in this retrospective study. The probable malignant tumor sites continuously underwent a 20min SPECT/CT and a 3min SPECT scan. A deep learning model was applied to generate algorithm-enhanced images (3min-DL SPECT). Two reviewers evaluated general image quality, 99m Tc-MDP distribution, artifacts, and diagnostic confidence independently. The sensitivity, specificity, accuracy, and inter-observer agreement were calculated. Linear regression was analyzed for lesion SUV max between 3min-DL and 20min SPECT. Peak signal-to-noise ratio (PSNR), image similarity (SSIM) were evaluated. Results The general image quality, 99m Tc-MDP distribution, artefact, and diagnostic confidence of 3min-DL images were significantly superior to those of 20min images (P < 0.0001). The sensitivity, specificity and accuracy of 20min and 3min-DL SPECT/CT had no difference by both reviewers (0.903 vs 0.806, 0.873 vs 0.873, 0.882 vs 0.853; 0.867 vs 0.806, 0.944 vs 0.936, 0.912 vs 0.920, P > 0.05). The diagnosis results of 20min and 3min-DL images showed a high inter-observer agreement (Kappa = 0.822, 0.732). PSNR and SSIM of 3min-DL images were significantly higher than 3min images (51.44 vs 38.44, 0.863 vs 0.752, P < 0.05). A strong linear relationship was found between the SUV max of 3min-DL and 20min images (r = 0.987; P < 0.0001). Conclusion An ultra-fast SPECT/CT with 1/7 scan time could be enhanced by deep learning method to have competitive image quality and equivalent diagnostic value to those of standard acquisition. Bone SPECT/CT Deep learning Ultra-fast Image quality Diagnostic efficiency Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Background Single-photon emission computed tomography (SPECT) bone scintigraphy is a frequently used imaging technology in nuclear medicine with broad diverse applications[ 1 ][ 2 ]. 99m Tc methylene diphosphonate ( 99m Tc-MDP) is a widely used radiopharmaceutical for bone scanning, Despite the high sensitivity of bone scintigraphy, the specificity is relatively poor[ 3 ]. This dilemma was improved by the introduction of hybrid SPECT/CT in 1997[ 4 ], which utilized the precise anatomical localization of registered CT[ 5 ][ 6 ][ 7 ]. However, the multimodality information obtained by SPECT/CT comes at a cost of long scanning time. In general, SPECT/CT for one patient in clinical practice takes at least 30 minutes. The patients examined are usually those with malignant tumors and are often intolerant of prolonged immobility. Body movement during the prolonged examination can result in inaccurate SPECT/CT fusion and motion artifacts[ 8 ]. Therefore, it is a pressing need to reduce the scanning time to improve patient comfort and suppress patient motion without sacrificing image quality. A short scanning time can also boost scanner throughput and improve clinical productivity. However, with the short scanning time, image quality will decrease due to the increased noise. To improve the trade-off between examination time and image quality, deep learning-based methods have been used in SPECT/CT reconstruction in recent years[ 9 ][ 10 ][ 11 ]. Yang et al. apply deep learning method to synthesize attenuation-corrected cardiac SPECT using noncorrected SPECT without undergoing additional image reconstruction process[ 12 ]. Ramon et al. simulated low dose SPECT myocardial perfusion imaging (MPI) scans by statistical subsampling of counts obtained from standard clinical dose to form a paired dataset necessary for training deep learning network[ 13 ]. They also improved simulate process by accepting or rejecting the photon projection during data collection with a given probability, and thereby obtaining 1/2, 1/4, 1/8, 1/16 of the original sampled data[ 14 ]. Deep learning-generated synthesized projections constructed by a deep convolution U-Net model were added in 177 Lu-SPECT with sparsely acquired projections, which circumvent image degradation considerably and reduce scanning time[ 15 ]. Another three-dimension residual U-Net model was used to reconstruct full-acquisition-time images from short-acquisition-time images, reducing the scan time of pediatric 99m Tc-dimercaptosuccinic acid SPECT[ 16 ]. Shiri et al. compared the performance of deep learning method to reduce scan time in SPECT MPI through two approaches, namely cutting off angular projections and reduction of acquisition time per projection[ 17 ]. Our previous work[ 18 ] showed image quality of SPECT bone scan could be significantly improved in terms of Peak Signal-to-Noise Ratio (PSNR) and Structure Similarity Index Measure (SSIM). However, the influence of diagnostic result on enhanced SPECT bone images was not assessed.In this study, we explored the performance of deep learning enhanced SPECT images on subjects with various diseases and evaluated whether the deep learning approach can meet the clinical diagnostic needs. Materials And Methods Subjects One hundred and two subjects with suspected bone metastases from March 2021 to March 2022 were enrolled in this prospective study. All patients signed informed consent before examination. This study was approved by the institutional review board of Shanghai East Hospital. The subjects with a history of renal insufficiency, hormone, endocrine therapy, chemotherapy, and other treatments affecting bone metabolism were excluded from the research. The subject information was obtained from the medical record. Spect/ct Acquisition All patients were injected with 19–22 MBq/kg 99m Tc-MDP. The whole-body scan and quantitative bone SPECT/CT imaging were performed using SPECT/CT (Siemens Symbia Intevo, Erlangen, Germany). Whole-body planar imaging was performed at about 3–4 hours post-injection. The areas with suspected malignancy were examined with a 20 minutes SPECT/CT (standard time) and followed by a 3 minutes SPECT (7 times reduction in scan time), the patient was instructed to remain still during the examination. Cases with motion artifacts were discarded after examination. The scanning matrix was 256×256, and the zoom factor was 1.0. Step-and-shoot mode with a total of 120 projections (60 steps) over 360° was used while 20s per step was adopted for standard time SPECT and 3s per step was for 1/7 standard time SPECT. Subsequently, a low-dose CT scan was performed at 130 kV and 10 mAs. CT data was reconstructed using a sharp bone kernel with 5mm slice thickness (B50s) and a smooth attenuation-correction kernel with 3mm slice thickness (B31s). SPECT reconstruction with attenuation correction was performed using the B31s CT attenuation map. The ordered subsets conjugate gradient enhanced xSPECT reconstruction algorithm (xSPECT/CT, Siemens Symbia Intevo) with 2 subsets and 28 iterations without post-smoothing was used to generate quantification measurement such as the maximum standard uptake value (SUV max ). Imaging Process A pretrained deep learning model[ 18 ] with integrated multi-scale and multi-modality features from 3min SPECT image and corresponding CT image was applied to generated enhanced SPECT images (3min-DL SPECT). The archiecture of the network was shown in Figure. 1. 3min SPECT SUV images were directly used as the network input without extra scaling. CT images were interpolated to the size of 3min SPECT and were normalized by image mean before network input. 3 consecutive SPECT and CT images were fed each time and average value was calculated if one slice was inferred multiple times. Image Evaluation 20min, 3min and 3min-DL SPECT/CT images were evaluated independently by 2 nuclear medicine physicians with 5 and 10 years’ experience respectively. 5-point Likert scale (1, unacceptable image quality; 2, suboptimal image quality; 3, acceptable image quality; 4, good image quality; 5, excellent image quality) was used to score the three groups of images to evaluate their overall image quality, 99m Tc-MDP details, presence of artifacts, and general diagnostic confidence. A score of 3 or higher indicate the requirement on image quality for clinical diagnosis was met. The lesion with the highest SUV in each subject was defined as the volume of interest (VOI), which was drawn using Siemens 3D Isocontour with SUV max automatically calculated. All analyses were performed blind to the image acquisition information. To quantitatively evaluate the performance of synthesized images, PSNR and SSIM are used as evaluation metrics. 20 min SPECT/CT were treated as the ground truth images. PSNR for synthesized image is defined as $$\text{P}\text{S}\text{N}\text{R}=10\cdot {\text{log}}_{10}\left(\frac{\text{M}\text{A}{\text{X}}_{\text{g}\text{t}}^{2}}{\text{M}\text{S}\text{E}}\right)$$ Where \(\text{M}\text{A}{\text{X}}_{\text{g}\text{t}}\) is the maximum pixel value of ground truth 20min SPECT. MSE is the mean square error of synthesized images compared to the 20min SPECT. SSIM for synthesized image is defined as $$\text{S}\text{S}\text{I}\text{M}\left(\text{x},\text{y}\right)=\frac{\left(2{{\mu }}_{\text{x}}{{\mu }}_{\text{y}}+{\text{c}}_{1}\right)\left(2{{\sigma }}_{\text{x}\text{y}}+{\text{c}}_{2}\right)}{\left({{\mu }}_{\text{x}}^{2}+{{\mu }}_{\text{y}}^{2}+{\text{c}}_{1}\right)\left({{\sigma }}_{\text{x}}^{2}+{{\sigma }}_{\text{y}}^{2}+{\text{c}}_{2}\right)}$$ Where \({{\mu }}_{\text{x}}\) and \({ {\sigma }}_{\text{x}}^{2}\) are average value and variance of input synthesized image. \({{\mu }}_{\text{y}}\) and \({{\sigma }}_{\text{y}}^{2}\) are the average value and variance of input 20min SPECT. \({{\sigma }}_{\text{x}\text{y}}\) is the covariance of the two images. \({\text{c}}_{1}\) and \({\text{c}}_{2}\) are small constants. SSIM is calculated using the scikit-image package. Diagnostic Performance 20min SPECT/CT images of all cases were read by the two reviewers. Benign and malignant lesions were determined based on pathological diagnosis, imaging examinations (20min SPECT/CT, CT, MRI), and clinical follow-up data. If the results were inconsistent, another senior physician was added to make a judgment, and the final benign and malignant lesions were determined as the gold standard. To dilute the memory effect, two readers independently read the disordered 3min and 3min-DL SPECT after one month. Benign (negative) and malignant (positive) were determined (a 5-point Likert scale of 1 was negative). Sensitivity, specificity, accuracy, and interobserver agreement were calculated. Statistical analysis Statistical analyses were performed using Graphpad Prism (8.0.0). Kappa consistency test was used to evaluate the consistency of the two reviewers in 20min, 3min, and 3min-DL SPECT images. Kappa ≥ 0.75 showed good consistency, 0.4 < Kappa < 0.75, moderate consistency, and Kappa ≤ 0.4, poor consistency. PSNR and RMSE in the quantitative analysis were compared with the paired Student’s t-test. Pearson and linear regression were used for demonstrating the consistency of the SUV max between the 3min, 3min-DL, and 20min SPECT/CT. Chi-square test or Fisher's exact test was assessed the differences in sensitivity, specificity, accuracy. P < 0.05 was considered statistically significant. Results Patient characteristics A total of 102 patients (55 males and 47 females, mean age 60 ± 12 years, age range 26–87 years, mean Body Mass Index (BMI) 23 ± 3 kg/m 2 , BMI range 16–32 kg/m 2 ), mainly diagnosed with cancer, were included in this study. Patient characteristics were summarized in Table 1 . Table 1 Patient characteristics N = 102 Overall Age Mean (SD) 60(12) [Min, Max] [26,87] Gender Male 55 Female 47 BMI Mean (SD) 23(3) [Min, Max] [15,32] Injected dose (MBq) [740,1110] Wating time (min) [168, 265] Diagnosis Lung cancer 33 Breast cancer 21 Prostate cancer 11 Gastric cancer 10 Colon cancer 5 Nasopharyngeal carcinoma 4 Liver cancer 4 Cholangiocarcinoma 4 Endometrial carcinoma 3 Others 7 Image Quantitative Analysis 3min images rated 1 point, insufficient for clinical diagnosis. 3min-DL and 20 min images displayed excellent image quality (Figure. 2, 3, 4). As shown in Figure. 5, the mean 5-point Likert scale scores by 2 reviewers of 3min-DL images were higher than those of 20min image in general image quality, Tc distribution, artifacts, and diagnostic confidence (mean ± SD 3.44 ± 0.64 vs. 3.10 ± 0.64, 3.45 ± 0.64 vs. 3.07 ± 0.62, 3.46 ± 0.63 vs. 3.10 ± 0.61, 3.78 ± 0.66 vs. 3.46 ± 0.68. P＜0.0001). The SUV max obtained from 3min-DL and 20min image have no statistical difference (P = 0.973), and a strong linear relationship were verified between SUV max of lesions in 3min-DL and 20min images (Y = 0.9881*X་0.2065, r = 0.987; P < 0.0001; Figure.6). PSNR and SSIM of 3min-DL images were significantly higher than those of 3min images (51.44 vs 38.44, 0.863 vs 0.752; P < 0.05) (Table 2 ). Table 2 PSNR and SSIM of 3min-DL, 3min images PSNR SSIM 3min 38.44 0.7522 3min-DL 51.44 0.8633 Diagnostic Performance Analysis The diagnosis results of 20min and 3min-DL showed a high consistency between two reviewers (Kappa = 0.822, 0.732). The sensitivity, specificity and accuracy of 20min and 3min-DL had no difference by both reviewers (0.903 vs 0.806, 0.873 vs 0.873, 0.882 vs 0.853; 0.867 vs 0.806, 0.944 vs 0.936, 0.912 vs 0.920, P > 0.05) (Table 3 ). Table 3 Sensitivity, specificity and accuracy of 20min and 3min-DL images by two reviewers reviewer1 reviewer2 Sensitivity Specificity Accuracy Sensitivity Specificity Accuracy 20min 0.903(28/31) 0.873(62/71) 0.882(90/102) 0.867(26/31) 0.936(67/72) 0.912(93/102) 3min-DL 0.806(25/31) 0.873(62/71) 0.853(87/102) 0.806(25/31) 0.944(67/71) 0.902(92/102) X 2 1.17 0 0.384 0.11 0.104 0.058 P value 0.279 ＞0.999 0.535 0.74 0.747 0.81 Discussion SPECT bone scintigraphy continues to be a great-volume nuclear imaging procedure, offering the advantage of total body examination with high sensitivity but poor specificity[ 19 ]. The use of additional SPECT/CT for evaluating suspicious or equivocal lesions has been shown to improve diagnostic confidence and specificity [ 5 ][ 7 ]. But the extra multimodal information obtained with SPECT/CT leads to an increased scanning time. Reduction of acquisition time and radioactive dose in SPECT/CT have become a major focus of attention. The short acquisition time may be considered more comfortable to patients especially children, obese people and painful patients those are intolerant of prolonged immobility during the acquisition period. Reduction of radioactive dose can reduce radiation exposure of children and patients who require multiple follow-up during examinations. Our previous study has generated high-quality bone scan SPECT images from 1/7 scan time SPECT images using deep learning method with a small sample and confirmed that this method yielded significant image quality improvement in the noise level, details of anatomical structure and SUV accuracy[ 18 ]. The previous deep learning based SPECT enhancement studies always build on simulated images from randomly under-sampled list mode data[ 12 ][ 13 ][ 14 ]. To imitate the real clinical fast scan environment, this work adopted continuous examination and use the same reconstruction method. The fast scan of 102 patients with different types of cancer were enhanced by the pretrained deep learning model. The SPECT/CT image quality in this study was assessed by two independent nuclear radiologists with 5 and 10 years experience in SPECT/CT diagnosis. The reading results of both radiologists showed that the image qualities of the 3min SPECT were not good enough to have diagnostic value with 1 point in 5-point Likert scale, and not included in the statistical evaluation. Though the standard acquisition always received the highest score in deep learning based PET image enhancement[ 20 ], the general image quality, detail of 99m Tc-MDP, presence of artifacts, and general diagnostic confidence of 3min-DL images were significantly superior to those of 20min images (P＜0.0001) in this SPECT bone scan study. A reduction of noise level of soft tissue and more uniform and coherent distribution of normal bone in 3min-DL SPECT images were observed compared with 20min SPECT as shown in Figure. 2, 3, 4. This could be a result of high noise level of SPECT system and deep learning method trend to generate expected pixel values[ 21 ]. The diagnosis results of both 20min and 3min-DL SPECT/CT showed substantial agreement in the 2 reviewers’ image evaluation (Kappa = 0.822, 0.732). A SPECT study showed that shorting acquisition SPECT scan time to 1/4 standard time resulted in an excellent inter-observer agreement without affecting the diagnosis[ 22 ]. Inconsistent with their findings, when the scanning time was reduced to 3 minutes, the image quality was greatly degraded and had a grainy appearance in our study. After deep-learning technology application, the algorithm-enhanced 3 min images showed excellent image quality and could meet the needs of clinical diagnosis in our study . 99m Tc-MDP chemisorbs and binds to the hydroxyapatite crystals and is marker of bone turnover and bone perfusion. It rapidly localizes to bone and clears quickly from background, making it favorable for imaging[ 23 ]. Even a 5% change in bone turnovercan be detected on bone imaging, bone imaging can often detect active bone formation in the skeleton related to malignant and benign disease [ 24 ]. For some diseases, such as infection, trauma, sclerosis or other origin, focus of increased radioactive tracer uptake could cause false-positive results[ 23 ]. Therefore, diagnosis based on SPECT/CT imaging relies to some extent on CT imaging and the readers’ experience. The sensitivity, specificity and accuracy of 3min-DL images have no significant difference from 20min images of both reviewers in our study (0.903 vs 0.806, 0.873 vs 0.873, 0.882 vs 0.853; 0.867 vs 0.806, 0.944 vs 0.936, 0.912 vs 0.920, P > 0.05). These results were similar to other diagnostic performance of traditional SPECT/CT bone scans studies[ 25 ][ 26 ]. This study showed that 3min-DL SPECT/CT images not only sped up the scanning time but also reached the diagnostic level of traditional SPECT/CT. Fast scanning would be particularly beneficial to the patients who are difficult to maintain a prolonged horizontal position for SPECT/CT imaging due to pain or other reasons. Our previous study has shown that SUV max obtained from SPECT/CT could play an important role in differentiating benign and malignant bone diseases[ 27 ]. In this study, the SUV max obtained from 3min-DL and 20min SPECT images has no statistical difference (P = 0.973), there was a strong linear relationship between the SUV max of lesions in 3min-DL and 20min images (r = 0.987; P < 0.0001). The result indicated that quantitative evaluation SUV max of 3min ultra-high-speed SPECT/CT bone scan enhanced by deep learning algorithm can also be used in clinical. Despite this research explored the clinical performance of deep learning method in SPECT bone scan, it has several limitations. We addressed most tumor types in this study, but few positive cases for bone metastasis were included. Sample capacity needs to be expanded in the future study. Pathological verification of some bone lesions was challenging, the gold standard based on the clinical history and other imaging studies might cause bias in the results. Our experiment bases on SPECT/CT bone images, more types of SPECT/CT studies such as SPECT MPI should be explored. Though dose reduction and fast scan has the same physical essence, minimize injection dosage while maintaining diagnostic accuracy should be further evaluated. Conclusion This study evaluated the image quality and diagnostic efficiency of ultra-fast SPECT/CT bone imaging based on deep learning approach with a large population. The results indicated that the image quality and diagnostic efficiency were comparable to those of standard time SPECT/CT and could meet the needs of clinical diagnosis. The reduction of acquisition time and radioactive dose in SPECT/CT will be potentially available for routine use with the help of deep learning approach. Abbreviations SPECT Single-photon Emission Computed Tomography 99mTc-MDP Technetium 99m-Methyl Diphosphonate MPI Myocardial Perfusion Imaging PSNR Peak Signal-to-Noise Ratio SSIM Structural Similarity Index SUV Standard Uptake Value SUV max Maximum Standardized Uptake Value VOI Volume of Interest Declarations 1. Ethics approval and consent to participate All methods were performed in accordance with the ethical standards as laid down in the Declaration of Helsinki and its later amendments or comparable ethical standards. Institutional Review Board approval from Ethical Review of Medical Ethics Committee of Shanghai East Hospital EC. D（BG）. 009. 02.1 was obtained. Written informed consent was obtained from all participants. 2. Consent for publication Not applicable. 3. Availability of data and material The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. 4. Competing interests Author NG is stockholder of RadioDynamic Healthcare. 5.Funding The study was partially supported by the Key Specialty Construction Project of Pudong Health and Family Planning Commission of Shanghai, PWZzk2017-24. 6.Authors' contributions NQ, BP, QM participated in its design and coordination, draft the manuscript; NQ, BP conducted statistical processing on the data; NQ, YY reviewed the images; JZ, NG determined the ideas of the article. QM, HC , and WW contributed to data collection; HL, TF, XJ, NG, and JZ provided critical review and substantially revised the manuscript. All authors read and approved the final manuscript. 7. Acknowledgements Not applicable References Ghanem MA, Dannoon S, Elgazzar AH. The added value of SPECT-CT in the detection of heterotopic ossification on bone scintigraphy. Skeletal Radiol. 49: 291–298. Brenner DO MMM CPE AI, Koshy MDJ, Morey MDJ, Lin MDC, DiPoce. MD J The Bone Scan Semin Nucl Med. 2012;42:11–26. Gnanasegaran MDG, Barwick MRCPFRCRT, Adamson MSc K, Mohan MRCPH, Sharp HNCD, Fogelman MDI. Multislice SPECT/CT in Benign and Malignant Bone Disease: When the Ordinary Turns Into the Extraordinary. Semin Nucl Med. 2009;39:431–42. Kalki K, Blankespoor SC, Brown JK, Hasegawa BH, Dae MW, Chin M, et al. Myocardial Perfusion Imaging with a Combined X-Ray CT and SPECT System. J Nucl Med. 1997;38:1535–40. UTSUNOMIYA D, SHIRAISHI S, IMUTA M, TOMIGUCHI S, KAWANAKA K, MORISHITA S, et al. Added value of SPECT/CT fusion in assessing suspected bone metastasis: Comparison with scintigraphy alone and nonfused scintigraphy and CT. Radiology. 2006;238:264–71. Mariani G, Bruselli L, Kuwert T, Kim EE, Flotats A, Israel O, et al. A review on the clinical uses of SPECT/CT. Eur J Nucl Med Mol Imaging. 2010;37:1959–85. Palmedo H, Marx C, Ebert A, Kreft B, Ko Y, Türler A, et al. Whole-body SPECT/CT for bone scintigraphy: diagnostic value and effect on patient management in oncological patients. Eur J Nucl Med Mol Imaging. 2014;41(1):59–67. Picone V, Makris N, Boutevin F, Roy S, Playe M, Soussan M. Clinical validation of time reduction strategy in continuous step-and-shoot mode during SPECT acquisition. EJNMMI Phys. 2021;8(1):10. Shao W, Rowe SP, Du Y. Artificial intelligence in single photon emission computed tomography (SPECT) imaging: a narrative review. Ann Transl Med. 2021;9(9):820. Cheng Z, Wen J, Huang G, Yan J. Applications of artificial intelligence in nuclear medicine image generation. Quant Imaging Med Surg. 2021;11(6):2792–822. Zhang Q, Sun J, Mok GSP. Low dose SPECT image denoising using a generative adversarial network. 2019. Yang J, Shi L, Wang R, Miller EJ, Sinusas AJ, Liu C, et al. Direct attenuation correction using deep learning for cardiac SPECT: A feasibility study. J Nucl Med. 2021;62. Ramon AJ, Yang Y, Pretorius PH, Johnson KL, King MA, Wernick MN. Initial Investigation of Low-Dose SPECT-MPI via Deep Learning. 2018 IEEE Nucl Sci Symp Med Imaging Conf NSS/MIC 2018 - Proc. 2018; 1–3. Ramon AJ, Yang Y, Pretorius PH, Johnson KL, King MA, Wernick MN. Improving Diagnostic Accuracy in Low-Dose SPECT Myocardial Perfusion Imaging with Convolutional Denoising Networks. IEEE Trans Med Imaging. 2020;39:2893–903. Rydén T, Van Essen M, Marin I, Svensson J, Bernhardt P, universitet G, et al. Deep-Learning Generation of Synthetic Intermediate Projections Improves 177 Lu SPECT Images Reconstructed with Sparsely Acquired Projections. J Nucl Med. 62: 528–535. Lin C, Chang Y-C, Chiu H-Y, Cheng C-H, Huang H-M. Reducing scan time of paediatric 99m Tc-DMSA SPECT via deep learning. Clin Radiol. 2021;76:315.e13. Shiri I, Sabet KA, Arabi H, Pourkeshavarz M, Teimourian B, Ay MR, et al. Standard SPECT myocardial perfusion estimation from half-time acquisitions using deep convolutional residual neural networks. J Nucl Cardiol. 2021;28(6):2761–79. Pan B, Qi N, Meng Q, Wang J, Peng S, Qi C, et al. Ultra high speed SPECT bone imaging enabled by a deep learning enhancement method: a proof of concept. EJNMMI Phys. 2022;9(1):43. Brenner AI, Koshy J, Morey J, Lin C, DiPoce J. The bone scan. Semin Nucl Med. 2012;42(1):11–26. Ly J, Minarik D, Jögi J, Wollmer P, Trägårdh E. Post-reconstruction enhancement of [18F]FDG PET images with a convolutional neural network. EJNMMI Res. 2021;11(1):48. Lehtinen J, Munkberg J, Hasselgren J, Laine S, Karras T, Aittala M, et al. Noise2Noise: Learning Image Restoration without Clean Data. 2018. https://go.exlibris.link/P27f6SpV . Zacho HD, Manresa JAB, Aleksyniene R, Ejlersen JA, Fledelius J, Bertelsen H, et al. Three-minute SPECT/CT is sufficient for the assessment of bone metastasis as add-on to planar bone scintigraphy: prospective head-to-head comparison to 11-min SPECT/CT. EJNMMI Res. 2017;7:1. Van den Wyngaert T, Strobel K, Kampen WU, Kuwert T, van der Bruggen W, Mohan HK, et al. The EANM practice guidelines for bone scintigraphy. Eur J Nucl Med Mol Imaging. 2016;43:1723–38. Vijayanathan MRCPFRCRS, Butt MBBSFRCRS, Gnanasegaran MDG, Groves MDAM. Advantages and Limitations of Imaging the Musculoskeletal System by Conventional Radiological, Radionuclide, and Hybrid Modalities. Semin Nucl Med. 2009;39:357–68. Zhang Y, Li B, Wu B, Yu H, Song J, Xiu Y, et al. Diagnostic performance of whole-body bone scintigraphy in combination with SPECT/CT for detection of bone metastases. Ann Nucl Med. 2020;34:549–58. Mostafa R, Abdelhafez YG, Abougabal M, Nardo L, Elkareem MA. Two-bed SPECT/CT versus planar bone scintigraphy: prospective comparison of reproducibility and diagnostic performance. Nucl Med Commun. 2021;42:360–8. Qi N, Meng Q, You Z, Chen H, Shou Y, Zhao J. Standardized uptake values of 99m Tc-MDP in normal vertebrae assessed using quantitative SPECT/CT for differentiation diagnosis of benign and malignant bone lesions. BMC Med Imaging. 2021;21:39. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-2190739","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":147642648,"identity":"66b4562e-fd35-4279-9ce0-1098f58aff03","order_by":0,"name":"Na Qi","email":"","orcid":"","institution":"Department of Nuclear Medicine, Shanghai East Hospital, Tongji University School of Medicine","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Na","middleName":"","lastName":"Qi","suffix":""},{"id":147642649,"identity":"fb48e519-54eb-40e4-bdfd-a1ff3138013e","order_by":1,"name":"Boyang Pan","email":"","orcid":"","institution":"RadynDynamic Healthcare","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Boyang","middleName":"","lastName":"Pan","suffix":""},{"id":147642650,"identity":"395a0078-e952-4b83-9ca1-679ce15f21b6","order_by":2,"name":"Qingyuan Meng","email":"","orcid":"","institution":"Department of Nuclear medicine, Shanghai East Hospital, Tongji University School of Medicine","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Qingyuan","middleName":"","lastName":"Meng","suffix":""},{"id":147642651,"identity":"26200315-66c5-4e89-a56f-e8ec622eb1c5","order_by":3,"name":"Yihong Yang","email":"","orcid":"","institution":"Department of Nuclear Medicine, Shanghai East Hospital, Tongji University School of Medicine","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Yihong","middleName":"","lastName":"Yang","suffix":""},{"id":147642652,"identity":"b52e3e1a-0628-4b81-8155-a8283f682a6b","order_by":4,"name":"Huiqian Chen","email":"","orcid":"","institution":"Department of Nuclear Medicine, Shanghai East Hospital, Tongji University School of Medicine","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Huiqian","middleName":"","lastName":"Chen","suffix":""},{"id":147642653,"identity":"010682b0-3574-4049-af4c-cae31f2e4003","order_by":5,"name":"Weilun Wang","email":"","orcid":"","institution":"Department of Nuclear Medicine, Shanghai East Hospital, Tongji University School of Medicine","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Weilun","middleName":"","lastName":"Wang","suffix":""},{"id":147642654,"identity":"d0013ccb-babb-4fa5-b91b-a775c648b59e","order_by":6,"name":"Tao Feng","email":"","orcid":"","institution":"Intelligent Medical Imaging Laboratory, Cross-Strait Tsinghua Research Institute","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Tao","middleName":"","lastName":"Feng","suffix":""},{"id":147642655,"identity":"b6e1d157-c69b-41ed-b5d5-67f8c0c41339","order_by":7,"name":"Hui Liu","email":"","orcid":"","institution":"Department of Engineering Physics, Tsinghua University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Hui","middleName":"","lastName":"Liu","suffix":""},{"id":147642656,"identity":"bd95ead9-a218-48bd-ae6a-21f68982707e","order_by":8,"name":"Nan-Jie Gong","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA60lEQVRIiWNgGAWjYJACgwQIzfgASPDwkaKF2QCkhY0U29gkwCQhZebsZw8UPKhgsOuXbr9W+TXHToaNgfnhoxt4tFj25CUYJJxhSJ4550zZbdltyUCHsRkb5+DRYnAgx8AgsY0h2eBGTtptyW3MQC08bNJ4tZx/A9TyD6KlWHJbPRFaboBsaWCwM7iRfozx47bDxGgB2pJwTCJBckYOszTjtuM8bMyE/HI+x8zwR42NPb9E+sOPP7dV2/OzNz98jE8LELABY1AC6DYeA2YeEJ8Zv3KwkgdAwp6Bgf0B4w/CqkfBKBgFo2AEAgBwq0Usgh2GLwAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0001-9249-413X","institution":"Intelligent Medical Imaging Laboratory, Cross-Strait Tsinghua Research Institute","correspondingAuthor":true,"submittingAuthor":false,"prefix":"","firstName":"Nan-Jie","middleName":"","lastName":"Gong","suffix":""},{"id":147642657,"identity":"6a6dfe05-3fc7-4904-9b80-d52b79e191d0","order_by":9,"name":"Jun Zhao","email":"","orcid":"https://orcid.org/0000-0002-9887-5512","institution":"Department of Nuclear Medicine, Shanghai East Hospital, Tongji University School of Medicine","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Jun","middleName":"","lastName":"Zhao","suffix":""}],"badges":[],"createdAt":"2022-10-21 13:26:33","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-2190739/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-2190739/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":28478594,"identity":"8422bd98-aebf-40b1-b569-a63b6a2d920a","added_by":"auto","created_at":"2022-10-31 21:33:29","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":59260,"visible":true,"origin":"","legend":"\u003cp\u003eSchematic of proposed network architecture with 3min SPECT/CT as input, 20min SPECT as output. The main architecture is a U-Net like Encoder-Decoder, where each stage is replaced by N-layer RSU (RSU-N in the figure, N=3, 4, 5, 6; N denote the number of down-sample layers) and small residual block (RS, no down-sample layer). The output of RSU in the decoder is further up-sampled to compare with the target image in the training process, but only the output of RSU-6 is used in the testing process. Conv: 2D convolution, ReLU: rectified linear unit.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-2190739/v1/15e95e656b05c849c7b7da2e.png"},{"id":28478423,"identity":"5d8451d7-0568-4a62-bece-ba6141fdb31f","added_by":"auto","created_at":"2022-10-31 21:28:29","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":284607,"visible":true,"origin":"","legend":"\u003cp\u003eAn 81-year-old man with lung cancer after chemotherapy. \u003cstrong\u003ea, b\u003c/strong\u003e 20min SPECT and SPECT/CT lumbar sagittal images, respectively. Showed the concentration of radiation distribution in the osteophyte of the 5th lumbar vertebra (SUV\u003csub\u003emax\u003c/sub\u003e=11.24). \u003cstrong\u003ec\u003c/strong\u003e the MIP showed the concentration of radiation distribution at the edge of the lumbar spine. From the two reviewers, the general image quality, detail of 99mTc-MDP, presence of artifacts, and general diagnostic confidence are scored 4,3,4,4; 4,3,3,4. \u003cstrong\u003ed, e\u003c/strong\u003e 3min-DL SPECT and SPECT/CT lumbar sagittal images, showing radioactive distribution and concentration in the osteophyte of the 5th lumbar vertebra (SUV\u003csub\u003emax\u003c/sub\u003e=11.03). \u003cstrong\u003ef\u003c/strong\u003e MIP, showed multiple lumbar marginal radioactive distribution and concentration. The general image quality, detail of 99mTc-MDP, presence of artifacts, and general diagnostic confidence are scored 4 points by both reviewers. \u003cstrong\u003eg,h\u003c/strong\u003e 3min SPECT and SPECT/CT lumbar sagittal images. \u003cstrong\u003ei\u003c/strong\u003e MIP. From the two reviewers, the general image quality, detail of 99mTc-MDP, presence of artifacts, and general diagnostic confidence are scored 1 point. \u003cstrong\u003ej-l\u003c/strong\u003e transverse, sagittal and coronal CT images showed hyperosteogeny at the edge of the 5th lumbar vertebrae(arrow). The overall radioactive concentration contrast is slightly lower than that of the 20min image, the Tc distribution is more uniform, and the left iliac crest edge of f (arrow) is smoother than that of c (arrow).\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-2190739/v1/23649d8daf75b5a890a29d6e.png"},{"id":28478424,"identity":"31cce77e-9628-4a03-8467-80bdd1138ae8","added_by":"auto","created_at":"2022-10-31 21:28:29","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":286511,"visible":true,"origin":"","legend":"\u003cp\u003eA 40-year-old woman after surgery for left breast cancer. \u003cstrong\u003ea, b\u003c/strong\u003e 20min SPECT and SPECT/CT lumbar sagittal images, showed abnormal concentration of radiation distribution in the second lumbar vertebrae (SUV\u003csub\u003emax\u003c/sub\u003e=23.98). \u003cstrong\u003ec\u003c/strong\u003e MIP, displayed abnormal concentration of radiation distribution in multiple thoracolumbar vertebrae and iliac bone on both sides. The two reviewers scored 4 point for general image quality, detail of 99mTc-MDP, presence of artifacts, and 5 point for general diagnostic confidence. \u003cstrong\u003ed, e\u003c/strong\u003e 3 min-DL SPECT and SPECT/CT lumbar sagittal images, respectively. Showed abnormal concentration of radiation distribution in the second lumbar vertebrae (SUV\u003csub\u003emax\u003c/sub\u003e) 24.04. \u003cstrong\u003ef\u003c/strong\u003e MIP, showed abnormal concentration of radiation distribution in multiple thoracolumbar vertebrae and iliac crest on both sides. This group was\u0026nbsp; scored consistent with the 20min image from two reviewers. \u003cstrong\u003eg, h\u003c/strong\u003e 3min SPECT and SPECT/CT lumbar sagittal images, showed abnormally concentrated radiation distribution in the second lumbar vertebrae (SUV\u003csub\u003emax\u003c/sub\u003e = 22.12). \u003cstrong\u003ei\u003c/strong\u003e MIP. From the two reviewers,\u0026nbsp; the general image quality, detail of 99mTc-MDP, presence of artifacts, and general diagnostic confidence scored 1,1,1,3; 1,1,1,2. \u003cstrong\u003ej-l \u003c/strong\u003etransverse, sagittal and coronal CT images showed osteogenic bone destruction of the second lumbar vertebrae(circle).\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-2190739/v1/6b4e79a93e7ab00ea0dad267.png"},{"id":28478425,"identity":"1adb995f-47f3-40ce-9a87-7be4127b46af","added_by":"auto","created_at":"2022-10-31 21:28:29","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":265115,"visible":true,"origin":"","legend":"\u003cp\u003eA 58-year-old man after lung cancer surgery. \u003cstrong\u003ea, b\u003c/strong\u003e 20min SPECT and SPECT/CT sagittal images, showed the concentration of radiation distribution in the left second rib (SUV\u003csub\u003emax\u003c/sub\u003e=16.86). \u003cstrong\u003ec\u003c/strong\u003e MIP,\u0026nbsp; showed the concentration of radiation distribution in the left second rib. From the two reviewers, the general image quality, detail of 99mTc-MDP, presence of artifacts, and general diagnostic confidence are scored 4,3,3,4;3,3,3,4, respectively. \u003cstrong\u003ed, e\u003c/strong\u003e 3min-DL SPECT and SPECT/CT sagittal images, showed the concentration of radiation distribution in the left second rib (SUV\u003csub\u003emax\u003c/sub\u003e=16.48). \u003cstrong\u003ef\u003c/strong\u003e MIP, showed the concentration of radiation distribution in the left second rib. The general image quality, detail of 99mTc-MDP, presence of artifacts, and general diagnostic confidence are scored 4, 4, 4, 5; 5, 4, 4, 4.\u003cstrong\u003e g, h\u003c/strong\u003e 3min SPECT and SPECT/CT lumbar sagittal images. \u003cstrong\u003eI\u003c/strong\u003e MIP. The general image quality, detail of 99mTc-MDP, presence of artifacts, and general diagnostic confidence were scored 1 point by both reviewers. \u003cstrong\u003ej-l\u003c/strong\u003e transverse, sagittal and coronal CT images showed osteogenic bone destruction of the second rib on the left (arrow). The radiation distribution of the soft tissue in the left shoulder (f, arrow) was more lightly concentrated. d (arrow) had a smoother distribution of radioactivity at the bone edge than a (arrow).\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-2190739/v1/47093ca697095f5ce2af787b.png"},{"id":28478421,"identity":"0778911f-7e16-449c-a4d9-af441d76aa97","added_by":"auto","created_at":"2022-10-31 21:28:29","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":71724,"visible":true,"origin":"","legend":"\u003cp\u003eBox plot of 5-point Likert scale scores of 20min and 3min-DL images. The general image quality, detail of 99mTc-MDP, presence of artifacts, and general diagnostic confidence are significantly superior to those of 20min images (P \u0026lt; 0.0001).\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-2190739/v1/1e7c806b26ee84a3f6b77851.png"},{"id":28478595,"identity":"6c613b0b-b217-4831-9521-d83bde5a4474","added_by":"auto","created_at":"2022-10-31 21:33:29","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":43967,"visible":true,"origin":"","legend":"\u003cp\u003eStrong linear relationship between the SUV\u003csub\u003emax\u003c/sub\u003e of 40 lesions in 3min-DL and 20min images\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-2190739/v1/5586846bddbba3e06f799046.png"},{"id":32196666,"identity":"88c73804-006f-4413-9403-55a25c98c728","added_by":"auto","created_at":"2023-01-30 09:13:15","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1305878,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-2190739/v1/201389ce-e4f3-41b4-afc1-2eefca81de25.pdf"}],"financialInterests":"","formattedTitle":"Deep learning enhanced ultra-fast SPECT/CT bone scan: quantitative assessment and clinical performance","fulltext":[{"header":"Background","content":"\u003cp\u003eSingle-photon emission computed tomography (SPECT) bone scintigraphy is a frequently used imaging technology in nuclear medicine with broad diverse applications[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e][\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. \u003csup\u003e99m\u003c/sup\u003eTc methylene diphosphonate (\u003csup\u003e99m\u003c/sup\u003eTc-MDP) is a widely used radiopharmaceutical for bone scanning, Despite the high sensitivity of bone scintigraphy, the specificity is relatively poor[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. This dilemma was improved by the introduction of hybrid SPECT/CT in 1997[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e], which utilized the precise anatomical localization of registered CT[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e][\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e][\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. However, the multimodality information obtained by SPECT/CT comes at a cost of long scanning time. In general, SPECT/CT for one patient in clinical practice takes at least 30 minutes. The patients examined are usually those with malignant tumors and are often intolerant of prolonged immobility. Body movement during the prolonged examination can result in inaccurate SPECT/CT fusion and motion artifacts[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Therefore, it is a pressing need to reduce the scanning time to improve patient comfort and suppress patient motion without sacrificing image quality. A short scanning time can also boost scanner throughput and improve clinical productivity. However, with the short scanning time, image quality will decrease due to the increased noise.\u003c/p\u003e \u003cp\u003eTo improve the trade-off between examination time and image quality, deep learning-based methods have been used in SPECT/CT reconstruction in recent years[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e][\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e][\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Yang et al. apply deep learning method to synthesize attenuation-corrected cardiac SPECT using noncorrected SPECT without undergoing additional image reconstruction process[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Ramon et al. simulated low dose SPECT myocardial perfusion imaging (MPI) scans by statistical subsampling of counts obtained from standard clinical dose to form a paired dataset necessary for training deep learning network[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. They also improved simulate process by accepting or rejecting the photon projection during data collection with a given probability, and thereby obtaining 1/2, 1/4, 1/8, 1/16 of the original sampled data[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Deep learning-generated synthesized projections constructed by a deep convolution U-Net model were added in \u003csup\u003e177\u003c/sup\u003eLu-SPECT with sparsely acquired projections, which circumvent image degradation considerably and reduce scanning time[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Another three-dimension residual U-Net model was used to reconstruct full-acquisition-time images from short-acquisition-time images, reducing the scan time of pediatric \u003csup\u003e99m\u003c/sup\u003eTc-dimercaptosuccinic acid SPECT[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. Shiri et al. compared the performance of deep learning method to reduce scan time in SPECT MPI through two approaches, namely cutting off angular projections and reduction of acquisition time per projection[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. Our previous work[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e] showed image quality of SPECT bone scan could be significantly improved in terms of Peak Signal-to-Noise Ratio (PSNR) and Structure Similarity Index Measure (SSIM). However, the influence of diagnostic result on enhanced SPECT bone images was not assessed.In this study, we explored the performance of deep learning enhanced SPECT images on subjects with various diseases and evaluated whether the deep learning approach can meet the clinical diagnostic needs.\u003c/p\u003e"},{"header":"Materials And Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eSubjects\u003c/h2\u003e \u003cp\u003eOne hundred and two subjects with suspected bone metastases from March 2021 to March 2022 were enrolled in this prospective study. All patients signed informed consent before examination. This study was approved by the institutional review board of Shanghai East Hospital. The subjects with a history of renal insufficiency, hormone, endocrine therapy, chemotherapy, and other treatments affecting bone metabolism were excluded from the research. The subject information was obtained from the medical record.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eSpect/ct Acquisition\u003c/h3\u003e\n\u003cp\u003eAll patients were injected with 19\u0026ndash;22 MBq/kg \u003csup\u003e99m\u003c/sup\u003eTc-MDP. The whole-body scan and quantitative bone SPECT/CT imaging were performed using SPECT/CT (Siemens Symbia Intevo, Erlangen, Germany). Whole-body planar imaging was performed at about 3\u0026ndash;4 hours post-injection. The areas with suspected malignancy were examined with a 20 minutes SPECT/CT (standard time) and followed by a 3 minutes SPECT (7 times reduction in scan time), the patient was instructed to remain still during the examination. Cases with motion artifacts were discarded after examination. The scanning matrix was 256\u0026times;256, and the zoom factor was 1.0. Step-and-shoot mode with a total of 120 projections (60 steps) over 360\u0026deg; was used while 20s per step was adopted for standard time SPECT and 3s per step was for 1/7 standard time SPECT. Subsequently, a low-dose CT scan was performed at 130 kV and 10 mAs. CT data was reconstructed using a sharp bone kernel with 5mm slice thickness (B50s) and a smooth attenuation-correction kernel with 3mm slice thickness (B31s). SPECT reconstruction with attenuation correction was performed using the B31s CT attenuation map. The ordered subsets conjugate gradient enhanced xSPECT reconstruction algorithm (xSPECT/CT, Siemens Symbia Intevo) with 2 subsets and 28 iterations without post-smoothing was used to generate quantification measurement such as the maximum standard uptake value (SUV\u003csub\u003emax\u003c/sub\u003e).\u003c/p\u003e\n\u003ch3\u003eImaging Process\u003c/h3\u003e\n\u003cp\u003eA pretrained deep learning model[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e] with integrated multi-scale and multi-modality features from 3min SPECT image and corresponding CT image was applied to generated enhanced SPECT images (3min-DL SPECT). The archiecture of the network was shown in Figure. 1. 3min SPECT SUV images were directly used as the network input without extra scaling. CT images were interpolated to the size of 3min SPECT and were normalized by image mean before network input. 3 consecutive SPECT and CT images were fed each time and average value was calculated if one slice was inferred multiple times.\u003c/p\u003e\n\u003ch3\u003eImage Evaluation\u003c/h3\u003e\n\u003cp\u003e20min, 3min and 3min-DL SPECT/CT images were evaluated independently by 2 nuclear medicine physicians with 5 and 10 years\u0026rsquo; experience respectively. 5-point Likert scale (1, unacceptable image quality; 2, suboptimal image quality; 3, acceptable image quality; 4, good image quality; 5, excellent image quality) was used to score the three groups of images to evaluate their overall image quality, \u003csup\u003e99m\u003c/sup\u003eTc-MDP details, presence of artifacts, and general diagnostic confidence. A score of 3 or higher indicate the requirement on image quality for clinical diagnosis was met. The lesion with the highest SUV in each subject was defined as the volume of interest (VOI), which was drawn using Siemens 3D Isocontour with SUV\u003csub\u003emax\u003c/sub\u003e automatically calculated. All analyses were performed blind to the image acquisition information.\u003c/p\u003e \u003cp\u003eTo quantitatively evaluate the performance of synthesized images, PSNR and SSIM are used as evaluation metrics. 20 min SPECT/CT were treated as the ground truth images. PSNR for synthesized image is defined as\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\text{P}\\text{S}\\text{N}\\text{R}=10\\cdot {\\text{log}}_{10}\\left(\\frac{\\text{M}\\text{A}{\\text{X}}_{\\text{g}\\text{t}}^{2}}{\\text{M}\\text{S}\\text{E}}\\right)$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eWhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\text{M}\\text{A}{\\text{X}}_{\\text{g}\\text{t}}\\)\u003c/span\u003e\u003c/span\u003e is the maximum pixel value of ground truth 20min SPECT. MSE is the mean square error of synthesized images compared to the 20min SPECT.\u003c/p\u003e \u003cp\u003eSSIM for synthesized image is defined as\u003cdiv id=\"Equb\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e\n$$\\text{S}\\text{S}\\text{I}\\text{M}\\left(\\text{x},\\text{y}\\right)=\\frac{\\left(2{{\\mu }}_{\\text{x}}{{\\mu }}_{\\text{y}}+{\\text{c}}_{1}\\right)\\left(2{{\\sigma }}_{\\text{x}\\text{y}}+{\\text{c}}_{2}\\right)}{\\left({{\\mu }}_{\\text{x}}^{2}+{{\\mu }}_{\\text{y}}^{2}+{\\text{c}}_{1}\\right)\\left({{\\sigma }}_{\\text{x}}^{2}+{{\\sigma }}_{\\text{y}}^{2}+{\\text{c}}_{2}\\right)}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eWhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({{\\mu }}_{\\text{x}}\\)\u003c/span\u003e\u003c/span\u003e and\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({ {\\sigma }}_{\\text{x}}^{2}\\)\u003c/span\u003e\u003c/span\u003e are average value and variance of input synthesized image. \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({{\\mu }}_{\\text{y}}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({{\\sigma }}_{\\text{y}}^{2}\\)\u003c/span\u003e\u003c/span\u003e are the average value and variance of input 20min SPECT. \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({{\\sigma }}_{\\text{x}\\text{y}}\\)\u003c/span\u003e\u003c/span\u003e is the covariance of the two images. \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({\\text{c}}_{1}\\)\u003c/span\u003e\u003c/span\u003eand \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({\\text{c}}_{2}\\)\u003c/span\u003e\u003c/span\u003e are small constants. SSIM is calculated using the scikit-image package.\u003c/p\u003e\n\u003ch3\u003eDiagnostic Performance\u003c/h3\u003e\n\u003cp\u003e20min SPECT/CT images of all cases were read by the two reviewers. Benign and malignant lesions were determined based on pathological diagnosis, imaging examinations (20min SPECT/CT, CT, MRI), and clinical follow-up data. If the results were inconsistent, another senior physician was added to make a judgment, and the final benign and malignant lesions were determined as the gold standard. To dilute the memory effect, two readers independently read the disordered 3min and 3min-DL SPECT after one month. Benign (negative) and malignant (positive) were determined (a 5-point Likert scale of 1 was negative). Sensitivity, specificity, accuracy, and interobserver agreement were calculated.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eStatistical analyses were performed using Graphpad Prism (8.0.0). Kappa consistency test was used to evaluate the consistency of the two reviewers in 20min, 3min, and 3min-DL SPECT images. Kappa\u0026thinsp;\u0026ge;\u0026thinsp;0.75 showed good consistency, 0.4\u0026thinsp;\u0026lt;\u0026thinsp;Kappa\u0026thinsp;\u0026lt;\u0026thinsp;0.75, moderate consistency, and Kappa\u0026thinsp;\u0026le;\u0026thinsp;0.4, poor consistency. PSNR and RMSE in the quantitative analysis were compared with the paired Student\u0026rsquo;s t-test. Pearson and linear regression were used for demonstrating the consistency of the SUV\u003csub\u003emax\u003c/sub\u003e between the 3min, 3min-DL, and 20min SPECT/CT. Chi-square test or Fisher's exact test was assessed the differences in sensitivity, specificity, accuracy. P\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003ePatient characteristics\u003c/h2\u003e \u003cp\u003eA total of 102 patients (55 males and 47 females, mean age 60\u0026thinsp;\u0026plusmn;\u0026thinsp;12 years, age range 26\u0026ndash;87 years, mean Body Mass Index (BMI) 23\u0026thinsp;\u0026plusmn;\u0026thinsp;3 kg/m\u003csup\u003e2\u003c/sup\u003e, BMI range 16\u0026ndash;32 kg/m\u003csup\u003e2\u003c/sup\u003e), mainly diagnosed with cancer, were included in this study. Patient characteristics were summarized in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePatient characteristics\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u0026nbsp;=\u0026nbsp;102\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOverall\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAge\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMean\u0026nbsp;(SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e60(12)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[Min,\u0026nbsp;Max]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e[26,87]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eGender\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e55\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e47\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eBMI\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMean (SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e23(3)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[Min, Max]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e[15,32]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eInjected dose (MBq)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e[740,1110]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eWating time (min)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e[168, 265]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDiagnosis\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLung cancer\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e33\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBreast\u0026nbsp;cancer\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e21\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eProstate cancer\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e11\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGastric cancer\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eColon cancer\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNasopharyngeal carcinoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLiver cancer\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCholangiocarcinoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEndometrial carcinoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOthers\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eImage Quantitative Analysis\u003c/h3\u003e\n\u003cp\u003e3min images rated 1 point, insufficient for clinical diagnosis. 3min-DL and 20 min images displayed excellent image quality (Figure. 2, 3, 4). As shown in Figure. 5, the mean 5-point Likert scale scores by 2 reviewers of 3min-DL images were higher than those of 20min image in general image quality, Tc distribution, artifacts, and diagnostic confidence (mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD 3.44\u0026thinsp;\u0026plusmn;\u0026thinsp;0.64 vs. 3.10\u0026thinsp;\u0026plusmn;\u0026thinsp;0.64, 3.45\u0026thinsp;\u0026plusmn;\u0026thinsp;0.64 vs. 3.07\u0026thinsp;\u0026plusmn;\u0026thinsp;0.62, 3.46\u0026thinsp;\u0026plusmn;\u0026thinsp;0.63 vs. 3.10\u0026thinsp;\u0026plusmn;\u0026thinsp;0.61, 3.78\u0026thinsp;\u0026plusmn;\u0026thinsp;0.66 vs. 3.46\u0026thinsp;\u0026plusmn;\u0026thinsp;0.68. P＜0.0001). The SUV\u003csub\u003emax\u003c/sub\u003e obtained from 3min-DL and 20min image have no statistical difference (P\u0026thinsp;=\u0026thinsp;0.973), and a strong linear relationship were verified between SUV\u003csub\u003emax\u003c/sub\u003e of lesions in 3min-DL and 20min images (Y\u0026thinsp;=\u0026thinsp;0.9881*X་0.2065, r\u0026thinsp;=\u0026thinsp;0.987; P\u0026thinsp;\u0026lt;\u0026thinsp;0.0001; Figure.6). PSNR and SSIM of 3min-DL images were significantly higher than those of 3min images (51.44 vs 38.44, 0.863 vs 0.752; P\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePSNR and SSIM of 3min-DL, 3min images\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePSNR\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSSIM\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3min\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e38.44\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.7522\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3min-DL\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e51.44\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8633\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e\n\u003ch3\u003eDiagnostic Performance Analysis\u003c/h3\u003e\n\u003cp\u003eThe diagnosis results of 20min and 3min-DL showed a high consistency between two reviewers (Kappa\u0026thinsp;=\u0026thinsp;0.822, 0.732). The sensitivity, specificity and accuracy of 20min and 3min-DL had no difference by both reviewers (0.903 vs 0.806, 0.873 vs 0.873, 0.882 vs 0.853; 0.867 vs 0.806, 0.944 vs 0.936, 0.912 vs 0.920, P\u0026thinsp;\u0026gt;\u0026thinsp;0.05) (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSensitivity, specificity and accuracy of 20min and 3min-DL images by two reviewers\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003ereviewer1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c7\" namest=\"c5\"\u003e \u003cp\u003ereviewer2\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSensitivity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSpecificity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSensitivity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eSpecificity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e20min\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.903(28/31)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.873(62/71)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.882(90/102)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.867(26/31)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.936(67/72)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.912(93/102)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3min-DL\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.806(25/31)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.873(62/71)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.853(87/102)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.806(25/31)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.944(67/71)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.902(92/102)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eX\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.384\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.104\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.058\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u0026nbsp;value\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.279\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\n＞0.999\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.535\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.747\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.81\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eSPECT bone scintigraphy continues to be a great-volume nuclear imaging procedure, offering the advantage of total body examination with high sensitivity but poor specificity[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. The use of additional SPECT/CT for evaluating suspicious or equivocal lesions has been shown to improve diagnostic confidence and specificity [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e][\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. But the extra multimodal information obtained with SPECT/CT leads to an increased scanning time. Reduction of acquisition time and radioactive dose in SPECT/CT have become a major focus of attention. The short acquisition time may be considered more comfortable to patients especially children, obese people and painful patients those are intolerant of prolonged immobility during the acquisition period. Reduction of radioactive dose can reduce radiation exposure of children and patients who require multiple follow-up during examinations.\u003c/p\u003e \u003cp\u003eOur previous study has generated high-quality bone scan SPECT images from 1/7 scan time SPECT images using deep learning method with a small sample and confirmed that this method yielded significant image quality improvement in the noise level, details of anatomical structure and SUV accuracy[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. The previous deep learning based SPECT enhancement studies always build on simulated images from randomly under-sampled list mode data[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e][\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e][\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. To imitate the real clinical fast scan environment, this work adopted continuous examination and use the same reconstruction method. The fast scan of 102 patients with different types of cancer were enhanced by the pretrained deep learning model.\u003c/p\u003e \u003cp\u003eThe SPECT/CT image quality in this study was assessed by two independent nuclear radiologists with 5 and 10 years experience in SPECT/CT diagnosis. The reading results of both radiologists showed that the image qualities of the 3min SPECT were not good enough to have diagnostic value with 1 point in 5-point Likert scale, and not included in the statistical evaluation. Though the standard acquisition always received the highest score in deep learning based PET image enhancement[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e], the general image quality, detail of \u003csup\u003e99m\u003c/sup\u003eTc-MDP, presence of artifacts, and general diagnostic confidence of 3min-DL images were significantly superior to those of 20min images (P＜0.0001) in this SPECT bone scan study. A reduction of noise level of soft tissue and more uniform and coherent distribution of normal bone in 3min-DL SPECT images were observed compared with 20min SPECT as shown in Figure. 2, 3, 4. This could be a result of high noise level of SPECT system and deep learning method trend to generate expected pixel values[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. The diagnosis results of both 20min and 3min-DL SPECT/CT showed substantial agreement in the 2 reviewers\u0026rsquo; image evaluation (Kappa\u0026thinsp;=\u0026thinsp;0.822, 0.732). A SPECT study showed that shorting acquisition SPECT scan time to 1/4 standard time resulted in an excellent inter-observer agreement without affecting the diagnosis[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Inconsistent with their findings, when the scanning time was reduced to 3 minutes, the image quality was greatly degraded and had a grainy appearance in our study. After deep-learning technology application, the algorithm-enhanced 3 min images showed excellent image quality and could meet the needs of clinical diagnosis in our study .\u003c/p\u003e \u003cp\u003e \u003csup\u003e99m\u003c/sup\u003eTc-MDP chemisorbs and binds to the hydroxyapatite crystals and is marker of bone turnover and bone perfusion. It rapidly localizes to bone and clears quickly from background, making it favorable for imaging[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. Even a 5% change in bone turnovercan be detected on bone imaging, bone imaging can often detect active bone formation in the skeleton related to malignant and benign disease [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. For some diseases, such as infection, trauma, sclerosis or other origin, focus of increased radioactive tracer uptake could cause false-positive results[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. Therefore, diagnosis based on SPECT/CT imaging relies to some extent on CT imaging and the readers\u0026rsquo; experience. The sensitivity, specificity and accuracy of 3min-DL images have no significant difference from 20min images of both reviewers in our study (0.903 vs 0.806, 0.873 vs 0.873, 0.882 vs 0.853; 0.867 vs 0.806, 0.944 vs 0.936, 0.912 vs 0.920, P\u0026thinsp;\u0026gt;\u0026thinsp;0.05). These results were similar to other diagnostic performance of traditional SPECT/CT bone scans studies[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e][\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. This study showed that 3min-DL SPECT/CT images not only sped up the scanning time but also reached the diagnostic level of traditional SPECT/CT. Fast scanning would be particularly beneficial to the patients who are difficult to maintain a prolonged horizontal position for SPECT/CT imaging due to pain or other reasons.\u003c/p\u003e \u003cp\u003eOur previous study has shown that SUV\u003csub\u003emax\u003c/sub\u003e obtained from SPECT/CT could play an important role in differentiating benign and malignant bone diseases[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. In this study, the SUV\u003csub\u003emax\u003c/sub\u003e obtained from 3min-DL and 20min SPECT images has no statistical difference (P\u0026thinsp;=\u0026thinsp;0.973), there was a strong linear relationship between the SUV\u003csub\u003emax\u003c/sub\u003e of lesions in 3min-DL and 20min images (r\u0026thinsp;=\u0026thinsp;0.987; P\u0026thinsp;\u0026lt;\u0026thinsp;0.0001). The result indicated that quantitative evaluation SUV\u003csub\u003emax\u003c/sub\u003e of 3min ultra-high-speed SPECT/CT bone scan enhanced by deep learning algorithm can also be used in clinical.\u003c/p\u003e \u003cp\u003eDespite this research explored the clinical performance of deep learning method in SPECT bone scan, it has several limitations. We addressed most tumor types in this study, but few positive cases for bone metastasis were included. Sample capacity needs to be expanded in the future study. Pathological verification of some bone lesions was challenging, the gold standard based on the clinical history and other imaging studies might cause bias in the results. Our experiment bases on SPECT/CT bone images, more types of SPECT/CT studies such as SPECT MPI should be explored. Though dose reduction and fast scan has the same physical essence, minimize injection dosage while maintaining diagnostic accuracy should be further evaluated.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study evaluated the image quality and diagnostic efficiency of ultra-fast SPECT/CT bone imaging based on deep learning approach with a large population. The results indicated that the image quality and diagnostic efficiency were comparable to those of standard time SPECT/CT and could meet the needs of clinical diagnosis. The reduction of acquisition time and radioactive dose in SPECT/CT will be potentially available for routine use with the help of deep learning approach.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSPECT\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSingle-photon Emission Computed Tomography\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e99mTc-MDP\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eTechnetium 99m-Methyl Diphosphonate\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMPI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMyocardial Perfusion Imaging\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePSNR\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ePeak Signal-to-Noise Ratio\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSSIM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eStructural Similarity Index\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSUV\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eStandard Uptake Value\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSUV\u003csub\u003emax\u003c/sub\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMaximum Standardized Uptake Value\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eVOI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eVolume of Interest\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003e1. Ethics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll methods were performed in accordance with the ethical standards as laid down in the Declaration of Helsinki and its later amendments or comparable ethical standards.\u0026nbsp;Institutional Review Board approval from Ethical Review of Medical Ethics Committee of Shanghai East Hospital EC. D（BG）. 009. 02.1 was obtained.\u0026nbsp;Written informed consent was obtained from all participants.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2. Consent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3. Availability of data and material\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4. Competing interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAuthor NG is stockholder of RadioDynamic Healthcare.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e5.Funding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study was partially supported by the Key Specialty Construction Project of Pudong Health and Family Planning Commission of Shanghai, PWZzk2017-24.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e6.Authors\u0026apos; contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNQ, BP, QM participated in its design and coordination, draft the manuscript; NQ, BP conducted statistical processing on the data; NQ, YY reviewed the images; JZ, NG determined the ideas of the article. QM, HC , and WW contributed to data collection; HL, TF, XJ, NG, and JZ provided critical review and substantially revised the manuscript. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e7. Acknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eGhanem MA, Dannoon S, Elgazzar AH. The added value of SPECT-CT in the detection of heterotopic ossification on bone scintigraphy. Skeletal Radiol. 49: 291\u0026ndash;298.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrenner DO MMM CPE AI, Koshy MDJ, Morey MDJ, Lin MDC, DiPoce. MD J The Bone Scan Semin Nucl Med. 2012;42:11\u0026ndash;26.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGnanasegaran MDG, Barwick MRCPFRCRT, Adamson MSc K, Mohan MRCPH, Sharp HNCD, Fogelman MDI. Multislice SPECT/CT in Benign and Malignant Bone Disease: When the Ordinary Turns Into the Extraordinary. Semin Nucl Med. 2009;39:431\u0026ndash;42.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKalki K, Blankespoor SC, Brown JK, Hasegawa BH, Dae MW, Chin M, et al. Myocardial Perfusion Imaging with a Combined X-Ray CT and SPECT System. J Nucl Med. 1997;38:1535\u0026ndash;40.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUTSUNOMIYA D, SHIRAISHI S, IMUTA M, TOMIGUCHI S, KAWANAKA K, MORISHITA S, et al. Added value of SPECT/CT fusion in assessing suspected bone metastasis: Comparison with scintigraphy alone and nonfused scintigraphy and CT. Radiology. 2006;238:264\u0026ndash;71.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMariani G, Bruselli L, Kuwert T, Kim EE, Flotats A, Israel O, et al. A review on the clinical uses of SPECT/CT. Eur J Nucl Med Mol Imaging. 2010;37:1959\u0026ndash;85.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePalmedo H, Marx C, Ebert A, Kreft B, Ko Y, T\u0026uuml;rler A, et al. Whole-body SPECT/CT for bone scintigraphy: diagnostic value and effect on patient management in oncological patients. Eur J Nucl Med Mol Imaging. 2014;41(1):59\u0026ndash;67.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePicone V, Makris N, Boutevin F, Roy S, Playe M, Soussan M. Clinical validation of time reduction strategy in continuous step-and-shoot mode during SPECT acquisition. EJNMMI Phys. 2021;8(1):10.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShao W, Rowe SP, Du Y. Artificial intelligence in single photon emission computed tomography (SPECT) imaging: a narrative review. Ann Transl Med. 2021;9(9):820.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCheng Z, Wen J, Huang G, Yan J. Applications of artificial intelligence in nuclear medicine image generation. Quant Imaging Med Surg. 2021;11(6):2792\u0026ndash;822.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang Q, Sun J, Mok GSP. Low dose SPECT image denoising using a generative adversarial network. 2019.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang J, Shi L, Wang R, Miller EJ, Sinusas AJ, Liu C, et al. Direct attenuation correction using deep learning for cardiac SPECT: A feasibility study. J Nucl Med. 2021;62.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRamon AJ, Yang Y, Pretorius PH, Johnson KL, King MA, Wernick MN. Initial Investigation of Low-Dose SPECT-MPI via Deep Learning. 2018 IEEE Nucl Sci Symp Med Imaging Conf NSS/MIC 2018 - Proc. 2018; 1\u0026ndash;3.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRamon AJ, Yang Y, Pretorius PH, Johnson KL, King MA, Wernick MN. Improving Diagnostic Accuracy in Low-Dose SPECT Myocardial Perfusion Imaging with Convolutional Denoising Networks. IEEE Trans Med Imaging. 2020;39:2893\u0026ndash;903.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRyd\u0026eacute;n T, Van Essen M, Marin I, Svensson J, Bernhardt P, universitet G, et al. Deep-Learning Generation of Synthetic Intermediate Projections Improves 177 Lu SPECT Images Reconstructed with Sparsely Acquired Projections. J Nucl Med. 62: 528\u0026ndash;535.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLin C, Chang Y-C, Chiu H-Y, Cheng C-H, Huang H-M. Reducing scan time of paediatric 99m Tc-DMSA SPECT via deep learning. Clin Radiol. 2021;76:315.e13.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShiri I, Sabet KA, Arabi H, Pourkeshavarz M, Teimourian B, Ay MR, et al. Standard SPECT myocardial perfusion estimation from half-time acquisitions using deep convolutional residual neural networks. J Nucl Cardiol. 2021;28(6):2761\u0026ndash;79.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePan B, Qi N, Meng Q, Wang J, Peng S, Qi C, et al. Ultra high speed SPECT bone imaging enabled by a deep learning enhancement method: a proof of concept. EJNMMI Phys. 2022;9(1):43.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrenner AI, Koshy J, Morey J, Lin C, DiPoce J. The bone scan. Semin Nucl Med. 2012;42(1):11\u0026ndash;26.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLy J, Minarik D, J\u0026ouml;gi J, Wollmer P, Tr\u0026auml;g\u0026aring;rdh E. Post-reconstruction enhancement of [18F]FDG PET images with a convolutional neural network. EJNMMI Res. 2021;11(1):48.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLehtinen J, Munkberg J, Hasselgren J, Laine S, Karras T, Aittala M, et al. Noise2Noise: Learning Image Restoration without Clean Data. 2018. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://go.exlibris.link/P27f6SpV\u003c/span\u003e\u003cspan address=\"https://go.exlibris.link/P27f6SpV\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZacho HD, Manresa JAB, Aleksyniene R, Ejlersen JA, Fledelius J, Bertelsen H, et al. Three-minute SPECT/CT is sufficient for the assessment of bone metastasis as add-on to planar bone scintigraphy: prospective head-to-head comparison to 11-min SPECT/CT. EJNMMI Res. 2017;7:1.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVan den Wyngaert T, Strobel K, Kampen WU, Kuwert T, van der Bruggen W, Mohan HK, et al. The EANM practice guidelines for bone scintigraphy. Eur J Nucl Med Mol Imaging. 2016;43:1723\u0026ndash;38.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVijayanathan MRCPFRCRS, Butt MBBSFRCRS, Gnanasegaran MDG, Groves MDAM. Advantages and Limitations of Imaging the Musculoskeletal System by Conventional Radiological, Radionuclide, and Hybrid Modalities. Semin Nucl Med. 2009;39:357\u0026ndash;68.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang Y, Li B, Wu B, Yu H, Song J, Xiu Y, et al. Diagnostic performance of whole-body bone scintigraphy in combination with SPECT/CT for detection of bone metastases. Ann Nucl Med. 2020;34:549\u0026ndash;58.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMostafa R, Abdelhafez YG, Abougabal M, Nardo L, Elkareem MA. Two-bed SPECT/CT versus planar bone scintigraphy: prospective comparison of reproducibility and diagnostic performance. Nucl Med Commun. 2021;42:360\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQi N, Meng Q, You Z, Chen H, Shou Y, Zhao J. Standardized uptake values of 99m Tc-MDP in normal vertebrae assessed using quantitative SPECT/CT for differentiation diagnosis of benign and malignant bone lesions. BMC Med Imaging. 2021;21:39.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":true,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Bone, SPECT/CT, Deep learning, Ultra-fast, Image quality, Diagnostic efficiency","lastPublishedDoi":"10.21203/rs.3.rs-2190739/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-2190739/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eTo evaluate clinical performance of deep learning enhanced ultra-fast SPECT/CT bone scan.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eOne hundred and two patients were enrolled in this retrospective study. The probable malignant tumor sites continuously underwent a 20min SPECT/CT and a 3min SPECT scan. A deep learning model was applied to generate algorithm-enhanced images (3min-DL SPECT). Two reviewers evaluated general image quality, \u003csup\u003e99m\u003c/sup\u003eTc-MDP distribution, artifacts, and diagnostic confidence independently. The sensitivity, specificity, accuracy, and inter-observer agreement were calculated. Linear regression was analyzed for lesion SUV\u003csub\u003emax\u003c/sub\u003e between 3min-DL and 20min SPECT. Peak signal-to-noise ratio (PSNR), image similarity (SSIM) were evaluated.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe general image quality, \u003csup\u003e99m\u003c/sup\u003eTc-MDP distribution, artefact, and diagnostic confidence of 3min-DL images were significantly superior to those of 20min images (P\u0026thinsp;\u0026lt;\u0026thinsp;0.0001). The sensitivity, specificity and accuracy of 20min and 3min-DL SPECT/CT had no difference by both reviewers (0.903 vs 0.806, 0.873 vs 0.873, 0.882 vs 0.853; 0.867 vs 0.806, 0.944 vs 0.936, 0.912 vs 0.920, P\u0026thinsp;\u0026gt;\u0026thinsp;0.05). The diagnosis results of 20min and 3min-DL images showed a high inter-observer agreement (Kappa\u0026thinsp;=\u0026thinsp;0.822, 0.732). PSNR and SSIM of 3min-DL images were significantly higher than 3min images (51.44 vs 38.44, 0.863 vs 0.752, P\u0026thinsp;\u0026lt;\u0026thinsp;0.05). A strong linear relationship was found between the SUV\u003csub\u003emax\u003c/sub\u003e of 3min-DL and 20min images (r\u0026thinsp;=\u0026thinsp;0.987; P\u0026thinsp;\u0026lt;\u0026thinsp;0.0001).\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eAn ultra-fast SPECT/CT with 1/7 scan time could be enhanced by deep learning method to have competitive image quality and equivalent diagnostic value to those of standard acquisition.\u003c/p\u003e","manuscriptTitle":"Deep learning enhanced ultra-fast SPECT/CT bone scan: quantitative assessment and clinical performance","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2022-10-31 21:28:26","doi":"10.21203/rs.3.rs-2190739/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"62d10ad6-7e74-4fe2-b96b-98364049bcd9","owner":[],"postedDate":"October 31st, 2022","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2023-01-30T09:13:07+00:00","versionOfRecord":[],"versionCreatedAt":"2022-10-31 21:28:26","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-2190739","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-2190739","identity":"rs-2190739","version":["v1"]},"buildId":"cBFmMYwuxLRRLfASyISRj","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}