{"paper_id":"28e2ca46-49cd-4a1a-b3b3-c27cadbd9572","body_text":"1\n1 Title: \n2 Auto-detection of motion artifacts on CT pulmonary angiograms with a physician-trained AI \n3 algorithm \n4 Giridhar Dasegowda1,2\n5 Bernardo C Bizzo1,2\n6 Parisa Kaviani 1,2\n7 Lina Karout1,2\n8 Shadi Ebrahimian1\n9 Subba R Digumarthy1\n10 Nir Neumark2\n11 James Hillis1,2\n12 Mannudeep K Kalra1,2\n13 Keith J Dreyer1,2\n14\n15 1- Massachusetts General Hospital and Harvard Medical School\n16 2- Mass General Brigham Data Science Office\n17 Corresponding author: Dr. Bernardo C Bizzo\n18 Email: bbizzo@mgh.harvard.edu\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \nNOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.\n\n2\n19 Abstract\n20 Purpose: Motion-impaired CT images can result in limited or suboptimal diagnostic \n21 interpretation (with missed or miscalled lesions) and patient recall. We trained and tested an \n22 artificial intelligence (AI) model for identifying substantial motion artifacts on CT pulmonary \n23 angiography (CTPA) that have a negative impact on diagnostic interpretation. \n24 Methods: With IRB approval and HIPAA compliance, we queried our multicenter radiology \n25 report database (mPower, Nuance) for CTPA reports between July 2015 - March 2022 for the \n26 following terms: \"motion artifacts,\" \"respiratory motion,\" \"technically inadequate,\" and \n27 \"suboptimal\" or \"limited exam.\" All CTPA reports belonged to two quaternary (Site A, n= 335; B, \n28 n= 259) and a community (C, n= 199) healthcare sites. A thoracic radiologist reviewed CT \n29 images of all positive hits for motion artifacts (present or absent) and their severity (no \n30 diagnostic effect or major diagnostic impairment). Coronal multiplanar images belonging to 793 \n31 CTPA exams were de-identified and exported offline into an AI model building prototype \n32 (Cognex Vision Pro, Cognex Corporation) to train an AI model to perform two-class classification \n33 (\"motion\" or \"no motion\") with data from the three sites (70% training dataset, n= 554; 30% \n34 validation dataset, n= 239). Separately, data from Site A and Site C were used for training and \n35 validating; testing was performed on the Site B CTPA exams. A 5-fold repeated cross-validation \n36 was performed to evaluate the model performance with accuracy and receiver operating \n37 characteristics analysis (ROC). \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \n\n3\n38 Results: Among the CTPA images from 793 patients (mean age 63 ± 17 years; 391 males, 402 \n39 females), 372 had no motion artifacts, and 421 had substantial motion artifacts. The statistics \n40 for the average performance of the AI model after 5-fold repeated cross-validation for the two-\n41 class classification included 94% sensitivity, 91% specificity, 93% accuracy, and 0.93 area under \n42 the ROC curve (AUC: 95% CI 0.89-0.97). \n43 Conclusion: The AI model used in this study can successfully identify CTPA exams with \n44 diagnostic interpretation limiting motion artifacts in multicenter training and test datasets. \n45 Clinical relevance: The AI model used in the study can help alert the technologists about \n46 the presence of substantial motion artifacts on CTPA where a repeat image acquisition can help \n47 salvage diagnostic information.   \n48\n49\n50\n51\n52\n53\n54\n55\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \n\n4\n56 Introduction\n57 The chest represents one of the most frequently scanned body parts in CT, but also ranks high \n58 among the most challenging parts to obtain an optimal diagnostic quality. Recent technical \n59 advancements in multidetector-row CT scanners have led to tremendous improvements in \n60 diagnostic quality with lower noise, higher contrast, and fewer motion artifacts. Although with \n61 faster scanning times and better reconstruction techniques, most chest CT exams are generally \n62 optimal for diagnostic interpretation, several artifacts can still have a negative impact on \n63 diagnostic interpretations of chest CT [1]. The common artifacts include beam hardening, \n64 photon starvation, partial volume, metal, motion, and cone-beam artifacts [2, 3]. Such artifacts \n65 can hinder the optimal evaluation of lung abnormalities as well as mediastinal and vascular \n66 findings.  \n67 Prior studies have reported that motion artifacts are frequent, especially on legacy scanners, \n68 and can limit the diagnostic information from both routine chest CT as well as CT pulmonary \n69 angiography (CTPA). Motion artifacts can result in misinterpretation as the artifacts can mimic \n70 an embolus, or the artifact can cause an apparent abrupt vessel cut-off [4, 5]. In the lung \n71 parenchyma, artifacts can reduce the ability to detect and characterize both focal (such as lung \n72 nodules) and diffuse parenchymal processes. Such artifacts are especially common in CTPA of \n73 critically ill patients and patients with shortness of breath and/or persistent cough [6]. The use \n74 of wide-area detector scanners and scan capabilities such as high non-overlapping pitch and \n75 faster rotation time reduce scanning duration and have lower artifacts [3, 7]. Such fast \n76 acquisition modes are not compatible with dual-energy CT and in patients with large body \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \n\n5\n77 habitus. Furthermore, such scanners and advanced techniques still represent a minority of \n78 clinically deployed scanners even in developed countries such as the United States [8]. \n79 An automated method of detecting substantial motion artifacts at the time of scanning can \n80 help re-scan patients during the same imaging session with better coaching or faster scanning \n81 techniques. Recent advances in machine and deep learning (DL) have led to the creation of \n82 several AI algorithms in medical imaging including in the areas of image reconstruction, triaging, \n83 quality control and pathology detection [9-14]. Therefore, we trained and tested an artificial \n84 intelligence (AI) model to identify substantial motion artifacts on CTPA that have a negative \n85 impact on diagnostic interpretation.\n86\n87 Materials and Methods \n88 Study Design\n89 Our retrospective study was conducted after receiving approval from Institutional Review \n90 Board (IRB). The study was Human Insurance Portability and Accountability Act (HIPAA) \n91 compliant. The study methodology has been described in accordance to CLAIM guidelines [15]. \n92 The model was trained and tested by physicians without any prior knowledge of, or training in \n93 machine learning or coding.  \n94\n95\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \n\n6\n96 Data Definitions\n97 The study included adult patients (≥19 years) who underwent CTPA at one of the three \n98 hospitals (quaternary hospitals: Massachusetts General Hospital, Brigham and Women's \n99 Hospital; community hospital: Cooley Dickinson Hospital) within an integrated health system. A \n100 commercial radiology report search engine, Nuance mPower, was used to identify radiology \n101 reports with a mention of motion artifacts in CTPA examinations performed between Jan 2015 \n102 to November 2021. The following keywords were used to identify the eligible CTPA: \"motion \n103 artifacts,\" \"respiratory motion,\" \"technically inadequate,\" and \"suboptimal\" or \"limited exam.\" \n104 The search was optimized to identify positive CTPA with the keyword hits (without mention of \n105 “absence” or “no” within a few words of the keywords). A negative search with the exact \n106 keywords was used to identify the control CTPA exams without motion artifacts. \n107 All CTPA exams were performed on one of the 28 CT scanners in the three participating sites \n108 with single or dual-energy CT protocols using the standard of care scan protocols. For each \n109 CTPA, a thin slice coronal multiplanar image at the level of descending thoracic aorta was de-\n110 identified and exported offline from the PACS workstation. \n111\n112 Ground truth\n113 In addition to the radiology reports, a thoracic radiologist (MKK with 16 years of subspecialty \n114 experience) reviewed all CTPA and opined on the presence or absence of substantial motion \n115 artifacts. Each CTPA thus had the opinions of two radiologists, the reporting radiologist, and the \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \n\n7\n116 study coinvestigator radiologist. Substantial motion artifacts were defined as CTPA exams with \n117 the presence of motion artifacts involving both lungs (at least 50% of each lung) and limiting \n118 the ability to assess pulmonary embolism or parenchyma. Minor motion artifacts involving a \n119 single lobe or smaller portions of lungs without effect on the diagnostic evaluation of \n120 pulmonary embolism or parenchyma were labeled as negative for substantial motion artifacts. \n121 Although extremely common, minor artifacts are not as important since they should not trigger \n122 repeat imaging. CTPA with evidence of “white lungs” (diffuse parenchymal opacities), a \n123 substantial bilateral lung volume loss, or pneumonectomies were excluded from the study (n= \n124 50 CTPA exams).  These cases were excluded since it is difficult to assess motion artifacts’ \n125 impact on the pulmonary evaluation in such cases. CTPA exams without the complete inclusion \n126 of lung apex and bases were excluded from the study.\n127\n128 Model\n129 The AI model was trained on a deep learning model-building platform, Vision Pro Deep Learning \n130 (VPDL, COGNEX Corporation, Natick, MA). The software enables users to train DL models based \n131 on a labeled image dataset using a vision-optimized deep neural network. The users require no \n132 formal programming or coding knowledge or experience. The VPDL platform has two different \n133 options for training classification models: High detail Mode (HDM) and Focused Mode (FM). The \n134 HDM enables model training for challenging or complex applications (such as pixel-level \n135 information in image domain) and provides higher accuracy. A heat map is also generated in \n136 HDM that indicates the image region that was most influential in the classification decision. FM \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \n\n8\n137 enables fast training of models for simple applications (such as the distinction between \n138 different image types or body parts). We used the HDM classification model for identifying \n139 CTPA examinations with substantial motion artifacts. \n140\n141 Training\n142 A physician co-investigator (GD with one-year post-doctoral research fellowship experience in \n143 thoracic imaging) trained the AI model on the VPDL platform without prior programming or \n144 data science knowledge. The images were de-identified, exported from the PACS workstation \n145 (Visage), and then uploaded onto the software platform which was installed on a virtual \n146 machine within the hospital intranet to maintain data security and privacy. Within the platform, \n147 the study coinvestigator labeled each uploaded image as \"motion\" (with substantial motion \n148 artifacts) or \"no motion\" (without substantial motion artifacts) based on the assessment from \n149 the thoracic radiologist. In the first training, all CTPA examinations from three hospitals were \n150 included. The software randomized the training and validation data set with a 70%-30% \n151 distribution, respectively. We performed five-fold repeated cross-validation to evaluate the \n152 robustness of the model. The output was recorded and separately analyzed.\n153 To establish the inter-institutional generalizability of the model, we trained a model using \n154 images from two hospitals (A and C, after excluding the data from site B), and then tested the \n155 algorithm on CTPA data from the third hospital (site B). \n156\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \n\n9\n157 Statistical Analysis\n158 Information on the distribution of true positive, true negative, false positive, and false negative \n159 CTPA was recorded in Microsoft Excel worksheets (Microsoft Inc). Data were analyzed with \n160 SPSS statistical software, version 26 (IBM Inc.). The performance of the AI model was evaluated \n161 using sensitivity, specificity, accuracy, and area under the curve (AUC) for the receiver operating \n162 characteristic (ROC) analysis. For the five-fold repeated cross-validation, the average of the five \n163 models was considered. We also estimated the F-score for AI model performance, as a measure \n164 of the harmonic average of precision and recall/sensitivity.\n165\n166 Results\n167 Our study included 793 CTPA examinations from 793 adult patients (mean age= 63 ± 17; 391 \n168 men, 402 women). The distribution of CTPA across each site is as follows: Site A, n= 335; Site B, \n169 n= 259; Site C, n= 199. A total of 455/793 (57%) CTPA examinations were performed during the \n170 emergency visit, while 277/793 (35%) and 111/793 (14%) examinations were among inpatients \n171 and outpatients, respectively. Most CTPA with substantial motion artifacts were either from the \n172 emergency department (n = 213/455) or inpatients (n = 146/227) with a minority of patients \n173 coming with outpatient referrals (n = 33/111). There was no discrepancy between the reporting \n174 radiologists and the research radiologist for the presence of substantial motion artifacts.\n175\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \n\n10\n176 Model validation at three sites\n177 Among CTPA exam datasets from all three sites in model training, 471 CTPA (n= 421/793, \n178 53.1%) had substantial motion artifacts and 372 CTPA (n= 372/793, 46.9%) were without \n179 substantial artifacts. For the average performance after 5-fold repeated cross-validation for the \n180 two-class classification (as either with or without substantial motion artifacts), on the 30% \n181 validation data set, the AI model had a sensitivity of 94%, specificity of 91%, 93% accurate with \n182 AUC of 0.93 (95% confidence interval [95% CI] 0.89 - 0.97). The best performing model had F-\n183 scores of 96% and 95% for identifying CTPA with and without substantial motion artifacts, \n184 respectively. The two-class classification of the AI model is summarized in Fig 1.\n185\n186 Fig 1. Coronal MPR images of CTPA examinations in two patients with (A) and without \n187 substantial motion artifacts. The AI algorithm correctly classified the images as with motion (A: \n188 with 100% confidence score) and “no motion” (B: with 99.4% confidence score). \n189\n190 Model testing\n191 External testing with Sites A and C training datasets and Site B as the test site, the model \n192 performance statistics were 85% sensitivity, 90% specificity, 86% accuracy, and an AUC of 0.87 \n193 (95% CI 0.82 - 0.92). One hundred forty-seven CTPA exams were correctly classified into those \n194 with substantial motion artifacts (true positive), and nine CTPA were mislabeled as positive for \n195 motion artifacts (false positive). Seventy-seven CTPA exams were correctly classified as without \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \n\n11\n196 motion artifacts (true negative), and twenty-six CTPA exams were misclassified as without \n197 motion (false negative). The confusion matrix of model performance is represented in Fig 2 and \n198 the model performance for validation and test data is summarized in Table 1.\n199\n200 Fig 2: Confusion matrix generated by the AI model for the testing (A true positive rate \n201 [sensitivity] – AI correctly labeled images with motion; B false-negative rate – AI incorrectly \n202 labeled “motion” images as without motion; C false positive rate – AI incorrectly labeled “no \n203 motion” images as with substantial motion artifacts; D true negative rate [specificity]– AI \n204 correctly identified images without motion artifacts). \n205\n206 Table 1. Summary of our AI model’s performance for motion artifacts detection in CTPA exams \n207 from the three participating sites (A, B, C). (key: + present; - absent; CI confidence interval).\n208\nTraining Test Sensitivity Specificity Accuracy AUC 95% CI (AUC)\nMotion + 295 126Model validation\n(training + validation \ndata from all sites) Motion - 259 113\n94% 91% 93% 0.93 0.89-0.97\nMotion + 196 173Model testing\n(training from A, C \nand testing on B) Motion - 231 86\n85% 90% 86% 0.87 0.82-0.92\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \n\n12\n209 Discussion\n210 We report high accuracy, sensitivity, and specificity of physician-trained and tested AI model for \n211 identifying substantial motion artifacts on CTPA examinations. Prior studies have reported on \n212 the ability of AI algorithms to identify anatomic regions with motion artifacts [16]. Our study \n213 uses a single coronal multiplanar reformatted image per CT exam and therefore could be more \n214 time-efficient, while ignoring region-specific, sparse motion artifacts which do not require \n215 repeat acquisition [17-19] . To our best knowledge, there are no peer-reviewed reports on the \n216 use of AI models for identifying motion artifacts in CTPA or chest CT examinations.  \n217 Beri et al. reported that their AI algorithm trained to identify the motion affected regions by \n218 segmenting the entire image series had an  AUC of 0.81 [18]. We achieved a similar AUC of 0.88 \n219 with a single coronal MPR image per CTPA examination. Other studies [17, 19] on motion \n220 artifact detection in CT images have focused in coronary CT angiography (CCTA) rather than \n221 CTPA examinations. Ma et al. reported 91% sensitivity and 71% specificity with 87% accuracy \n222 for detecting motion artifacts in CCTA examinations [17]. Likewise, Elss et al. trained an AI \n223 algorithm for identifying motion artifacts on CCTA and reported an accuracy of 94% [19]. Xu et \n224 al. reported a fully automatic AI for grading image quality (motion artifacts) of CCTA using semi-\n225 automatic labeling and tracking of the coronary arteries [20] Based on the identification and \n226 estimation of motion artifact, other investigators have reported on motion artifact correction \n227 and compensation solutions for head and cardiac CT examinations [21, 22]. \n228 Although not yet cleared by the US Food and Drug Administration (FDA), our proof-of-concept \n229 study and the AI model used in the study may have potential clinical implications. Firstly, given \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \n\n13\n230 the high frequency of motion artifacts in chest CT and CTPA examinations, our study highlights \n231 the role of AI models in identifying motion artifacts. If integrated with CT scanners, such AI \n232 models can efficiently detect and alert the CT technologist to the presence of motion artifacts \n233 that are likely to have a substantial effect on the diagnostic interpretation. Secondly, such \n234 artifacts can be present on chest CT exams acquired on both the older, less advanced scanners \n235 and newer, faster scanners. On the latter, modification of scanning parameters can enable \n236 faster scanning of the entire chest in under 1 second [23]. On the older scanner, the AI-\n237 generated surveillance for motion-impaired CTPA or chest CT examinations can prompt \n238 technologist to give better breath-hold instructions [24] and modify scan parameters or when \n239 possible scan at-risk patients on faster scanners. Thirdly, most diagnostic CT scanners in our \n240 institution (MGH) have two scan protocols – one for patients who can hold their breath and the \n241 other for those who are unable to hold their breath or have substantial motion artifacts on \n242 their initial CT acquisition. Despite instructions to our CT technologists to always review the CT \n243 images for motion artifacts before taking the patient off the CT table, most technologists \n244 cannot or do not comply with the recommendation especially during pandemic times. As a \n245 result, interpreting radiologists either report CT with a disclaimer on motion-limited diagnostic \n246 value or request patient recall and rescanning. By automating the detection of motion artifacts, \n247 AI models such as the one reported in our study could potentially help address compliance and \n248 reacquisition, when appropriate.  Fourthly, several modern scanners automatically generate \n249 multiplanar reformatted images as soon as the data acquisition is complete, so the use of \n250 coronal MPR image for our model is not a rate-limiting step. Although our model would still \n251 require the identification of the single image at the descending aorta level. Finally, in hospitals \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \n\n14\n252 with high CT volumes, the task of quality assessment for image quality is time-consuming and \n253 labor-intensive. In such sites, the AI model used in this study can be used to analyze image \n254 quality in a retrospective manner. Derived statistics can then be used to develop faster scan \n255 protocols and track their impact on diagnostic evaluability.  Our study has limitations. We did \n256 not perform a power analysis to determine the number of training and testing cases needed to \n257 prove our hypothesis. Although the high level of performance associated with the AI model \n258 suggests that our sample size was adequate, it is conceivable that more training data could \n259 generate better results. Likewise, for external testing, we had only a single site with completely \n260 independent clinical operations, and we would require additional external sites to support the \n261 claim of generalizability of the AI model. The AI model’s performance can however vary with \n262 the change in scan protocols (low dose chest CT versus CTPA protocols) and scanners (for those \n263 scanners without input training data). Exclusion of CTPA with “white lungs” (diffuse \n264 parenchymal opacities), a substantial bilateral lung volume loss, or pneumonectomies also \n265 limits the application of our AI model in such patients.\n266 Another limitation of our study pertains to the use of a single coronal MPR image per CT for \n267 assessing substantial motion artifacts as opposed to the entire image series for prior studies \n268 [18]. It is therefore possible that the performance of our AI model can differ on the entire \n269 image series (transverse or coronal) as compared to its current performance on a single image. \n270 Despite a high model performance, it is possible that motion artifacts in other anatomic \n271 location can affect evaluation of key findings. Given the full longitudinal coverage of anatomy in \n272 the coronal plane, we believe that such “missed motion artifacts in key location” is less likely. \n273 The AI building platform at the time of manuscript preparation cannot easily group entire image \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \n\n15\n274 series and is limited to 2D image input, which impacted our decision to train on the coronal \n275 MPR image instead of an axial series. Finally, our model training and testing were limited to \n276 CTPA and might not apply to other chest CT protocols or body regions. \n277 In conclusion, the physician-trained and tested AI model can help identify substantial motion \n278 artifacts on CT pulmonary angiography. Automatic recognition of such artifacts can help CT \n279 technologists apply faster scan protocols and reacquire images to mitigate the impact of \n280 substantial motion impairment on diagnostic evaluability.  \n281\n282\n283\n284\n285\n286\n287\n288\n289\n290\n291\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \n\n16\n292 References  \n293 1. Veikutis V, Budrys T, Basevicius A, Lukosevicius S, Gleizniene R, Unikas R, et al. Artifacts in \n294 computer tomography imaging: how it can really affect diagnostic image quality and confuse clinical \n295 diagnosis? Journal of Vibroengineering. 2015;17(2):995-1003.\n296 2. Boas FE, Fleischmann D. CT artifacts: causes and reduction techniques. Imaging Med. \n297 2012;4(2):229-40.\n298 3. Barrett JF, Keat N. Artifacts in CT: recognition and avoidance. Radiographics. 2004;24(6):1679-\n299 91.\n300 4. Johnson PT. Artifacts mimicking pulmonary embolism. Pearls and Pitfalls in Cardiovascular \n301 Imaging: Pseudolesions, Artifacts, and Other Difficult Diagnoses. 2015:134.\n302 5. Wittram C, Maher MM, Yoo AJ, Kalra MK, Shepard JA, McLoud TC. CT angiography of pulmonary \n303 embolism: diagnostic criteria and causes of misdiagnosis. Radiographics. 2004;24(5):1219-38.\n304 6. Jones SE, Wittram C. The indeterminate CT pulmonary angiogram: imaging characteristics and \n305 patient clinical outcome. Radiology. 2005;237(1):329-37.\n306 7. Aviram G, Levy G, Fishman JE, Blank A, Graif M. Pitfalls in the diagnosis of acute pulmonary \n307 embolism on spiral computer tomography. Current problems in diagnostic radiology. 2004;33(2):74-84.\n308 8. Global Imaging Market Outlook Report 2019 [Available from: https://imvinfo.com/product/imv-\n309 2019-global-imaging-market-outlook-report/.\n310 9. Huhtanen H, Nyman M, Mohsen T, Virkki A, Karlsson A, Hirvonen J. Automated detection of \n311 pulmonary embolism from CT-angiograms using deep learning. BMC Medical Imaging. 2022;22(1):1-10.\n312 10. Soffer S, Klang E, Shimon O, Barash Y, Cahan N, Greenspana H, et al. Deep learning for \n313 pulmonary embolism detection on computed tomography pulmonary angiogram: a systematic review \n314 and meta-analysis. Scientific reports. 2021;11(1):1-8.\n315 11. Ebrahimian S, Digumarthy SR, Homayounieh F, Bizzo BC, Dreyer KJ, Kalra MK. Predictive values \n316 of AI-based triage model in suboptimal CT pulmonary angiography. Clinical Imaging. 2022;86:25-30.\n317 12. Singh R, Wu W, Wang G, Kalra MK. Artificial intelligence in image reconstruction: The change is \n318 here. Physica Medica. 2020;79:113-25.\n319 13. Lenfant M, Chevallier O, Comby P-O, Secco G, Haioun K, Ricolfi F, et al. Deep learning versus \n320 iterative reconstruction for CT pulmonary angiography in the emergency setting: improved image \n321 quality and reduced radiation dose. Diagnostics. 2020;10(8):558.\n322 14. Dasegowda G, Bizzo B, Gupta R, Kaviani P, Ebrahimian S, Ricciardelli D, et al. Successful creation \n323 of clinical AI without data scientists or software developers: radiologist-trained AI model for identifying \n324 suboptimal chest-radiographs. Research Square; 2022.\n325 15. Mongan J, Moy L, Kahn Jr CE. Checklist for artificial intelligence in medical imaging (CLAIM): a \n326 guide for authors and reviewers. Radiology Artificial Intelligence. 2020;2(2).\n327 16. Lossau T, Nickisch H, Wissel T, Bippus R, Schmitt H, Morlock M, et al. Motion artifact recognition \n328 and quantification in coronary CT angiography using convolutional neural networks. Medical image \n329 analysis. 2019;52:68-79.\n330 17. Ma H, Gros E, Baginski SG, Laste ZR, Kulkarni NM, Okerlund D, et al. Automated quantification \n331 and evaluation of motion artifact on coronary CT angiography images. Medical Physics. \n332 2018;45(12):5494-508.\n333 18. BERI P. Detection of Motion Artifacts in Thoracic CT Scans. Available from \n334 https://pure.tue.nl/ws/portalfiles/portal/168210888/Puneet_B..pdf\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \n\n17\n335 19. Elss T, Nickisch H, Wissel T, Schmitt H, Vembar M, Morlock M, et al., editors. Deep-learning-\n336 based CT motion artifact recognition in coronary arteries. Medical Imaging 2018: Image Processing; \n337 2018: International Society for Optics and Photonics.\n338 20. Xu Y, Sushmit A, Lyu Q, Li Y, Cao X, Maltz JS, et al. Cardiac CT motion artifact grading via semi-\n339 automatic labeling and vessel tracking using synthetic image-augmented training data. Journal of X-Ray \n340 Science and Technology. 2022(Preprint):1-13.\n341 21. Maier J, Lebedev S, Erath J, Eulig E, Sawall S, Fournié E, et al. Deep learning-based coronary \n342 artery motion estimation and compensation for short-scan cardiac CT. Medical Physics. \n343 2021;48(7):3559-71.\n344 22. Su B, Wen Y, Liu Y, Liao S, Fu J, Quan G, et al. A deep learning method for eliminating head \n345 motion artifacts in computed tomography. Medical Physics. 2022;49(1):411-9.\n346 23. Doda Khera R, Nitiwarangkul C, Singh R, Homayounieh F, Digumarthy SR, Kalra MK. \n347 Multiplatform, Non-Breath-Hold Fast Scanning Protocols: Should We Stop Giving Breath-Hold \n348 Instructions for Routine Chest CT? Canadian Association of Radiologists Journal. 2021;72(3):505-11.\n349 24. Doda Khera R, Singh R, Homayounieh F, Stone E, Redel T, Savage CA, et al. Deploying clinical \n350 process improvement strategies to reduce motion artifacts and expiratory phase scanning in chest CT. \n351 Scientific Reports. 2019;9(1):1-7.\n352\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \n\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint \n\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 23, 2022. ; https://doi.org/10.1101/2022.06.23.22276818doi: medRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}