Table 1.

Overview of Deep Learning Algorithms Developed For Use in the Emergency and Clinical Setting.

Title/Author	Journal/Year/Type	Data	Data Processing	Application	Model	Performance	Reference
Pelvic Fractures: Epidemiology and Predictors of Associated Abdominal Injuries and Outcomes Demetriades et al. [9]	J. Am. Coll. Surg. 2002 Original	No DL					Demetriades D, et al. Pelvic fractures: epidemiology and predictors of associated abdominal injuries and outcomes. J Am Coll Surg. 2002 Jul;195(1):1–10. doi:10.1016/s1072-7515(02)01197-3. PMID: 12113532.
Detecting pelvic fracture on 3D-CT using deep convolutional neural networks with multi-orientated slab images Ukai et al. [10]	Scientific Reports 2021 Original	Multisource CT images acquired from 93 subjects who had one or more pelvic fractures. Multisource CT images acquired from 112 subjects identified by orthopedic surgeons as not having any fractures.	Voxel size and Intensity range harmonization	Automatically detect pelvic fractures from pelvic CT images of an evaluating subject.	DCNN: YOLOv3	Area under the curve (AUC) was 0.824, with 0.805 recall and 0.907 precision.	Ukai K, et al. Detecting pelvic fracture on 3D-CT using deep convolutional neural networks with multi-orientated slab images. Sci Rep 11, 11716 (2021). https://doi.org/10.1038/s41598-021-91144-z.
Accuracy of Abdominal Radiography in Acute Small-Bowel Obstruction: Does Reviewer Experience Matter? Thompson et al. [11]	Abdominal Imaging 2007 Original	No DL					Thompson WM, et al. Accuracy of abdominal radiography in acute small-bowel obstruction: does reviewer experience matter? AJR Am J Roentgenol. 2007 Mar;188(3):W233-8. doi:10.2214/AJR.06.0817. PMID: 17312028.
Abdominal Radiography Findings in Small-Bowel Obstruction: Relevance to Triage for Additional Diagnostic Imaging Lappas et al. [12]	AJR 2001 Original	No DL					Lappas JC, et al. Abdominal radiography findings in small bowel obstruction: relevance to triage for additional diagnostic imaging. AJR 2001; 176:167–174.
Detection of high-grade small bowel obstruction on conventional radiography with convolutional neural networks Cheng et al. [13]	Ab. Radiol. 2018 Original	3663 supine abdominal radiographs	Pixel size and Intensity range harmonization	Determine whether a deep CNN can be trained with limited image data to detect high-grade small bowel obstruction patterns on supine abdominal radiographs.	Inception v3 CNN	The neural network achieved an AUCof 0.84 on the test set (95% CI 0.78–0.89). At the maximum Youden index (sensitivity + specificity-1), the sensitivity of the system for small bowel obstruction was 83.8%, with a specificity of 68.1%.	Cheng PM, et al. Detection of high-grade small bowel obstruction on conventional radiography with convolutional neural networks. Abdom Radiol (NY) 2018;43(5):1120–1127.
Performance of a Deep Learning Algorithm for Automated Segmentation and Quantification of Traumatic Pelvic Hematomas on CT Dreizin et al. [14]	Journal of Digital Imaging 2021 Original	253 C/A/P admission trauma CT	Pixel size and Intensity range harmonization	Determine if RSTN would result in sufficiently high Dice similarity coefficients to facilitate accurate and objective volumetric measurements for outcome prediction (arterial injury requiringangioembolization).	Recurrent Saliency Transformation Network vs. 3D U-Net	Dice scores in the test set were 0.71 (SD ± 0.10) using RSTN, compared to 0.49 (SD ± 0.16) using a baseline Deep Learning Tool Kit (DLTK) reference 3D U-Net architecture.	Dreizin D, et al. “A Multiscale Deep Learning Method for Quantitative Visualization of Traumatic Hemoperitoneum at CT: Assessment of Feasibility and Comparison with Subjective Categorical Estimation.” Radiology. Artificial intelligence vol. 2,6 e190220. 11 Nov. 2020, doi:10.1148/ryai.2020190220.
Image Segmentation and Machine Learning for Detection of Abdominal Free Fluid in Focused Assessment With Sonography for Trauma Examinations A Pilot Study Sjogren et al. [15]	J. Ultrasound Med. 2016 Original	20 cross-sectional abdominal US videos (FAST)	None	Test the feasibility of automating the detection of abdominal free fluid in focused assessment with sonography for trauma (FAST) examinations.	ML: SVM	The sensitivity and specificity (95% confidence interval) were 100% (69.2–100%) and 90.0% (55.5–99.8%), respectively.	Sjogren AR, et al. “Image Segmentation and Machine Learning for Detection of Abdominal Free Fluid in Focused Assessment With Sonography for Trauma Examinations: A Pilot Study.” Journal of ultrasound in medicine: official journal of the American Institute of Ultrasound in Medicine vol. 35,11 (2016): 2501–2509. doi:10.7863/ultra.15.11017.
Quantitative Assessment of Abdominal Aortic Aneurysm Geometry Shum et al. [16]	Ann. Biomed. Eng. 2011 Original	76 CTs of patients with aneurysms	None	Test the feasibility that aneurysm morphology and wall thickness are more predictive of rupture risk and can be the deciding factors in the clinical management.	ML: Decision Tree	The model correctly classified 65 datasets and had an average prediction accuracy of 86.6% (κ = 0.37).	Shum J, et al. “Quantitative assessment of abdominal aortic aneurysm geometry.” Annals of biomedical engineering vol. 39,1 (2011): 277–286. doi:10.1007/s10439-010-0175-3.
Detection and Diagnosis of Colitis on Computed Tomography Using Deep Convolutional Neural Networks Liu et al. [17]	Med Phys. 2017 Original	CT images of 80 patients with colitis	None	Develop deep convolutional neural networks methods for lesion-level colitis detection and a support vector machine (SVM) classifier for patient-level colitis diagnosis on routine abdominal CT scans.	Faster Region-based Convolutional Neural Network (Faster RCNN) with ZF net	For patient-level colitis diagnosis, with ZF net, the average areas under the ROC curve (AUC) were 0.978 ± 0.009 and 0.984 ± 0.008 for RCNN and Faster RCNN method, respectively.	Liu J, et al. Detection and diagnosis of colitis on computed tomography using deep convolutional neural networks. Med Phys 2017;44(9):4630–4642.
Convolutional-neural-network-based diagnosis of appendicitis via CT scans in patients with acute abdominal pain presenting in the emergency department Park et al. [18]	Scientific Reports 2020 Original	667 CT image sets from 215 patients with acute appendicitis and 452 patients with a normal appendix	Data augmentation to prevent over-fitting	Test feasibility of a neural network-based diagnosis algorithm of appendicitis by using CT for patients with acute abdominal pain visiting the emergency room (ER).	Deep CNN	Diagnostic performance was excellent inboth the internal and external validation with an accuracy larger than 90%.	Park JJ, et al. Convolutional-neural-network-based diagnosis of appendicitis via CT scans in patients with acute abdominal pain presenting in the emergency department. Sci Rep. 2020 Jun 12;10(1):9556. doi:10.1038/s41598-020-66674-7. PMID: 32533053; PMCID: PMC7293232.
Deep learning algorithms for detecting and visualizing intussusception on plain abdominal radiography in children: a retrospective multicenter study Kwon et al. [19]	Scientific Reports 2021 Original	9935 X-rays	None	Verify a deep CNN algorithm to detect intussusception in children using a human-annotated dataset of plain abdominal X-rays.	Single Shot MultiBox Detector and ResNet	The internal test values after training with two hospital datasets were 0.946 to 0.971 for the area under the receiver operating characteristic curve (AUC), 0.927 to 0.952 for the highest accuracy, and 0.764 to 0.848 for the highest Youden index.	Kwon G, et al. Deep learning algorithms for detecting and visualising intussusception on plain abdominal radiography in children: a retrospective multicenter study. Sci Rep 10, 17582 (2020). https://doi.org/10.1038/s41598-020-74653-1.
An artificial intelligence deep learning model for identification of small bowel obstruction on plain abdominal radiographs Kim et al. [20]	British Journal of Radiology 2021 Original	990 plain abdominal radiographs	None	Detect small bowel obstructions of plain abdominal X-rays.	VGG16, Densenet121, NasNetLarge, InceptionV3, and Xception	The model showed an AUC of 0.961, corresponding to sensitivity and specificity of 91 and 93%, respectively.	Kim DH, et al. “An artificial intelligence deep learning model for identification of small bowel obstruction on plain abdominal radiographs.” The British journal of radiology vol. 94,1122 (2021): 20201407. doi:10.1259/bjr.20201407.
Performance of deep learning-based algorithm for detection of ileocolic intussusception on abdominal radiographs of young children Kim et al. [19]	Scientific Reports 2019 Original	Abdominal radiographs of 681 children	Intensity normalization using z-score	Detect ileocolic intussusception on abdominal radiographs of young children.	YOLO v3	The sensitivity of the algorithm was higher compared with that of the radiologists (0.76 vs. 0.46, p = 0.013), while specificity was not different between the algorithm and the radiologists (0.96 vs. 0.92, p = 0.32).	Kim S, et al. Performance of deep learning-based algo-rithm for detection of ileocolic intussusception on abdominal radiographs of young children. Sci Rep. 2019 Dec 19;9(1):19420. doi:10.1038/s41598-019-55536-6. PMID: 31857641; PMCID: PMC6923478.
Using Machine Learning to Predict the Diagnosis, Management and Severity of Pediatric Appendicitis Marcinkevics et al. [21]	Frontiers in Pediatrics 2021 Original	430 children and adolescents	None	Detect pediatric appendicitis.	Logistic regression, random forests, and gradient boosting machines	A random forest classifier achieved areas under the precision-recall curve of 0.94, 0.92, and 0.70, respectively, for the diagnosis, management, and severity of appendicitis.	Marcinkevics R, et al. (2021). Using Machine Learning to Predict the Diagnosis, Management and Severity of Pediatric Appendicitis [Original Research]. Frontiers in Pediatrics, 9. https://doi.org/10.3389/fped.2021.662183.
Development and Validation of a Deep Neural Network for Accurate Identification of Endoscopic Images From Patients With Ulcerative Colitis and Crohn’s Disease Ruan et al. [22]	Frontiers in Medicine 2022 Original	49,154 colonoscopy images from 1772 patients	Data augmentation using operations such as horizontal flipping, vertical flipping, random cropping, random rotation, brightness adjustment, contrast adjustment, and saturation adjustment, CutMix algorithm	Detect ulcerative colitis and Crohn’s disease using endoscopic images.	ResNet50	The identification accuracy achieved by the deep learning model was superior to that of experienced endoscopists per patient (deep model vs. trainee endoscopist, 99.1% vs. 78.0% and per lesion (deep model vs. trainee endoscopist, 90.4% vs. 59.7%. While the difference between the two was lower when an experienced endoscopist was included, the deep learning still performed significantly (p < 0.001) better.	Ruan G, et al. (2022). Development and Validation of a Deep Neural Network for Accurate Identification of Endoscopic Images From Patients With Ulcerative Colitis and Crohn’s Disease [Original Research]. Frontiers in Medicine, 9. https://doi.org/10.3389/fmed.2022.854677.
A Multiscale Deep Learning Method for Quantitative Visualization of Traumatic Hemoperitoneum at CT: Assessment of Feasibility and Comparison with Subjective Categorical Estimation Dreizin et al. [23]	Radiology AI 2020 Original	CT images of 130 patients	Pixel size and Intensity range harmonization	Evaluate the feasibility of a multiscale deep learning algorithm for quantitative visualization and measurement of traumatic hemoperitoneum and compare diagnostic performance for relevant outcomes with categorical estimation.	MSAN TensorFlow	AUCs for automated volume measurement and categorical estimation were 0.86 and 0.77, respectively (p = 0.004). An optimal cutoff of 278.9 mL yielded accuracy of 84%, sensitivity of 82%, specificity of 93%, positive predictive value of 86%, and negative predictive value of 83%.	Dreizin D, et al. “A Multiscale Deep Learning Method for Quantitative Visualization of Traumatic Hemoperitoneum at CT: Assessment of Feasibility and Comparison with Subjective Categorical Estimation.” Radiology. Artificial intelligence vol. 2,6 e190220. 11 Nov. 2020, doi:10.1148/ryai.2020190220.
A scalable physician-level deep learning algorithm detects universal trauma on pelvic radiographs Cheng et al. [24]	Nat. Comm. 2021 Original	5204 pelvic radiographs	Zero-padding and resizing, Data augmentation such as translation, flipping, scaling, rotation, brightness and contrast	Detect most types of trauma-related radiographic findings on pelvic radiographs.	PelviXNet	PelviXNet yielded an area under the receiver operating characteristic curve (AUROC) of 0.973 (95% CI, 0.960–0.983) and an area under the precision-recall curve (AUPRC) of 0.963 (95% CI, 0.948–0.974) in the clinical population test set of 1888 PXRs. The accuracy, sensitivity, and specificity at the cutoff value were 0.924 (95% CI, 0.912–0.936), 0.908 (95% CI, 0.885–0.908), and 0.932 (95% CI, 0.919–0.946), respectively.	Cheng CT, et al. A scalable physician-level deep learning algorithm detects universal trauma on pelvic radiographs. Nat Commun 12, 1066 (2021). https://doi.org/10.1038/s41467-021-21311-3.
Automated Screening for Abdominal Aortic Aneurysm in CT Scans under Clinical Conditions Using Deep Learning Golla et al. [25]	Diagnostics (Basel) 2021 Original	187 heterogenous CT scans.	Pixel size and Intensity range harmonization, Data augmentation	Develop and validate an easily trainable and fully automated deep learning 3D AAA screening algorithm, which can run as a background process in the clinic workflow.	ResNet, VGG-16 and AlexNet	The 3D ResNet outperformed both other networks and achieved an accuracy of 0.953 and an AUC of 0.971 on the validation dataset.	Golla AK, et al. “Automated Screening for Abdominal Aortic Aneurysm in CT Scans under Clinical Conditions Using Deep Learning.” Diagnostics (Basel, Switzerland) vol. 11,11 2131. 17 Nov. 2021, doi:10.3390/diagnostics11112131.