a b s t r a c t

Introduction: Our objective was to measure the diagnostic accuracy of a novel software technology to detect pneumothorax on Brightness (B) mode and Motion (M) mode ultrasonography.

Methods: Ultrasonography fellowship-trained emergency physicians performed thoracic ultrasonography at baseline and after surgically creating a pneumothorax in eight intubated, spontaneously breathing porcine sub- jects. Prior to pneumothorax induction, we captured sagittal M-mode still images and B-mode videos of each in- tercostal space with a linear array transducer at 4 cm of depth. After collection of baseline images, we placed a chest tube, injected air into the pleural space in 250 mL increments, and repeated the ultrasonography for pneu- mothorax volumes of 250 mL, 500 mL, 750 mL, and 1000 mL. We confirmed pneumothorax with intrapleural dig- ital manometry and ultrasound by expert sonographers. We exported collected images for interpretation by the software. We treated each individual scan as a single test for interpretation by the software.

Results: Excluding indeterminate results, we collected 338 M-mode images for which the software demonstrated a sensitivity of 98% (95% confidence interval [CI] 92-99%), specificity of 95% (95% CI 86-99), positive likelihood ratio (LR+) of 21.6 (95% CI 7.1-65), and negative likelihood ratio of 0.02 (95% CI 0.008-0.046). Among 364 B-mode videos, the software demonstrated a sensitivity of 86% (95% CI 81-90%), specificity of 85% (81-

91%), LR+ of 5.7 (95% CI 3.2-10.2), and LR- of 0.17 (95% CI 0.12-0.22).

Conclusions: This novel technology has potential as a useful adjunct to diagnose pneumothorax on thoracic ultrasonography.

1. Introduction
   1. Background

bedside thoracic ultrasonography is a valuable tool for the diagnosis of pneumothorax but is prone to operator error [1]. Ultrasonography demonstrates higher diagnostic accuracy for pneumothorax when com- pared to physical examination and supine chest radiography [2,3]. How- ever, skilled operators may not be routinely available to provide training and quality assurance. There is potential for computer technology to aid

* Corresponding author at: 3551 Roger Brooke Dr., Department of Emergency Medicine, San Antonio Uniformed Services Health Education Consortium, Joint Base San Antonio, TX Study objectives“>78234, USA

E-mail address: [email protected] (M.D. April).

¹ Disclaimer: The view(s) expressed herein are those of the author(s) and do not reflect the official policy or position of Brooke Army Medical Center, the U.S. Army Medical Department, the U.S. Army Office of the Surgeon General, the Department of the Army, the Department of Defense or the U.S. Government.

in the diagnosis of this potentially life threatening condition in the ab- sence of expert sonographers such as the prehospital setting.

To assist the novice ultrasonographer, biomedical engineers devel- oped the Intelligent Focused assessment with sonography in trauma (iFAST) software algorithm [4]. The iFAST is a computerized diagnostic assistant designed to automatically detect pneumothorax on standard ultrasonography imagery. This software systematically analyzes Bright- ness (B) mode video clips and Motion (M) mode still images for the presence of sliding lung and seashore signs, respectively [5]. In a recent study, the iFAST demonstrated a sensitivity of 79% and a specificity of 87% to detect pneumothorax as compared to expert physician sonographers [6]. Limitations of this study included retrospective de- sign and lack of definitive reference standard for diagnosis.

Study objectives

The primary objective of this investigation was to estimate the diag- nostic accuracy of the iFAST computer algorithm to detect pneumothorax

http://dx.doi.org/10.1016/j.ajem.2017.03.073 0735-6757/

on M-mode and B-mode ultrasonography imagery. The secondary objec- tive was to assess the association between algorithm sensitivity and pneumothorax volume.

1. Methods
   1. Study design and setting

We conducted an experimental study assessing the accuracy of computerized interpretation of thorax ultrasound in a porcine pneu- mothorax model. We conducted the study as part of a parallel inves- tigation of chest seal placement in an animal vivarium. We complied with the regulations and guidelines of the Animal Welfare Act and the American Association for Accreditation of Laboratory Animal Care. The Institutional Animal Care and Use Committee approved the investigation.

Animal subjects

The veterinary service prepared eight female crossbred Yorkshire swine (Sus scrofa) in similar fashion to a prior study evaluating chest seals in pneumothorax [7]. Swine weighing 35 to 45 kg underwent se- dation with a mixture of Intramuscular ketamine, tiletamine- zolazepam, and midazolam. After induction with 2-3% isoflurane in 30% oxygen, technicians intubated the animals with a cuffed 7.5 mm en- dotracheal tube and set the ventilator to assist-control mode, tidal vol- ume (Vt) 8 mL/kg, respiratory rate (RR) 12 to 16 breaths/min, and fraction of inspired oxygen (FiO2) 21 to 30%.

Veterinary technicians then established carotid and femoral Arterial lines (Millar Instruments, Houston, TX) and an internal jugular vein Swan Ganz catheter (Edwards Lifesciences, Irvine, CA) for continuous hemodynamic monitoring as part of the parallel investigation. We de- livered isotonic crystalloid at a maintenance rate and titrated to urine output. After instrumentation, we discontinued the isoflurane and pro- vided sedation and analgesia with propofol at 3-6 mg/kg/h and buprenorphine at 2-8 ug/kg/h. The animals were supine and spontane- ously breathing (transitioned to pressure support ventilation) for all ul- trasonography examinations.

Study procedures

Two ultrasonography fellowship-trained emergency physicians pro- spectively collected sagittal thoracic ultrasonography images using a Sonosite M-Turbo (Bothell, WA) equipped with a linear array transduc- er (5 to 10 MHz). We collected images at baseline and after surgically creating a pneumothorax in eight intubated, spontaneously breathing porcine subjects. We induced only one pneumothorax in each subject to avoid Hemodynamic compromise. We arbitrarily chose the left chest in all subjects to maintain consistency. We recorded one M- mode still image along with a six second B-mode video clip of each in- tercostal space of the left chest, all in the sagittal plane, beginning with the second intercostal space in the mid-clavicular line and continu- ing in an inferior-lateral direction until we visualized the diaphragm. If we visualized the heart or lung pulse, we moved the probe laterally until these findings were no longer present. There were no instances in which we were unable to capture these measurements in each inter- costal space.

After we obtained baseline images, a research physiologist inserted a 14-gauge pleural catheter into the left sixth intercostal space in the mid- axillary line. We then attached this catheter to a three-way stopcock connected to a digital manometer (NETECH, Digimano, Farmingdale, NY), injected 250 mL of air in the pleural space, and confirmed a non- negative resting intrapleural pressure to establish the presence of pneu- mothorax. We always confirmed presence of pneumothorax through identification (by one of the two ultrasonography fellowship-trained emergency physicians) of absent lung sliding, barcode sign, and lung

point. The Lung point location varied according to pneumothorax volume; we visualized the lung point at higher intercostal spaces with lower pneu- mothorax volumes. Following confirmation of pneumothorax, we repeat- ed the left-sided Ultrasonography examinations beginning with the second intercostal space and continuing in an inferior-lateral direction until visualization of the lung point. Because the sonographic lung point defines the terminal edge of the pneumothorax, we did not collect images at intercostal spaces below this level [5].

We continued with 250 mL incremental injections of air into the pleural space and repeated the ultrasonography examinations for pneu- mothorax volumes of 500 mL, 750 mL, and 1000 mL. Finally, we exported all images from the ultrasonography machine and arranged them in random order on a compact disc for iFAST interpretation. The biomedical engineer, blinded to the diagnosis and the timing of the ul- trasonography examinations, applied the iFAST alogorithm to each image and recorded the software interpretation on a standardized data collection instrument. For each image, the iFAST rendered an inter- pretation of positive, negative, or indeterminate for pneumothorax.

We have described the iFAST computer algorithm in detail previous- ly [4,6]. Briefly, in M-mode the iFAST identifies the pleural line as the most hyperechoic horizontal line on the image and then analyzes below that line for pixel granularity or “bar code” pattern (Fig. 1). The iFAST will report negative for pneumothorax if there is a significant pro- portion of sub-pleural speckling in M-mode resembling a “sandy beach.” In B-mode, the iFAST first identifies the rib shadows to locate the pleural line. Once identified, the iFAST dynamically scans video clips at 30 frames per second for pixel movement along the pleural line as well reverberation artifacts extending below (Fig. 2). The iFAST reports negative for pneumothorax if it identifies the presence of sliding lung. When the iFAST is unable to identify a pleural line, it reports an in- determinate result. The algorithm remained fixed during this study.

Outcomes

The primary outcome was the diagnostic accuracy of the iFAST com- puter algorithm interpretation of M-mode and B-mode scans, consid- ered independently. As we did not have access to computed tomography, the reference standard was presence of surgically-induced pneumothorax as confirmed by intrapleural pressure monitor [7-9]. One of the two ultrasonography fellowship-trained emergency physi- cians also confirmed presence of pneumothorax by identification of all three of the following: absent lung sliding, barcode sign, and lung point.

Analysis

We analyzed the data using non-parametric descriptive statistics to report intrapleural Pressure measurements stratified by lung volume (Fig. A.1). We calculated iFAST Diagnostic test characteristics separately for M-mode and B-mode. Characteristics calculated included sensitivity, specificity, positive likelihood ratio (LR+) and negative likelihood ratio (LR-). We further stratified these calculations by pneumothorax vol- ume. We excluded indeterminate results from these calculations. We also repeated these calculations assuming all indeterminate results were inaccurate to generate conservative estimates of the algorithm’s diagnostic accuracy [10]. We calculated confidence intervals using jack- knife resampling methods [11].

1. Results
   1. Main results

We collected 343 M-mode and 364 B-mode thoracic ultrasonogra- phy images for interpretation by the iFAST (Table 1). We excluded three (0.9%) M-mode images and two (0.5%) B-mode images which were irretrievable from the Sonosite hard drive. The iFAST was indeter- minate for 2 (0.6%) M-mode images and 14 (3.8%) B-mode, leaving

Fig. 1. For the M-mode image on the left, the Intelligent Focused Assessment with Sonography in Trauma (iFAST) software identifies “sandy” granular appearance (yellow) bellow the pleural line (neon green) and reports negative for pneumothorax. For the image on the right, the iFAST identifies “bar code” appearance with less granularity and reports positive for pneumothorax.

338 M-mode and 348 B-mode for the primary analysis (Figs. 3-4). In M- mode, the iFAST demonstrated a high diagnostic accuracy with a sensi- tivity of 98% and specificity of 95%. In B-mode, the iFAST demonstrated more modest diagnostic accuracy with a sensitivity of 86% and specific- ity of 85% (Table 2).

For the secondary analysis, we examined the diagnostic perfor- mance of the iFAST at increasing volumes of pneumothorax (Fig. 5). There were no marked differences in False negative rates between vol- umes of 250 mL, 500 mL, 750 mL, and 1000 mL. We observed a trend to- wards increased false negative results with increasing pneumothorax volume.

Sensitivity analyses

For M-mode, if we make the assumption that the 2 indeterminate re- sults reported after surgical creation of pneumothorax were false nega- tives, the test characteristics would be the following: sensitivity 97% (95% CI 95-99), specificity 95% (95% CI 86-99), LR+ 21.4 (95% CI 7.1-

65), and LR- 0.026 (95% CI 0.013-0.055). For B-mode there were 14 in-

determinate results. If we assume that all indeterminate results were in- accurate (i.e., the 7 indeterminate results reported at baseline were false positives and the 7 reported after induction of pneumothorax were false negatives) then the test characteristics would be the following: sensitiv- ity 84% (95% CI 79-88), specificity 77% (95% CI 66-86), LR+ 3.64 (95% CI 2.39-5.55), and LR- 0.21 (95% CI 0.16-0.28).

Fig. 2. In this negative B-mode image, the Intelligent Focused Assessment with Sonography in Trauma (iFAST) identifies rib shadows (blue rectangles), pixel movements along the pleural line (red dots), and reverberations extending below (red vertical lines) consistent with normal sliding lung.

1. Discussion
   1. Overview

With advances in technology, computer-aided diagnostic sys- tems have become increasingly common over the past two decades [12-14]. Software algorithms now exist to assist physicians with au- tomated detection of diseases such as coronary artery disease, acute myocardial infarction, arrhythmias, pulmonary edema, retinopathy, thyroid disorders, Alzheimer’s dementia, Pulmonary nodules, colon- ic polyposis, and various malignancies [15-26]. To our knowledge, the iFAST is the first to analyze thoracic ultrasonography for sono- graphic signs of pneumothorax, a condition of particular relevance to emergency physicians. Although this technology is not a substi- tute for the experienced emergency physician, there is a potential for the iFAST to supplement the medical decision making for novice users of bedside ultrasonography. There is also potential for impact in the prehospital setting, austere environments, and as a training aid for residency programs.

For the primary analysis, our results demonstrate that iFAST is highly accurate when interpreting good quality M-mode images obtained in a laboratory setting. B-mode was moderately accurate in our study but significantly outperformed by M-mode. We believe M-mode was supe- rior because it provides a graphical representation of pleural movement over time which may better lend itself to computer analysis. Our results support M-mode as the optimum choice for future deployment of the iFAST unless significant enhancements can be made to the B-mode algo- rithm. One caveat is that M-mode requires placement of the scan line di- rectly over the region of the interest which adds an additional step for the operator.

Much like with other computer-aided diagnosis and detection sys- tems, we envision that the iFAST will serve to guide the operator to areas of interest in the ultrasonography image and alert to the possibil- ity of pneumothorax. The operator will then apply the iFAST interpreta- tion to the Clinical context and make a determination if immediate

Table 1

3 x 2 contingency table including indeterminate results for Intelligent Focused Assess- ment with Sonography in Trauma (iFAST).

iFAST result	Pneumothorax absent	Pneumothorax present
M-mode (n = 340)
Positive	3	267
Indeterminate	0	2
Negative	63	5
B-mode (n = 362)
Positive	10	241
Indeterminate	7	7
Negative	57	40

Fig. 3. Flowchart for collection of M-mode images with Intelligent Focused Assessment with Sonography in Trauma (iFAST) software interpretation of each image.

treatment or Confirmatory imaging is necessary. For future directions, the ideal algorithm would also provide an assessment of diagnostic cer- tainty based on quality of the image, the proportion of far-field granu- larity in M-mode, and the degree of pixel movement along the pleural line in B-mode.

For the secondary analysis, the iFAST demonstrated no significant dif- ferences in sensitivity for higher volumes of pneumothorax, but there was a trend towards increased false negative results with higher volumes. We hypothesize that this was related to the tachypnea of the animal. As the respiratory rate and Work of breathing increased, intercostal muscle

Fig. 4. Flowchart for collection of B-mode images with Intelligent Focused Assessment with Sonography in Trauma (iFAST) software interpretation of each image.

Table 2 Test characteristics of Intelligent Focused Assessment with Sonography in Trauma (iFAST) in M-mode and B-mode for the detection of pneumothorax.

Limitations

Our use of images by fellowship-trained ultrasonographers is an im-

Test characteristics

iFAST in M-mode (95% CI)

iFAST in B-mode (95% CI)

portant limitation. Because novice sonographers rarely acquire “standard”

images in clinical practice, we cannot draw conclusions on the diagnostic

Sensitivity, % 98 (92-99) 86 (81-90)

Specificity, % 95 (86-99) 85 (81-91)

LR+ 21.6 (7.1-65) 5.7 (3.2-10.2)

LR- 0.02 (0.008-0.046) 0.17 (0.12-0.22)

Abbreviations: LR-likelihood ratio.

retraction may have led to artifact mimicking motion along the pleu- ral line. Thus, the iFAST algorithm may require further refinement to optimize the signal to noise ratio to account for times of respiratory distress.

Our study suggests that iFAST is highly accurate for diagnosing pneumothorax. Our study design provides more ultrasound results in the presence of pneumothorax and so provides more data on sensitivity than specificity. This is of particular value from the perspective of seeking to rule out pneumothorax pathology. In M- mode, we report a higher sensitivity (98% vs 91%) and slightly lower specificity (95% vs 98%) when compared to lung ultrasonog- raphy performed by trained physician sonographers as reported by the meta-analysis by Alrajhi et al. [27]. Similarly, we report a higher sensitivity and specificity than that reported in a previous in- vestigation of this software using images obtained by resident sonographers [6]. The higher diagnostic accuracy values in our stud- ies may reflect the fact that fellowship-trained ultrasonographers obtained the images.

performance of iFAST in real world situations among providers with vary- ing levels of ultrasonography experience and skill. We suspect that the iFAST would render more indeterminate results for novice sonographers. A related point is that the iFAST algorithm assists only with identification of pneumothorax and provides no diagnostic information regarding other potential findings from ultrasound of the lung or elsewhere. For these rea- sons, we doubt that a perfected iFAST algorithm will ever completely ob- viate the need for some basic ultrasonography training.

Our study also did not assess the performance of iFAST in human subjects. That said, we believe a porcine model is similar to human sub- jects in terms of sonographic signs of pneumothorax [28-32].

Another limitation is our assumption that images obtained from dif- ferent intercostal spaces in the same animal were independent observa- tions. We believe this a reasonable assumption because each intercostal space renders a unique and distinct ultrasonography image in terms of depth of the pleural line, number and quality of reverberation artifacts, gain, and the proportion of granularity in M-mode. However, this ap- proach does not reflect the reality of bedside ultrasonography in which providers commonly scan multiple intercostal spaces to inform their determination of whether a pneumothorax is present. Our deci- sion to instead study individual intercostal spaces in isolation may therefore underestimate the potential accuracy of iFAST.

Given this final limitation, future research may benefit from examin- ing alternative iFAST computer algorithms incorporating data from multiple intercostal spaces to make an overall determination regarding the presence or absence of pneumothorax. Such integration might also

Fig. 5. Performance of Intelligent Focused Assessment with Sonography in Trauma (iFAST) software at baseline and pneumothorax (PTX) volumes of 250 mL, 500 mL, 750 mL, and 1000 mL.

link data from M-mode images to B-mode images to further increase di- agnostic accuracy. Such refinements to the iFAST algorithms would then ideally lead a prospective study in trauma patients among clinicians with varying levels of ultrasonography training.

Conclusions

The iFAST software technology has potential to be a useful tool to as- sist with diagnosis of pneumothorax on bedside thoracic ultrasonogra- phy. The iFAST demonstrated high sensitivity and specificity when interpreting M-mode images obtained by experienced sonographers in a porcine model. Our estimates of sensitivity did not significantly change with higher volumes of pneumothorax. The literature would benefit from further refinement of this technology to optimize diagnos- tic accuracy followed by prospective evaluation in human subjects to as- sess the applicability and accuracy of this technology in clinical practice.

Funding

This work was supported by the United States Army Medical Re- search Materiel Command (MRMC) [grant number #H13017].

Conflicts of interest

SMS, EJC, RDG, JS, and LHB are co-inventors of the iFAST software. We have filed a provisional patent for the technology PCT/US2014/ 058374. However, as active duty military personal and employees of the Department of Defense, we have assigned to the Government of The United States, as represented by The Secretary of The Army, the en- tire right, title, and interest of the invention. We have received no royal- ties for the invention. MDA, JAL, and BSK have no conflicts of interest to disclose.

Appendix A

Fig. A.1Box-and-whiskers plot of mean intrapleural pressure mea- surements for 8 swine at baseline and after incremental injection of air into the pleural space.

References

1. Ding W, Shen Y, Yang J, He X, Zhang M. Diagnosis of pneumothorax by radiography and ultrasonography: a meta-analysis. Chest 2011;140:859-66.
2. Kong VY, Sartorius B, Clarke DL. The accuracy of physical examination in identifying significant pathologies in penetrating Thoracic trauma. Eur J Trauma Emerg Surg 2015;41:647-50.
3. Ebrahimi A, Yousefifard M, Mohammad Kazemi H, et al. Diagnostic accuracy of chest ultrasonography versus chest radiography for identification of pneumothorax: a sys- tematic review and meta-analysis. Tanaffos 2014;13:29-40.
4. Blackbourne L, Chin E, Grisell R, Salinas J, Summers S. Automatic focused assessment with sonography for trauma exams. Google patents; 2016.
5. Volpicelli G, Elbarbary M, Blaivas M, et al. International evidence-based recommen- dations for point-of-care Lung ultrasound. Intensive Care Med 2012;38:577-91.
6. Summers SM, Chin EJ, Long BJ, et al. Computerized diagnostic assistant for the auto- matic detection of pneumothorax on ultrasound: a pilot study. West J Emerg Med 2016;17:209-15.
7. Kheirabadi BS, Terrazas IB, Koller A, et al. Vented versus unvented chest seals for treatment of pneumothorax and prevention of Tension pneumothorax in a swine model. J Trauma Acute Care Surg 2013;75:150-6.
8. Arnaud F, Maudlin-Jeronimo E, Higgins A, et al. Adherence evaluation of vented chest seals in a swine skin model. Injury 2016;47:2097-104.
9. Barton ED, Rhee P, Hutton KC, Rosen P. The pathophysiology of tension pneumotho- rax in ventilated swine. J Emerg Med 1997;15:147-53.
10. Shinkins B, Thompson M, Mallett S, Perera R. Diagnostic accuracy studies: how to re- port and analyse inconclusive test results. BMJ 2013;346:f2778.
11. Efron B, Stein C. The jackknife estimate of variance. Ann Stat 1981;9:586-96.
12. Petrick N, Sahiner B, Armato 3rd SG, et al. Evaluation of computer-aided detection and diagnosis systems. Med Phys 2013;40:087001.
13. Giger ML, Chan HP, Boone J. Anniversary paper: history and status of CAD and quan- titative image analysis: the role of medical physics and AAPM. Med Phys 2008;35: 5799-820.
14. Ghobadi CW, Hayman EL, Finkle JH, Walter JR, Xu S. Radiological medical device in- novation: approvals via the premarket approval pathway from 2000 to 2015. J Am Coll Radiol 2017;14:24-33.
15. Faust O, Acharya UR, Sudarshan VK, et al. Computer aided diagnosis of coronary ar- tery disease, myocardial infarction and carotid atherosclerosis using ultrasound im- ages: a review. Phys Med 2016.
16. Hashoul S, Gaspar T, Halon DA, et al. Automated computer-assisted diagnosis of ob- structive coronary artery disease in emergency department patients undergoing 256-slice coronary computed tomography angiography for acute chest pain. Am J Cardiol 2015;116:1017-21.
17. Al-Zaiti SS, Callaway CW, Kozik TM, Carey MG, Pelter MM. Clinical utility of ventric- ular repolarization dispersion for real-time detection of non-ST elevation myocardial infarction in emergency departments. J Am Heart Assoc 2015;4.
18. Brattain LJ, Telfer BA, Liteplo AS, Noble VE. Automated B-line scoring on thoracic so- nography. J Ultrasound Med 2013;32:2185-90.
19. Corradi F, Brusasco C, Vezzani A, et al. Computer-aided quantitative ultrasonography for detection of pulmonary edema in mechanically ventilated cardiac surgery pa- tients. Chest 2016;150:640-51.
20. ElTanboly A, Ismail M, Shalaby A, et al. A computer aided diagnostic system for de- tecting diabetic retinopathy in optical coherence tomography images. Med Phys 2016.
21. Rhee SJ, Hong HS, Kim CH, Lee EH, Cha JG, Jeong SH. Using acoustic structure quan- tification during B-mode sonography for evaluation of Hashimoto thyroiditis. J Ultra- sound Med 2015;34:2237-43.
22. Beheshti I, Maikusa N, Matsuda H, Demirel H, Anbarjafari G. Histogram-based fea- ture extraction from individual gray matter similarity-matrix for Alzheimer’s disease classification. J Alzheimers Dis 2017;55:1571-82.
23. Perandini S, Soardi GA, Motton M, et al. Enhanced characterization of solid solitary pulmonary nodules with Bayesian analysis-based computer-aided diagnosis. World J Radiol 2016;8:729-34.
24. G k, R c. Automatic colorectal polyp detection in colonoscopy video frames. Asian Pac J Cancer Prev 2016;17:4869-73.
25. Saha M, Mukherjee R, Chakraborty C. Computer-aided diagnosis of breast cancer using cytological images: a systematic review. Tissue Cell 2016;48:461-74.
26. Liu L, Tian Z, Zhang Z, Fei B. Computer-aided detection of prostate cancer with MRI: technology and applications. Acad Radiol 2016;23:1024-46.
27. Alrajhi K, Woo MY, Vaillancourt C. Test characteristics of ultrasonography for the de- tection of pneumothorax: a systematic review and meta-analysis. Chest 2012;141: 703-8.
28. Oveland NP, Lossius HM, Wemmelund K, Stokkeland PJ, Knudsen L, Sloth E. Using thoracic ultrasonography to accurately assess pneumothorax progression during positive pressure ventilation: a comparison with CT scanning. Chest 2013;143: 415-22.
29. Oveland NP, Sloth E, Andersen G, Lossius HM. A porcine pneumothorax model for teaching ultrasound diagnostics. Acad Emerg Med 2012;19:586-92.
30. Oveland NP, Soreide E, Lossius HM, et al. The intrapleural volume threshold for ultra- sound detection of pneumothoraces: an experimental study on Porcine models. Scand J Trauma Resusc Emerg Med 2013;21:11.
31. Sanchez-de-Toledo J, Renter-Valdovinos L, Esteves M, Fonseca C, Villaverde I, Rosal

M. Teaching chest ultrasound in an experimental porcine model. Pediatr Emerg Care 2016;32:768-72.

1. Bloch AJ, Bloch SA, Secreti L, Prasad NH. A porcine Training model for ultrasound di- agnosis of pneumothoraces. J Emerg Med 2011;41:176-81.