## Abstract

The coronavirus disease of 2019 (COVID-19) has altered medical practice around the globe and revealed critical deficiencies in hospital supply chains ranging from adequate personal protective equipment to life-sustaining ventilators for critically ill hospitalized patients. We developed the CRISIS ventilator, a gas-powered resuscitator that functions without electricity, and which can be manufactured using hobby-level three-dimensional (3D) printers and standard off-the-shelf equipment available at the local hardware store. CRISIS ventilators were printed and used to ventilate sedated female Yorkshire pigs over 24-h. Pulmonary and hemodynamic values were recorded throughout the 24-h run, and serial arterial blood samples were obtained to assess ventilation and oxygenation. Lung tissue was obtained from each pig to evaluate for signs of inflammatory stress. All five female Yorkshire pigs survived the 24-h study period without suffering from hypoxemia, hypercarbia, or severe hypotension requiring intervention. One animal required rescue at the beginning of the experiment with a traditional ventilator due to leakage around a defective tracheostomy balloon. The wet/dry ratio was 6.74 ± 0.19 compared to historical controls of 7.1 ± 4.2 (not significantly different). This proof-of-concept study demonstrates that our 3D-printed CRISIS ventilator can ventilate and oxygenate a porcine model over the course of 24-h with stable pulmonary and hemodynamic function with similar levels of ventilation-related inflammation when compared with a previous control porcine model. Our work suggests that virtual stockpiling with just-in-time 3D-printed equipment, like the CRISIS ventilator, can temporize shortages of critical infrastructure needed to sustain life for hospitalized patients.

## Introduction

COVID-19 is an ongoing worldwide respiratory pandemic caused by the SARS-CoV-2 virus complicated by ongoing infectious surges threatening hospital capacity and prompting crisis standards of care in hospital systems across the United States [1]. Indeed, ventilator shortages driven by an increase in hospitalized patients requiring ventilator support necessitated mobilization of strategic stockpiles early in the pandemic and prompted partnerships with automotive companies to manufacture these life-sustaining devices for patients with respiratory failure [24]. The difficulties experienced with mobilizing the national strategic stockpile of ventilators revealed antiquated equipment with missing parts while retooling automotive companies for the manufacture of medical equipment illustrated the need for faster and nimbler just-in-time capabilities [5,6]. State-to-state and country-to-country transfers of ventilators to balance supply and demand were problematic given dynamic demands, logistical challenges with the shipment, and equipment quality assurance amongst other concerns [7].

In response to early supply shortages during COVID-19, the CRISIS ventilator is a three-dimensional (3D)-printed device designed to function as a traditional ventilator for critically ill patients with respiratory failure. The components of the CRISIS ventilator use the pressure head of the gas supply to operate rather than electricity, making it suitable for applications in austere environments where electricity may be scarce or intermittent [8]. Additionally, the CRISIS ventilator is manufactured using hobbyist-level 3D printers and assembled with standard off-the-shelf components (i.e., machine screws), enabling virtual stockpiling of the device for organizations or individuals who have access to hobbyist-level 3D printers and a small collection of standard parts.

While our group has previously published on the performance of the device compared to existing technologies using in vitro test models, we sought to validate the use of the CRISIS ventilator in vivo using a swine model [9]. Briefly, the CRISIS ventilator uses an adjustable huddling chamber which comprises a static mixer and adjustable size air outlet to create a stable breathing pattern for a wide arrange of patients and pressures. By adjusting the pressure adjustment dial on the top of the device, peak inspiratory pressure (PIP) is adjusted. PIP is mechanically linked to positive end expiratory pressure (PEEP) in our device. By iterating the design of the huddling chamber, physiologically desired PEEP to PIP ratios were achieved. A closed-loop spline calibration system was used to linearize the output range of the CRISIS ventilator. Briefly, the spiral cam pathway that controls the orifice size of the air outlet enables minute adjustments to the pressure outlet curve and the calibration system ensures a linear adjustment in pressure outputs for the end-user. The calibration data generated for each valve is fed into a parameterized spline model in Autodesk Inventor (Autodesk, Inc., San Rafael, CA), enabling consistent pressure output curves for each valve and consistent across different valves.

One of the key benefits of ventilators like the CRISIS ventilator and Go2Vent (Vortaran, Sacramento, CA) is the fact that they do not require electricity; however, one important consideration is that the mechanical work of the ventilator is coupled with the pressure head of the gas, thereby using more gas per unit time when compared with a traditional ventilator [10]. Additionally, these pressure-actuated ventilators need to be placed as close as possible to the patient in the respiratory circuit to minimize dead space; this is to aid with ventilation and to minimize CO2 retention. Traditional ventilators have CO2 removal systems and circuit architecture which obviates these problems but rely on electricity, ongoing regular service, and are in short supply given the ongoing respiratory pandemic [11].

We hypothesized that our 3D-printed CRISIS ventilator could support life in a swine model undergoing prolonged ventilation.

## Methods

Five 3D-printed CRISIS ventilators were created using polyethylene terephthalate on a hobbyist-level extrusion printer (Prusa MK3, Prusa Research, Prague, CZ). There were no significant postprocessing modifications and the semi-automated spline calibration system was used to print the spiral cam component with linearized pressure output curves. The ventilator was assembled using a stainless-steel type 316 spring (Lee Spring, Brookyln, NY) and a drag-cut shore D-50 0.1-mm silicone membrane (Jiawanshun, Dancheng, China) and installed using standard ventilator circuit tubing (Fig. 1). Each calibrated CRISIS ventilator took less than an hour to assemble with a five-hour print time.

Fig. 1
Fig. 1
Close modal

Five female Yorkshire pigs (weight 57 ± 12 kg average ± standard deviation; range 49–78 kg) underwent induction of anesthesia using inhaled isoflurane followed by intubation via tracheostomy to form a secure airway. The internal carotid artery and external jugular vein were cannulated for blood gas sampling and administration of fluids. The animal was transitioned to intravenous sedation using a combination of propofol, midazolam, and fentanyl to mimic the sedation package of a human patient. An open suprapubic Foley catheter was then inserted. Baseline arterial blood gases were collected after the complete transition from inhaled anesthetic to intravenous sedation, approximately 1 h. The animal was then connected to the CRISIS ventilator.

Tidal volume, PIP, respiratory rate, and PEEP were all adjusted initially utilizing an in-line monitor via the Fluke VT650 gas analyzer (Fluke Biomedical, Cleveland, OH). High-resolution airway pressure, flow, and volume curves were recorded every 6 h and with any changes in ventilator settings. Arterial blood gases were collected every 6 h and with any changes in the ventilator settings. End-tidal carbon dioxide, heart rate, invasive blood pressure, and urine output were recorded. Due to the known effects of this intravenous sedation regimen causing bradycardia in this animal model, any animal that did develop bradycardia was treated with intravenous atropine [12].

After 24 h, the animals were euthanized, and the anterior right lower lobe lungs were biopsied to determine the wet/dry ratios as a surrogate marker of inflammation and lung injury [13]. Biopsies were obtained from each of the six lobes for histologic preparation. To compare the efficacy of the CRISIS ventilator against traditional ventilators, we curated data from previous studies in our lab where animals were cannulated and anesthetized using the same fluid and cannulation protocol though anesthetized using a combination of inhaled isoflurane and intravenous ketamine while ventilated on a Dräger anesthesia machine (Drägerwerk AG & Co, Lübeck Germany). While these animals were ventilated for 48 h before euthanasia, the first 24 h of blood gases were compared alongside the lung wet/dry ratio.

This study was conducted in compliance with the institutional animal care and use committee to ensure all animals were treated humanely (STUDY# IP00003101).

## Results

All five animals in the study survived and there were no major complications with regards to hypotension or inability to sustain respiratory function with the CRISIS ventilator. The first animal in the study required rescue with a traditional ventilator due to a leak in the circuit at the endotracheal tube balloon. Once the circuit was exchanged, no further rescue ventilation was required. With animal 1 censored, the PEEP for the remaining four animals is 6.4 ± 0.3 mmHg.

Each animal required an average of two adjustments of the ventilator during the 24-h period of the study either due to leaks around the endotracheal tube or tidal volume drift. The additional dead-space ventilation required due to using the Fluke VT650 gas analyzer necessitated using higher minute ventilation.

The physiologic data of the subjects are summarized in Table 1 includes heart rate (69 ± 5 beats per minute), mean arterial pressure (78 ± 6 mmHg), respiratory rate (15 ± 1 breaths per minute), end-tidal CO2 (44 ± 1 mmHg), pH at 24 h (7.47 ± 0.03), PaCO2 at 24 h (42 ± 3 mmHg), and PaO2 at 24 h (541 ± 45 mmHg). The mechanical ventilation data is summarized in Table 2 including tidal volume (9.4 ± 0.5 mL/kg), minute ventilation (141 ± 9.2 mL/kg-min), PIP (23 ± 2.6 cm H2O), and PEEP (8.7 ± 5.1 cm H2O). The lung wet/dry ratio for the study animals was 6.74:1 ± 0.19 and not different from historical controls (7.1:1 ± 4.2 unitless). None of the animals experienced hypoxemia nor hypercarbia while all animals did require atropine for bradycardia.

Table 1

Physiologic parameters for five Yorkshire pigs in the study including heart rate, mean arterial pressure, respiratory rate, end tidal CO2, and arterial blood gas parameters at 24 h (end of the experiment) including pH, PaCO2, and PaO2

Heart rate (beats/min)Mean arterial pressure (mmHg)Respiratory rate (breaths/min)End-tidal CO2 (mmHg)pH at 24 hPaCO2 at 24 h (mmHg)PaO2 at 24 h (mmHg)
Animal 165 ± 1687 ± 1514.7 ± 2.044 ± 57.4839434
Animal 266 ± 1078 ± 1714.0 ± 1.444 ± 87.5239490
Animal 369 ± 1172 ± 816.1 ± 4.544 ± 37.4445511
Animal 470 ± 1174 ± 915.8 ± 1.545 ± 27.4648546
Animal 577 ± 981 ± 1014.6 ± 1.644 ± 27.4643541
Study Mean69 ± 578 ± 615 ± 144 ± 17.47 ± 0.0342 ± 3504 ± 45
Heart rate (beats/min)Mean arterial pressure (mmHg)Respiratory rate (breaths/min)End-tidal CO2 (mmHg)pH at 24 hPaCO2 at 24 h (mmHg)PaO2 at 24 h (mmHg)
Animal 165 ± 1687 ± 1514.7 ± 2.044 ± 57.4839434
Animal 266 ± 1078 ± 1714.0 ± 1.444 ± 87.5239490
Animal 369 ± 1172 ± 816.1 ± 4.544 ± 37.4445511
Animal 470 ± 1174 ± 915.8 ± 1.545 ± 27.4648546
Animal 577 ± 981 ± 1014.6 ± 1.644 ± 27.4643541
Study Mean69 ± 578 ± 615 ± 144 ± 17.47 ± 0.0342 ± 3504 ± 45

All data presented as mean  ±  standard deviation.

Table 2

Ventilatory parameters for five Yorkshire pigs in the study including tidal volume, minute ventilation, PIP, and PEEP

Tidal volume (mL/kg)Minute ventilation (mL/kg-min)Peak inspiratory pressure (cm H2O)Positive end-expiratory pressure (cm H2O)
Animal 19.1 ± 1.4133 ± 2126 ± 4.818 ± 4.4
Animal 210.1 ± 1.4141 ± 2025 ± 2.16.2 ± 0.7
Animal 38.9 ± 1.2143 ± 2024 ± 1.56.9 ± 0.6
Animal 49.9 ± 0.4156 ± 7.122 ± 1.16.3 ± 0.5
Animal 59.1 ± 0.4133 ± 5.119 ± 2.76.1 ± 0.4
Study Mean9.4 ± 0.5141 ± 9.223 ± 2.68.7 ± 5.1
Tidal volume (mL/kg)Minute ventilation (mL/kg-min)Peak inspiratory pressure (cm H2O)Positive end-expiratory pressure (cm H2O)
Animal 19.1 ± 1.4133 ± 2126 ± 4.818 ± 4.4
Animal 210.1 ± 1.4141 ± 2025 ± 2.16.2 ± 0.7
Animal 38.9 ± 1.2143 ± 2024 ± 1.56.9 ± 0.6
Animal 49.9 ± 0.4156 ± 7.122 ± 1.16.3 ± 0.5
Animal 59.1 ± 0.4133 ± 5.119 ± 2.76.1 ± 0.4
Study Mean9.4 ± 0.5141 ± 9.223 ± 2.68.7 ± 5.1

All data presented as mean  ±  standard deviation. Large PEEP in animal 1 multifactorial including requiring rescue ventilation and replacement of circuit before proceeding with CRISIS ventilator trial and due to the first animal on a novel ventilator. The mean PEEP for the study if animal 1 is censored is 6.4 ± 0.3.

Representative airflow volume, pressure, and flow rate curves using the CRISIS ventilator are shown in Fig. 2, which are like traditional ventilators in a pressure-regulated volume control mode. Representative histological specimens for animals from this study and historical controls are shown in Fig. 3 demonstrating normal alveolar architecture with no inflammation for lung specimens from the CRISIS ventilator study after 24 h and minimal inflammation for lung specimens from historical controls using the traditional ventilator.

Fig. 2
Fig. 2
Close modal
Fig. 3
Fig. 3
Close modal

## Discussion

The CRISIS ventilator is a 3D-printed device that can be extruded from hobby-level 3D printers and assembled using standard off-of-the-shelf equipment and used to maintain normal animal physiology in a porcine model over the course of 24 h. The CRISIS ventilator was able to maintain adequate ventilation and oxygenation for all animals. Hemodynamic and pulmonary function parameters were stable through reasonable adjustments and interventions as may be expected in routine clinical practice. Blood gas analysis was reassuring against hypercarbia and hypoxemia throughout the experiment for all animals while wet-dry ratios, a measure of pulmonary edema and potentially ventilator-induced lung injury, were similar between biopsy specimens from the CRISIS ventilator and historical controls. Histological evaluation between these two groups demonstrates the qualitative similarity between the lung tissue at the end of the CRISIS ventilator experiments and historical controls using traditional ventilation. Quantitative histological analysis of our specimens and the historical controls is ongoing and will be utilized in future analysis.

One important lesson learned from this pilot study was that the CRISIS device is notably sensitive to leaks in the circuit. Several animals developed leaks in their endotracheal cuffs which resulted in stalling of the device. It is worth noting that in the event of a stall the device is designed to provide a minimal positive continuous pressure, akin to continuous positive airway pressure, and that the inspiratory circuit would allow for a patient to continue spontaneous ventilation. Further, this ventilator is dependent on being as close as possible to the airway in a patient as any additional dead space will result in breath-recycling due to the lack of a built-in carbon dioxide scavenger system. Both have been described in other commercially available devices [14]. Of note, the ventilator itself could overcome increased dead space to a degree, as became necessary during this pilot study to obtain appropriate breath-level data. Another crucial factor in the utilization of this device is the rate of gas consumption. Due to constraints in our veterinary lab, we were only able to provide 100% oxygen with a rate of consumption of 30 liters per minute. In an ideal setting, a patient would be provided a blended gas source at a rate of 25–35 liters per minute for an adult utilizing standard adjustable flow valves found in the hospital setting (depending on the desired ventilator parameters). In contrast, most traditional ventilators (not including gas-powered resuscitators) require 1–2 liters per minute of oxygen flow. Clearly, the tradeoff here for independence from electricity for the function will depend on which resources are more limited in the desired environment.

As the development of the device moves forward to full-scale production, there are several quality control level innovations that are ongoing. This includes a robotic testing platform that uses high-resolution stepper motors as well as pressure and flow sensors to create highly accurate cam profiles. This advance will allow not only the rapid and repeatable testing of ventilators but also allows for a much less labor-intensive tuning process. In turn, this would allow for a wider variety of parameters to be evaluated and the use of much higher count splines, which would be far too time intensive for a human to feasibly generate themselves.

In our experiment, we utilized a hobby-level printer with off-of-the shelf materials purchased online during the height of the pandemic. The cost for the spring, printer materials, and silicon membrane was approximately $8 USD per unit. The cost of the connector and manometer setup was approximately$38 USD per complete setup. It is conceivable that a sufficiently skilled hobbyist could utilize the print design files and manufacturing instructions to manufacture our device in the setting of a mass casualty event or for employment in the austere environment where electricity is limited. The CRISIS-ventilator is a novel device that could be an example of virtual stockpiling for just-in-time equipment in the hospital, but which is pending further work before commercial and clinical use. The current regulatory framework for medical devices is complex for in-hospital manufacturing and further regulatory innovation is required to adequately address engineering at the patient's bedside [15].

There are two notable limitations to this pilot study. First, the number of samples available for this preliminary study was too small to generate meaningful statistical analysis and comparisons to historical control data. Second, the group of historical controls underwent ventilation for a longer period. This could theoretically increase the degree of inflammation seen on histologic evaluation and in wet/dry ratios. Overall, as a preliminary proof of concept evaluation of our device, this appeared to be a successful deployment of a 3D printed ventilator created largely on hobby level equipment.

## Conclusion

The CRISIS ventilator is a novel and innovative gas-powered resuscitator that functions independent of electricity. It can be manufactured using hobby level fused-deposition 3D printers in addition to off-the-shelf springs and silicone membranes. This pilot study demonstrated that the ventilator can indeed support the respiratory functions of large mammals. Further studies are required to evaluate the device in pulmonary disease states; however, this study suggests that the ventilator could be used in a mass-casualty scenario to support patients without significant pulmonary pathology and to liberate traditional ventilators for patients in the setting of a respiratory surge such as the third wave of COVID-19. The rapidity of constructing and validating CRISIS ventilators supports the notion that virtual stockpiles can bridge the gap between supply and demand for life-sustaining equipment using just-in-time manufacturing principles within the hospital setting [16,17].

## Acknowledgment

The authors would like to thank the Veterinarian staff of our Division of Comparative Medicine for their assistance in providing safe and humane environments in addition to the animals utilized in this experiment for their contributions to scientific progress.

## Funding Data

• David Lloyd Foundation.

## Conflict of Interest

All authors have no financial conflicts of interest. The research was conducted using internal institutional funding.

## References

1.
Hertelendy
,
A. J.
,
Ciottone
,
G. R.
,
Mitchell
,
C. L.
,
Gutberg
,
J.
, and
Burkle
,
F. M.
,
2021
, “
Crisis Standards of Care in a Pandemic: Navigating the Ethical, Clinical, Psychological and Policy-Making Maelstrom
,”
Int. J. Qual. Health Care
,
33
(
1
), epub. 10.1093/intqhc/mzaa094
2.
Kaliya-Perumal
,
A. K.
,
Kharlukhi
,
J.
, and
Omar
,
U. F.
,
2020
, “
The Second Wave of COVID-19: Time to Think of Strategic Stockpiles
,”
Can. J. Public Health
,
111
(
4
), pp.
486
487
.10.17269/s41997-020-00371-w
3.
Ramachandran
,
P.
,
Swamy
,
L.
,
Kaul
,
V.
, and
Agrawal
,
A.
,
2020
, “
A National Strategy for Ventilator and ICU Resource Allocation During the Coronavirus Disease 2019 Pandemic
,”
Chest
,
158
(
3
), pp.
887
889
.10.1016/j.chest.2020.04.050
4.
Wilgis
,
J.
,
2008
, “
Strategies for Providing Mechanical Ventilation in a Mass Casualty Incident: Distribution Versus Stockpiling
,”
Respir. Care
,
53
(
1
), pp.
96
100
(discussion 100-103).http://rc.rcjournal.com/content/53/1/96
5.
Handfield
,
R.
,
,
D. J.
,
Schneller
,
E. S.
,
Godfrey
,
A. B.
, and
Guinto
,
P.
,
2020
, “
A Commons for a Supply Chain in the Post-COVID-19 Era: The Case for a Reformed Strategic National Stockpile
,”
Milbank Q.
,
98
(
4
), pp.
1058
1090
.10.1111/1468-0009.12485
6.
Trovao
,
J. P.
,
2020
, “
Automotive Electronics Under the COVID-19 Shadow [Automotive Electronics]
,”
IEEE Veh. Technol. Mag.
,
15
(
3
), pp.
101
108
.10.1109/MVT.2020.2998710
7.
Bertsimas
,
D.
,
Boussioux
,
L.
,
Cory-Wright
,
R.
,
Delarue
,
A.
,
Digalakis
,
V.
,
Jacquillat
,
A.
,
Kitane
,
D. L.
,
Lukin
,
G.
,
Li
,
M.
,
Mingardi
,
L.
,
,
O.
,
Orfanoudaki
,
A.
,
Papalexopoulos
,
T.
,
,
I.
,
Pauphilet
,
J.
,
Lami
,
O. S.
,
Stellato
,
B.
,
Bouardi
,
H. T.
,
Carballo
,
K. V.
,
Wiberg
,
H.
, and
Zeng
,
C.
,
2021
, “
From Predictions to Prescriptions: A Data-Driven Response to COVID-19
,”
Health Care Manag. Sci.
,
24
(
2
), pp.
253
272
.10.1007/s10729-020-09542-0
8.
Blakeman
,
T.
,
Rodriquez
,
D.
,
Johannigman
,
J.
, and
Branson
,
R.
,
2019
, “
Pulsed Dose Oxygen Delivery During Mechanical Ventilation: Impact on Oxygenation
,”
Mil. Med.
,
184
(
5–6
), pp.
e312
e318
.10.1093/milmed/usy362
9.
Brito
,
A.
,
Fontaine
,
E.
,
,
S. J.
, and
Chi
,
A.
, “
Crisis Ventilator: A 3D Printed Option for Pressure Controlled Ventilation
,”
ASME
Paper No. DMD2021-1084.10.1115/DMD2021-1084
10.
Szlosarek
,
R.
,
Teichert
,
R.
,
Wetzel
,
A.
,
Fichtner
,
A.
,
Reuter
,
F.
, and
Kroger
,
M.
,
2021
, “
Design and Construction of a Simplified, Gas-Driven, Pressure-Controlled Emergency Ventilator
,”
Afr. J. Emer. Med.
,
11
(
1
), pp.
175
181
.10.1016/j.afjem.2020.09.018
11.
Terragni
,
P.
,
Maiolo
,
G.
, and
Ranieri
,
V. M.
,
2012
, “
Role and Potentials of Low-Flow CO2 Removal System in Mechanical Ventilation
,”
Curr. Opin. Crit. Care
,
18
(
1
), pp.
93
98
.10.1097/MCC.0b013e32834f17ef
12.
Moon
,
P. F.
,
Scarlett
,
J. M.
,
Ludders
,
J. W.
,
Conway
,
T. A.
, and
Lamb
,
S. V.
,
1995
, “
Effect of Fentanyl on the Minimum Alveolar Concentration of Isoflurane in Swine
,”
Anesthesiology
,
83
(
3
), pp.
535
542
.10.1097/00000542-199509000-00012
13.
Matute-Bello
,
G.
,
Frevert
,
C. W.
, and
Martin
,
T. R.
,
2008
, “
Animal Models of Acute Lung Injury
,”
Am. J. Physiol. Lung Cell Mol. Physiol.
,
295
(
3
), pp.
L379
399
.10.1152/ajplung.00010.2008
14.
Dar
,
M.
,
Swamy
,
L.
,
Gavin
,
D.
, and
Theodore
,
A.
,
2021
, “
Mechanical-Ventilation Supply and Options for the COVID-19 Pandemic. Leveraging All Available Resources for a Limited Resource in a Crisis
,”
Ann. Am. Thorac. Soc.
,
18
(
3
), pp.
408
416
.10.1513/AnnalsATS.202004-317CME
15.
Beitler
,
B. G.
,
Abraham
,
P. F.
,
Glennon
,
A. R.
,
Tommasini
,
S. M.
,
Lattanza
,
L. L.
,
Morris
,
J. M.
, and
Wiznia
,
D. H.
,
2022
, “
Interpretation of Regulatory Factors for 3D Printing at Hospitals and Medical Centers, or at the Point of Care
,”
3D Print Med.
,
8
(
1
), p.
7
.10.1186/s41205-022-00134-y
16.
Olivieri
,
L.
,
Su
,
L.
,
Hynes
,
C.
,
Krieger
,
A.
,
Alfares
,
F.
,
Ramakrishnan
,
K.
,
Zurakowski
,
D.
,
Marshall
,
B.
,
Kim
,
P.
,
Jonas
,
R.
, and
Nath
,
D.
,
2016
, “
Just-In-Time Simulation Training Using 3-D Printed Cardiac Models After Congenital Cardiac Surgery
,”
World J. Pediatr. Congen. Heart Surg.
,
7
(
2
), pp.
164
168
.10.1177/2150135115623961
17.
Shidid
,
D.
,
Leary
,
M.
,
Choong
,
P.
, and
Brandt
,
M.
,
2016
, “
Just-in-Time Design and Additive Manufacture of Patient-Specific Medical Implants
,”
Phys. Procedia
,
83
, pp.
4
14
.10.1016/j.phpro.2016.08.002