The patient thorax is simulated by a hollow mannequin torso filled with a set of orthotopically organized models of various organs (heart, aorta, lungs, trachea, diaphragm, ribs, spine) embedded in a gastight mediastinum10. To simulate minimally invasive cardiac surgery, a hole (~skin incision) was cut in the torso around the 5th intercostal space and the ribs were spread such that a soft retractor could be inserted (see Figure 1). An incision was made in the left atrium of the heart model and it was spread open to provide access to the left ventricle (LV). A gas sampling line was installed through the ceiling of the LV and routed via ports in the LV wall and diaphragm outside the thorax model, while two additional sampling lines were routed via the main incision and positioned at the bottom of the model underneath the gas diffuser position (cranial) and by the diaphragm (caudal), respectively.
The model was placed on a platform consisting of two wooden plates (60x60cm each) joint together with two hinges, to enable tilting and simulate the oblique inclination of the operating table at a fixed angle of 20°.
Model for schlieren images
A Z-type Schlieren optical test bench was previously developed for visualizing CO2 gas clouds in a transparent sternotomy model9. For the same purpose, a plexiglass MICS models was created from a half-cylindrical surgery training model, which was made gas tight by welding on flat plexiglass sidewalls and a polyurethane bottom sheet (see Figure 2). Unused trocar ports were sealed off with rubber plugs such that only the main thoracotomy window remained open. The same heart model (transparent with left atrial incision) as in the model described above was suspended orthotopically inside the thorax model on 3 strings.
The flat sidewalls of the model allow the collimated light beam of the test bench to transfer with minimal diffraction and show the Schlieren effect inside and around the thorax model. The curved surfaces of the LV model, however, strongly diffract the light beam, resulting in dark shadows blocking the view inside that LV. Details on the model and the test bench can be viewed in the supplemental video data.
The three sampling lines, composed of standard pressure line (1.5 mm ID, 150 cm length), were each connected to a membrane pump (ZR320-03PM, ZhengFuRui, Dongguan, China) and a wide-range CO2 concentration sensor (SprintIR – 6S, Gas Sensing Solutions, Cumbernauld, UK) to monitor local CO2 concentrations (range 0-100%) throughout the tests. The sensors were calibrated with medical grade CO2 gas before each set of test runs and during the runs the three pumps generated a suction flow of 0.25±0.05 l/min. In order to minimize disturbance of the gas atmosphere, this flow rate was applied intermittently for 5 seconds every 30 seconds via a timer board (XY-J02, XiongDi, China). A digital gas flowmeter with calibrations for air and CO2 in normalized liters/minute (MassView, Bronkhorst High-Tech BV, Ruurlo, The Netherlands) was used for control of the pump suction flow rate and recording of the delivery gas flow rate. A MATLAB (The MathWorks Inc., Natick, MA) was used to record the digital outputs of the CO2 sensors and the flow meter at a sampling rate of 20 and 10 Hz, respectively.
Gas delivery devices
Three commercially available diffusors and a trocar with sideport were tested as means of delivering CO2 gas for field-flooding (see Figure 1, bottom):
• Andocor Gas Diffusor (Andocor NV, Zoersel, Belgium)
• CarbonMini (CardiaInnovation AB, Kungens Kurva, Sweden)
• Temed Gas Diffusor (Temed Medical Ltd., Sidlesham, UK).
• Trocar (30120EX1, Karl Storz, Tuttlingen, Germany)
The diffusors all have a similar construction in that they connect to a CO2 source via a ~3 m long, ¼” PVC tube with an inline anti-bacterial filter. This tube connects to a thinner malleable tube and ends in the actual diffusor tip. The latter part exhibits the main differences between the devices: the malleable tube of the CarbonMini has an outer diameter of ~4 mm and is 15 cm long, while for the Temed and Andocor devices it is ~6 mm with 40 cm length. The CarbonMini tip consists of a soft foam cylinder (Ø7 x 17 mm), the Temed of a hard-porous cylinder (Ø8 x 13 mm) with a dome shaped end, and the Andocor consists of a hard-porous cylinder (Ø9 x 15 mm) with a proximal flange (Ø11 x 5 mm).
These 3 diffusors, for their respective tests, were inserted at the highest point of the main thoracotomy incision with the malleable tube folded upward so the tip would sit against the internal anterior thoracic wall. This position is typically used clinically and places the diffusors horizontally, in contrast with usage in a sternotomy procedure where they should, according to all ‘Instruction For Use’, be bent downwards.
The fourth gas delivery system, the threaded metal trocar with a lateral stopcock, was screwed through the plastic wall of the mannequin and also penetrated the gastight mediastinum compartment via a rubber seal. The stopcock was used to connect the CO2 supply such that the gap between the trocar inner lumen and the inserted instrument shaft acted as a tunnel for CO2 delivery inside the thorax.
Data collection Protocol
Data was recorded for all combinations of the following factors:
• 4 different gas delivery systems
• 4 different CO2 flows (1, 3 ,5 and 8 l/min)
• 2 tilting angles of the mannequin (0° and 20°)
for a total of 32 tests, executed sequentially in one run. Five runs (repetitions) with the 32 tests were performed, whereby the order of tests was randomized within each run. Each test starts by purging the model of CO2 by removing the lid representing the diaphragm and using a fan inside until all concentration sensors indicated <1% CO2. Next, the selected delivery tool was installed and the tilt angle of the model set according to the selected test. The timer boards and sampling pumps were then activated and finally the MATLAB (The MathWorks, Natick, MA) script for recording was started simultaneously with the start of CO2 delivery flow. The recording stopped automatically after 800 seconds, after which all active components were manually turned off and the model was drained again.
A similar protocol was used for the gas cloud visualization in the transparent thorax model, with the exception that no repetitions were performed and imaging was not performed at 8 l/min. The CO2 gas in this model was removed with a laboratory vacuum pump in between recordings.
All concentration data acquired in these tests follows a similar hyperbolic curve pattern: starting at 0%, a steep rise in concentration occurs and then tapers of, leading to a steady plateau which is maintained until the end of the recording9. Therefore, after filtering of the data (IIR low-pass filter of order 50 with 50 mHz cut-off frequency), two parameters were extracted: the maximum CO2 concentration achieved (i.e., the average of the final plateau), and the rise time (i.e., time needed to reach 90% of the plateau).
This exploratory study is based on the findings in one virtual patient, and is intended to guide parameter (factors and levels) selection for future clinical studies. In model data collected in a controlled environment, the variation is expected to be much smaller than in a clinical setting and for this and the limited repetitions, the statistical analysis of the model data is limited to non-parametric multi-level comparisons, where Kruskal-Wallis rank sum test with pairwise Dunn correction was used for evaluating the effect of a single factor and Friedmann’s test was used for two-factor comparisons. All tests were performed with MATLAB software and a value of p<0.05 was considered significant.