The Heart-Lung Machine

Set out the constraint before choosing hardware, because the constraint eliminates almost everything. A centrifugal impeller accelerates fluid in a tight clearance and the shear at the blade tip destroys red cells. A piston pump needs valves, and every valve seat is a place where cells are trapped and crushed as it closes. A gear pump meshes teeth through the fluid. A diaphragm pump is gentler but needs valves too. The roller pump does something different: the fluid stays inside a length of tubing, and the pump squeezes the tubing from outside. Nothing in the pump ever contacts the fluid. There are no valves because the occlusion point itself is the valve — the squeezed tube seals, and as the roller advances the seal travels with it. Sterility becomes trivial, because only the tubing needs to be sterile and the tubing is disposable. It also gives you flow directly. One revolution displaces the tube's internal volume over the arc the rollers sweep, so flow is revolutions times a fixed volume, independent of the pressure it is working against. That last property is why it was adopted and why it is dangerous: a roller pump will happily generate enough pressure to burst the circuit, because it does not know or care what is downstream. Occlusion setting is the one adjustment. Too loose and the pump backflows and the flow calibration is wrong. Too tight and the tubing is abraded, shedding particles into the circuit, and the cells in the squeezed zone are crushed. It is set just barely occlusive, and step 4 is how you find that point.

Occlusion first, because every later number depends on it. Fill the circuit with water and raise the outlet tubing vertically about a metre. Stop the pump. Watch the column. With the rollers over-occluded the column holds indefinitely; with them under-occluded it falls quickly. Back the occlusion off until the column falls at roughly one centimetre per minute. That is the standard setting and it means the rollers are just barely sealing — enough to pump, not enough to grind the tube. Now calibrate. Run the pump into a tared beaker on the scale for exactly sixty seconds at a set speed, and weigh what arrives. Water is one gram per millilitre so the mass is the volume directly. Repeat at five speeds across the range and plot flow against rpm. The plot should be a straight line through the origin. Where it bends away at the top, the tube is not refilling fast enough between rollers and you have found the pump's real ceiling — which is a property of the tubing, not the motor. Where it sits below the theoretical line from the notebook, the difference is your backflow past the occlusion. Repeat the whole calibration with glycerol solution at about four times the viscosity of water, which is closer to blood. The straight line should barely move — that is the roller pump's defining virtue, and it is worth seeing rather than being told.

The pump is gentle, not harmless, and you can measure how gentle with a surrogate that fails visibly. Make a suspension of intact cells you can afford to destroy — baker's yeast works, and so does a dilute suspension of oil in water stabilised just enough to hold. Run it around a closed loop for an hour at your working speed, drawing a sample every ten minutes. For yeast, count intact cells under the microscope. For the emulsion, watch the droplet size distribution shift downward as shear breaks droplets apart. Plot damage against total passes through the pump head, not against time — passes is the quantity that matters, and a small circuit at high flow racks them up fast. Then repeat at a deliberately over-occluded setting and watch the curve steepen. That comparison is the reason occlusion is set the way it is in step 4, and it is far more convincing measured than asserted. Where this stops. You have built and characterised a non-contact pumping circuit, which is genuine and useful engineering — the same circuit runs dialysis, fermentation, and dosing. It is not cardiopulmonary bypass. Real bypass requires anticoagulation of the whole patient because blood clots on contact with any foreign surface, an oxygenator that transfers gas without letting a single bubble through, temperature management, continuous blood gas and pressure monitoring, a trained perfusionist watching it, and a surgical team on the other end. Every one of those is a discipline in itself. Run water and yeast. The value here is the pump principle and the interlock logic, and both transfer to anything you build that must not damage what it moves.