I design devices and investigate why they fail.
ProblemPeople with severe COPD need breathing support they can wear. A biphasic cuirass needs negative and positive pressure on the chest, and that usually means a large machine.
SolutionOne blower and one rotary valve. The disc has four holes, joined in two pairs by curved slots. A quarter turn switches the shell from negative to positive pressure, so the device stays small.
ProblemBiphasic cuirass ventilation (BCV) applies negative pressure to a shell over the chest and abdomen for inspiration, and positive pressure for active expiration. Cuirass ventilators are bedside units, too large for daily ambulatory use by patients with COPD.
SolutionOne blower serves both phases: its inlet gives the vacuum and its outlet gives the pressure. A rotary valve with a four-port disc routes the two sides; a 90° turn swaps the side that goes to the shell and the side that vents to atmosphere. A sensor matrix detects breathing for triggering and synchronization. Developed under ISO 13485 and ISO 14971, tested to IEC 60601-1. US 11,839,587 B1.
ProblemAt the start of the COVID-19 pandemic, hospitals in Philadelphia expected to run out of ventilators, and parts were hard to get.
SolutionA fluidic oscillator with no moving parts, printed as two plastic parts. A wedge splits the flow, and a vortex holds the stream on the patient channel. At peak pressure the vortex collapses, and the stream moves to the second channel.
ProblemIn March 2020, hospitals in Philadelphia planned for a shortage of ventilators while the supply of valves, sensors and motors failed. The device had to run from a gas supply alone, with no moving parts, from parts that could be made locally.
SolutionA bistable wall-attachment (Coandă) fluidic amplifier, 3D-printed (FDM) as two parts. The supply jet attaches to the patient-side wall and fills the lungs. When the back-pressure gets to the switching threshold, which sets the peak inspiratory pressure (PIP), the separation bubble collapses and the jet flips to the exhaust channel. The patient exhales passively until the pressure falls and the jet returns. The jet also reacts to the effort of the patient, so the patient can start a breath.
IdeaA fun project: to make swimming with a monofin feel stronger and freer. A thruster on the fin itself is a more natural way to swim than holding on to a thruster in front of you.
How it worksA motion sensor detects each kick, and the controller adds a short burst of thrust in step with it.
IdeaHand-held underwater scooters pull the swimmer by the arms and break the line of the body. With the thruster on the monofin, the thrust acts where the kick of the swimmer acts, so the motor adds to the stroke and does not replace it.
How it worksAn IMU on the fin follows the kick cycle. The controller detects the power phase of each kick and gives the motor a short, timed pulse of thrust, then lets the fin glide. There are no buttons: the kick is the control input. US 2020/0070006 A1.
ProblemStroke studies in small animals need to see how the oxygenation of the blood in the brain changes during the stroke itself.
SolutionAn implant with a near-infrared source and a detector on the skull. The blood absorbs a part of the light, and the detected light shows the change in oxygenated and deoxygenated hemoglobin.
ProblemStroke models in small animals need a continuous measurement of the blood flow and oxygenation in the brain during the ischemic event itself, in the animal, without the size of a clinical fNIRS system.
SolutionAn implantable functional near-infrared spectroscopy (fNIRS) platform: a light source at two wavelengths and a photodetector, fixed to the skull. The absorption at the two wavelengths, through the modified Beer–Lambert law, gives the changes in the concentration of oxygenated (HbO2) and deoxygenated (HbR) hemoglobin during ischemia and reperfusion.
ProblemAn implanted intracranial pressure sensor drifts over time, and it cannot be taken out for calibration.
SolutionA vibration of known strength is applied to the head from outside. The response of the sensor gives the correction.
ProblemImplanted intracranial pressure (ICP) sensors drift from their zero over days and weeks. A sensor with drift gives a wrong pressure, and it cannot be taken out for a new zero without another operation.
SolutionA calibration from outside, with no operation: a mechanical excitation of known amplitude and frequency is applied to the head, and the response of the sensor to this known input is compared with the expected response to estimate the offset and the gain, in the patient.
ProblemA feeding tube placed through the nose can go into the airway, and the feeding fluid then goes into the lungs.
SolutionA magnet sits in the tip of the tube. A sensor on the outside of the body follows the magnet and shows the path of the tip on a screen in 3D. If the tube starts toward the airway, the operator sees it at once, pulls it back and guides it into the esophagus.
ProblemBlind bedside placement of nasogastric and nasoenteric feeding tubes ends in the airway in a small but real number of cases. Feeding into the lung causes aspiration pneumonia, and the tube itself can cause a pneumothorax. An X-ray confirms the position only after the placement.
SolutionA permanent magnet in the tip of the catheter is tracked by an external magnetic sensor. The system calculates the position of the tip in 3D in real time and draws its path, so a turn toward the trachea shows at once. The operator withdraws and redirects the tube before any feeding.
IncidentUsers of sleep-therapy devices found black particles in the air path.
FindingThe sound-abatement foam in the air path breaks down with age, heat and humidity. Particles of the decomposed foam were found in the tubing, in the masks and in the lungs of patients.
IncidentReports of black debris in the air path of CPAP and BiPAP devices.
FindingThe sound-abatement foam was a polyester-based polyurethane (PE-PUR). It degrades by hydrolysis, faster with heat and humidity, into particulate matter and volatile organic compounds (VOC). Decomposed foam particles were found in the patient tubing, in the masks and in the lungs of patients.
IncidentA home ventilator started breaths at random, out of step with the patient.
FindingDeposits on the flow sensor made it read more flow than there was. The ventilator took the error for patient effort and triggered breaths.
IncidentA home ventilator auto-triggered: it gave breaths with no effort from the patient, out of step with the patient (patient-ventilator asynchrony).
FindingDeposits on the flow sensor shifted its calibration and gave a positive flow offset. With flow triggering, the ventilator compares the inspiratory flow with the trigger sensitivity in L/min. The offset went above the threshold, so the ventilator read it as patient effort and started a breath.
IncidentInfections in patients after open-heart surgery.
FindingContaminated water in the tanks of the heater-cooler unit. The fan of the unit carries aerosol from the water into the operating room, and the aerosol can reach the operating field. Stop the fan to see the difference.
IncidentInfections with nontuberculous mycobacteria (NTM), such as Mycobacterium chimaera, months to years after open-heart surgery with cardiopulmonary bypass (CPB).
FindingThe water circuits and tanks of the heater-cooler unit (HCU) were colonized by an NTM biofilm. The exhaust fan aerosolizes the contaminated water, and the air currents of the operating room can carry the aerosol to the sterile field. Stop the fan to see the difference.
IncidentInfections at the surgical site after implant surgery in patients warmed with forced air.
FindingWarm air from the blanket rises at the blower side of the table, and the laminar flow from the ceiling pushes it down again. A slow vortex forms over the patient, and it can carry particles from the blower and the floor onto the surgical site. Stop the warming to see the difference.
IncidentPeriprosthetic joint infections (PJI) after joint-replacement surgery in patients warmed with forced-air warming (FAW).
FindingThe FAW blower releases waste heat at the side of the table. The buoyant warm plume rises against the downward unidirectional airflow (UDAF) of the operating room, and the two flows form a slow recirculation zone over the patient. This zone can carry particles from the blower and the floor and deposit them on the surgical site. Stop the warming to see the difference.
IncidentThe tip of an endoscope sheath came off during a procedure.
FindingThe collar of the tip is soldered to the tube. In each sterilization heat cycle the two parts expand and contract by different amounts. Over many cycles the solder joint fatigued and broke.
IncidentThe distal tip of a rigid endoscope (cystoscope) sheath came off during a procedure.
FindingThe tip collar is soldered to the sheath tube. The two parts have different coefficients of thermal expansion (CTE), so each steam sterilization cycle (autoclave, about 134 °C) loads the solder joint in shear, in the radial and the axial direction. Over many cycles, this low-cycle thermal fatigue started cracks in the joint and grew them until the tip separated.
Press a photograph to open it. Zoom with the wheel or a pinch, up to the full resolution of the file.
Engineer, Philadelphia.
Electronics, firmware, mechanics and test, from first concept to preproduction units.
Failure analysis of devices for litigation, through Alpha Devices.