Remote Magnetic Navigation and Applications in Ophthalmic Surgery

Abstract

Magnetism has entertained a close relationship with medicine throughout history, but its ability to navigate therapeutic devices inside the human body has emerged in the last few decades thanks to technological improvements in the fabrication, and control of magnetic devices. Remote magnetic navigation of untethered devices, also known as micro or nanorobots, or tethered surgical devices including catheters, endoscopes, and needles can be achieved by generating magnetic fields from outside the human body, using a magnetic navigation system. This thesis is divided in two parts. The first part discusses the prediction of generated magnetic fields, a fundamental task of remote magnetic navigation that is required for simulating, controlling, and localizing magnetically navigated devices. We first explore interpolation based methods, which create continuous representations of magnetic fields using pre-existing data. Several interpolation methods are compared based on their ability to accurately predict magnetic fields and magnetic field gradients, and how well they respect certain physical constraints obeyed by magnetic fields. Magnetic navigation systems using electromagnets that are large enough to perform magnetic navigation at human scales exhibit nonlinear magnetic saturation. We first propose a strategy that can correct for electromagnet saturation in existing linear models. Machine learning based methods are capable of modeling such complex nonlinear behavior with multiple inputs and outputs from data alone. We show an artificial neural network that achieved superior field prediction accuracy to both linear and corrected methods. This was followed by the application of a generative convolutional neural network that far outperformed all other methods. The second part of the thesis concerns the application of remote magnetic navigation for the control of tethered surgical devices in ophthalmology. Surgery on the retina is exceedingly challenging, involves movements and forces that are at the limits of human ability and perception, and for that reason has long been proposed as a candidate for the application of medical robotics. Differing from existing robots that use mechanical transfer of motion to navigate tools inside the cavity of the eye, this work presents flexible devices that are navigated using magnetic fields. Such devices combine fine position control, extreme miniaturization, and enhanced safety over existing rigid tools. We first describe a magnetically navigated laser probe that could be used for treating advanced forms of diabetic retinopathy, a rapidly growing and already leading cause of vision loss. By tracking the laser position in real-time using computer vision, the probe is navigated in closed-loop, and the procedure, which is repetitive, lengthy, and painful for patients, can be automated. There is active research in the development of new therapies for treating diseases that cause degeneration of the retina, particularly age-related macular degeneration, the leading cause of blindness in the developed world. New treatments including virus-carried gene therapies and stem cells need to be delivered close to the targeted areas of the retina for them to be effective. Subretinal injections have been proposed as the most promising pathway for delivery of such therapeutics, but they are a challenging surgical procedure with significant associated risks. We have developed a magnetically navigated cannula that can be navigated throughout the retina with micrometer precision, and with much greater ease than with handheld cannulas. Combined with optical coherence tomography, an increasingly popular imaging method in ophthalmology that enables high-definition 3D visualization of the retina, we show that a cannula can be placed precisely, for safe injections in the subretinal space. Injections were demonstrated in ex-vivo porcine eyes, as a first step towards subretinal delivery in human patients.