Ophthalmic ultrasound training is often done on healthy volunteers, limiting the scope of training on patients with a variety of conditions.The purpose of this project was to engineer an ophthalmic ultrasound training platform that simulated the human eye to enable clinically-relevant learning and practice of ocular ultrasound techniques. The most important requirement was the accuracy of the model’s ultrasound properties. Physiological accuracy of the model’s dimensions, as perceived on an ultrasound reading, was additionally identified as crucial. Lastly, the model was used to provide an easy and clinically-relevant user experience. To simulate the clinical experience, our group also aimed to incorporate 30-deg movement of the eye in each direction. In terms of productizing our design, our group aimed to create a design that was easily reproducible for homogeneous training by minimizing variability amongst models.

Model eyes were designed for healthy eyes and three disease conditions (melanoma, retinoblastoma and retinal detachment). These models were 3D printed using a variety of materials, resolutions, infill percentages and sizes to most accurately mimic a physiologically relevant ultrasound scan. The final eye model was printed out of polylactic acid with inner and outer diameters of 24 mm and 28 mm, respectively. The optic nerve was designed as a conically hollow cylinder and placed directly behind the iris to best simulate the physiological appearance of the eye in an ultrasound image while providing optimal space for the movement system. The base was modelled after a human face with an empty eye socket for the eye model to be placed in. The eyes were easily removable from these sockets in the base, allowing for easy cleaning and quick interchangeability of the training environment. To autotomize movement of the eyses, bipolar stepper motors were used with a pulley system to move the eye in different directions. This motor system and base was controlled by an Arduino through a MATLAB, allowing the operator to specify movement of the eye. This accurately simulated an ophthalmologists' test environment where patients are specifically instructed to move their eye.

Qualitative assessments were completed, comparing human eye ultrasound images to that of the model eyes; important physical landmarks were visible and accurate. Pixel measurements and pixel intensity were used to quantify the efficacy of our design. The measured diameter of our model eyes had a variation of only 0.5mm on average compared to the desired diameter. We also measured a significant difference in pixel intensity between the vitreous and disease characteristics. We completed a blind test whereby four ophthalmologists diagnosed our model eyes. Four of five eyes were correctly diagnosed 100% of the time, while the retinal detachment model was diagnosed correctly 25% of the time. This disease model was less accurate because the detachment cast a shadow behind it which looked like tumour cells, similar to retinoblastoma or melanoma. Based on the positive results seen from our design, we are certain our design will revolutionize the way ophthalmologists are trained by creating a reproducible, uniform model that accurately mimics healthy and diseased eyes.