Object Finder for a Retinal Prosthesis

The proposed project seeks to provide object recognition as a feature in a retinal implant system. Participants will be able to direct an object recognition application to find a desired object in the field of view of the head-mounted camera, and to direct the participant's view towards it through the presentation of a recognizable icon. A prototype system will be developed and evaluated in human subjects in phase I. A full system implementation and a second phase of the trial will be completed in phase II.

The investigators propose to add an object-finding feature to a retinal prosthesis system. To use this feature, the participant will enable a special mode and input the desired object from a set of pre-programmed object types. Imagery from the visible light camera in the system eyeglasses will be processed using object recognition software as the participant scans their head across the room scene. When the object is identified in the scene by the processor, a flashing icon will be output to the epiretinal array in the appropriate position to guide the participant to the physical location of the object. Once located, the system will track the location of the object. There will be two phases to the human subjects evaluation, each run initially through simulations in sighted human subjects, followed by tests in Argus II participants. In phase 1, system evaluation in human subjects at Johns Hopkins UNiversity (JHU) will explore performance in representative tasks and compare prosthetic visual performance without and with the new object finding feature. An important aspect of the evaluation will be the comparison of different icons and presentation modes to assist participants in locating and reaching objects. In phase 2, the system will be integrated into the Argus II video processing unit (VPU), and JHU will conduct human trials that include functional testing of the integrated prototype in representative environments and optimizing the ergonomics of the system, e.g. simultaneous finding and tracking of multiple objects/icons.

This is a small sample open label feasibility study for an object recognition and localization system based on machine learning

To investigate the utility of a device adaptation allowing Argus II users to detect the presence and location of desired objects. Performance of the unaided Argus II system will be compared with performance using the system augmented with object recognition.

The object recognition subsystem is an add-on to the Argus II retinal prosthesis system. in the early stage of the study the subsystem will run on a separate processor; in the later stage the subsystem will run in the Argus II user's video processing unit.

Inclusion Criteria: For healthy volunteers: Vision corrected to 20/25, good general health For retinitis pigmentosa (RP) patients: End-stage retinitis pigmentosa, recipient of an Argus II retinal prosthesis system Exclusion Criteria: None

Deidentified data will be available from the investigative team, following initial publication of the results

Any researchers wit ha legitimate reason for wishing to re-analyze the data or compare them to study data collected in similar experiments