Restoring Self-Feeding with a Sensor Driven Robotic System

Project Summary DESĪN LLC currently commercializes an Obi that successfully restores functional eating for thousands of people. However, many remaining severe UEMI individuals unfortunately do not possess a sufficient level of precise functional control to reliably move their head forward to obtain food from the Obi spoon taught to deliver food just in front of their lips. Furthermore, such patients may also be on a dysphagia diet, requiring a specific volume and texture of food and liquid delivered to their mouth for safety. Utilizing funding obtained from this NIH STTR opportunity, DESĪN LLC and Carnegie Mellon University (CMU) aim to test a key scientific question - "Is it possible to expand Obi’s patient population to include individuals with severe UEMI and minimal neck control by leveraging increased autonomy of a 3D vision system?" Automation will be developed and integrated in a logical 3-step progression: 1) Gauging the spoon’s food volume for optimized portioning; 2) Dynamic point of delivery adjustment; 3) Automated food to mouth insertion triggered by mouth opening detection. Feasibility of the resulting system will be evaluated via a non-interventional feasibility clinical investigation with a small number of subjects (n=15). We proposed the following Aims: Specific Aim 1: Integration of Visual Sensors: Sensors meeting specification will be identified, evaluated, then integrated with Obi. 3 candidates for each sensor will be assessed through benchtop characterization and in-lab testing. Acceptance: Criteria of performance (50%), cost (30%), and integration ease (20%) will determine final selection. Passing IQ, OQ, PQ test protocols will gauge successful sensor integration with Obi. Specific Aim 2: Development of AI-Models: Use of off-the-shelf models and development of new models will occur to carefully add automation in a logical 3-step progression: 1. Gauging correct spoon food volume (>95% accuracy); 2. Dynamic point of delivery adjustment (<1 cm error across workspace); 3) Automated food to mouth insertion (90% insertion success, <5 mm error). Acceptance will be measured by comparing predictions to ground truth labels. Specific Aim 3: Clinical Investigation (Non-Interventional): The enhanced Obi prototype will be tested in accordance with its intended use with 15 UEMI-affected subjects, each for a single meal, in a long term care dining hall. Endpoints include: 1. Rate of subject adherence; 2) Rate of correct food delivery to the mouth position; 3) Rate of successful insertion of the spoon into the mouth; 4) Subject and staff feedback regarding perceptions of safety, expected benefits, and ease of use. Feasibility will be established when 75% combined fidelity/favorability rate is achieved.