Topics in spinal cord injury rehabilitation · 2026
AI in the Wild: A Pilot Deployment of an AI-Enhanced Robot-Assisted Feeding System at a Rehab Hospital.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Miller Jacob, Foley Claire, Kubec Gina, Burkhart Ian, Zingale Nicholas, Schearer Eric
Аннотация
BACKGROUND: Assistive robots have the potential to mitigate the caregiver shortage crisis and increase independence for individuals with spinal cord injury (SCI). Advances in artificial intelligence (AI) have improved the capabilities of assistive robots, however they still lack evaluation outside of controlled laboratory settings.
OBJECTIVES: To evaluate the technical performance, usability, and workload of an AI-enhanced robot-assisted feeding (RAF) system deployed in an inpatient SCI rehabilitation unit.
METHODS: RAF-HI, a system incorporating AI-based food detection, natural language communication, and mouth tracking, was integrated into a rehabilitation hospital. Two participants with SCI used the system to eat 8 modified hospital meals in their private rooms. System performance metrics (bite acquisition and transfer rates), usability (System Usability Scale [SUS]), workload (NASA Task Load Index [NASA-TLX]), and researcher interventions were recorded.
RESULTS: Across the meals, the system had an acquisition rate of 88.4% and a transfer rate of 87.4%, similar to state-of-the-art benchmarks. However, usability and workload scores among participants were poor (NASA-TLX 42.9, SUS 47.5), and frequent researcher interventions were required to keep the system operating (36 total, 4.5 per meal). Participants cited low adaptability and complex setup requirements as obstacles to usability and demonstrated responsibility over their comfort within the interaction.
CONCLUSION: Our use of AI tools successfully achieved technical benchmarks in a hospital environment but fell short of achieving a truly independent meal. We argue advancing RAF systems toward human-inspired, situationally aware intelligence is necessary to translate technical capability into real-world clinical utility.
OBJECTIVES: To evaluate the technical performance, usability, and workload of an AI-enhanced robot-assisted feeding (RAF) system deployed in an inpatient SCI rehabilitation unit.
METHODS: RAF-HI, a system incorporating AI-based food detection, natural language communication, and mouth tracking, was integrated into a rehabilitation hospital. Two participants with SCI used the system to eat 8 modified hospital meals in their private rooms. System performance metrics (bite acquisition and transfer rates), usability (System Usability Scale [SUS]), workload (NASA Task Load Index [NASA-TLX]), and researcher interventions were recorded.
RESULTS: Across the meals, the system had an acquisition rate of 88.4% and a transfer rate of 87.4%, similar to state-of-the-art benchmarks. However, usability and workload scores among participants were poor (NASA-TLX 42.9, SUS 47.5), and frequent researcher interventions were required to keep the system operating (36 total, 4.5 per meal). Participants cited low adaptability and complex setup requirements as obstacles to usability and demonstrated responsibility over their comfort within the interaction.
CONCLUSION: Our use of AI tools successfully achieved technical benchmarks in a hospital environment but fell short of achieving a truly independent meal. We argue advancing RAF systems toward human-inspired, situationally aware intelligence is necessary to translate technical capability into real-world clinical utility.