EVALUACIÓN DE LA INTERACCIÓN PACIENTE-ROBOT EN ESCENARIO DE REHABILITACIÓN CARDIACA

Maria Insuasty, Nathalia Céspedes, Marcela Múnera, Carlos A. Cifuentes, Mónica Rincón Roncancio, Luisa F. Gutiérrez2021 · DOI

TL;DR

This study evaluates patient-robot interaction in cardiac rehabilitation (CR) using a NAO robot and a custom interface. The findings suggest that while patients exhibit a largely neutral attitude, they show interest in the robot's personalized feedback and safety alerts, indicating potential for social robots in CR.

What Problem Does This Paper Address?

Cardiac rehabilitation (CR) programs are crucial for improving the quality of life and recovery of patients with cardiovascular diseases. However, a significant challenge is ensuring patient adherence to these programs, with dropout rates ranging from 30-80%. Socially assistive robots (SARs) are emerging as a potential tool to enhance patient engagement and support in various healthcare settings. This paper addresses the gap in understanding how patients interact with SARs specifically within the context of cardiac rehabilitation, aiming to quantify this interaction and assess its impact on patient experience.

Methods

The researchers developed a human-robot interface that integrates a sensory module and a NAO robot. This interface monitors physiological parameters (heart rate) and provides cognitive support through social interaction. Key functionalities include:

  • Personalized Feedback: The robot performs facial recognition to personalize interactions, using the patient's name and recalling past performance.
  • Physiological Monitoring: Real-time heart rate monitoring is conducted.
  • Patient Input: Patients are prompted for their Borg scale rating (perceived exertion).
  • Feedback Events: Based on patient responses, the robot triggers various feedback events.
  • Interaction Classification: Interactions were categorized into two types:
  • Unidirectional Interactions (IU): Actions from one agent (robot or patient) to the other without requiring a direct response (e.g., robot's motivational phrases, patient's eye contact).
  • Bidirectional Interactions (IB): Interactions requiring a patient response through the interface (e.g., confirming identity, responding to alerts).
  • Video Analysis: 7 patients undergoing CR participated in 11 sessions, and their interactions were analyzed from video recordings.

Datasets

Not discussed in the paper. The study utilized data collected from 7 patients participating in 11 cardiac rehabilitation sessions.

Key Insights and Ideas

  • Personalization Matters: The robot's ability to recognize patients by face and use their names significantly contributed to a personalized experience, generating positive reactions.
  • Neutral but Engaged: While patients' attitudes were predominantly neutral (89%), they showed interest in the robot's requests and warnings, suggesting a level of engagement beyond passive observation.
  • Unidirectional Dominance: The majority of interactions were unidirectional (64%), indicating that patients primarily engaged with the robot through their own actions or responses to its prompts, rather than initiating complex dialogues.
  • Safety First: Correction of posture and warnings about elevated heart rate or Borg scale were the most frequent interaction events, highlighting the robot's role in ensuring patient safety and adherence to exercise protocols.
  • Novelty Effect: Initial sessions saw a higher number of interactions, likely due to the novelty of the robot, which then stabilized as patients became accustomed to the technology.

Key Results

  • Interaction Volume: An average of 16 interactions per session were observed, with a peak of 22.6 in the first session.
  • Interaction Type Distribution: 64% of interactions were unidirectional (IU), and 36% were bidirectional (IB).
  • Event Frequency: The most frequent events were posture correction (28%), eye contact (22%), motivation (21%), Borg scale requests (19%), and warnings (10%).
  • Patient Attitude: 89% of patient attitudes were classified as neutral, 11% as positive (including smiles), and 3% as negative.
  • Interaction Time: The percentage of session time involving interaction was relatively low, ranging from 4.43% to 7.23%, suggesting that the robot's presence did not significantly disrupt the rehabilitation process.

Limitations and Future Work

  • Limitations:
    • The study involved a small sample size (7 patients).
  • The majority of participants were male, limiting generalizability to female patients.
  • The study focused on a specific phase (Phase II) of cardiac rehabilitation.
  • The authors acknowledge that the initial sessions were influenced by a "novelty effect."
  • Future Work:
    • The authors suggest varying the robot's behaviors and physical appearance to identify preferred interaction styles and further enhance human-robot interaction.
  • Exploring new scenarios with modified robot aspects and dialogues is recommended to discover optimal configurations for different patient populations in rehabilitation settings.

Summary updated

Abstract

La robótica de asistencia social se ha introducido en ámbitos tales como: la salud, la educación y la industria. Este estudio busca evaluar y cuantificar la interacción que tienen los pacientes hacia un robot de asistencia social en el área de Rehabilitación Cardiaca (RC). Siguiendo este objetivo, se usó una interfaz humano-robot que integra un módulo sensorial y un robot NAO. Dentro de las principales funcionalidades de la interfaz se encuentran: el monitoreo online de parámetros fisiológicos y el soporte cognitivo a través de la interacción social durante el ejercicio desempeñado en la RC. Inicialmente, el robot realiza un reconocimiento facial del paciente, con el fin de personalizar las realimentaciones durante la sesión al utilizar el nombre del paciente y la información recopilada del desempeño en las sesiones anteriores. Durante el entrenamiento, la interfaz realiza la vigilancia de la frecuencia cardiaca y pregunta al paciente sobre su escala de Borg. Dependiendo de la respuesta dada por el paciente, el robot puede activar diferentes eventos de realimentación. Este estudio incluye el análisis de video de 7 pacientes que participaron activamente en las sesiones de RC asistida por el robot. Para esto se tuvieron en cuenta dos tipos de interacciones: (i) interacciones unidireccionales e (ii) interacciones bidireccionales. La primera corresponde a una acción de un agente (Robot o paciente) con respecto al otro, sin una respuesta (e.g. cuando el robot dice frases de motivación, el contacto visual del paciente hacia el robot, etc.). Por otra parte, en las interacciones bidireccionales se requiere una respuesta del paciente por medio de la interfaz tras una acción del robot (e.g. confirmación sobre el reconocimiento del paciente, alertas sobre el ritmo cardíaco, la solicitud de la escala de Borg, etc.). Los resultados del estudio se encontraron al evaluar el efecto del robot sobre la terapia, realizando un análisis sobre la cantidad de eventos de contacto visual, la ejecución y seguimiento de instrucciones y la respuesta humano-robot (asociada a las interacciones bidireccionales), de igual forma se midió la actitud del paciente respecto a cada una de las interacciones presentadas. Respecto al porcentaje de interacción obtenido durante cada sesión se obtuvo un valor máximo de 16.11% y un promedio de 11.38%, se obtuvo un promedio del 64% de interacciones unidireccionales y un 36% de bidireccionales. Fue posible observar que, si bien la actitud de los pacientes fue en promedio 89% neutra, ellos mostraron interés sobre las solicitudes del robot y sobre las advertencias realizadas en caso de ser necesario.
BibTeX
@misc{insuasty2021evaluacion,
  author = {Maria Insuasty and Nathalia Céspedes and Marcela Múnera and Carlos A. Cifuentes and Mónica Rincón Roncancio and Luisa F. Gutiérrez},
  title = {{EVALUACIÓN DE LA INTERACCIÓN PACIENTE-ROBOT EN ESCENARIO DE REHABILITACIÓN CARDIACA}},
  year = {2021},
  doi = {10.26507/ponencia.1950},
}
Figure