Welcome back to AIMedily! I’m happy you’re here.
Last month, I went to RehabWeek. An international congress where clinicians, engineers, researchers, and industry come together.
The quality of speakers, posters, and workshops was amazing. I even got to visit the Shirley Ryan AbilityLab.
But what I enjoyed the most was that everyone was open to sharing - and connecting. To work together to improve the quality of life of people with disabilities.
But let’s get started with today’s issue. Are you ready?
🤖AIBytes: AI and prosthetics, Robotic exoskeletons in spinal cord injury, and AI rehab in stroke.
🦾TechTool: One camera - complete gate analysis. A Socially Assistive Walker
🧬AIMedily Snaps: Interesting links
🧩TriviaRX: One question (just for fun)
🤖 AIBytes
A team trained an AI model to create accurate, comfortable, and wearable sockets using a 3D scan of the residual limb.
🔬 Methods
Phase 1 – Building the AI with 3D scans from 116 transtibial amputees. Along with socket designs crafted by skilled prosthetists.
The AI model could predict the shape of an optimal socket based on limb geometry.
Phase 2 – Testing AI in the Real World
Ten patients tested both sockets:
A manually made socket
An AI-designed socket- based on the residual limb, 3D-printed.
The prosthetist, physical therapist, and the user evaluated the performance using:
Activity tracking using Actigraph sensors
Skin integrity and pressure mapping
Socket fit and need for modifications
User satisfaction via questionnaires
📊 Results
Precision: AI-generated sockets deviated only 2.5 mm from expert designs.
Wearability: 8 out of 10 AI sockets were viable.
Comfort: Scores matched those of traditional sockets.
User/Evaluator Preference: In 2 cases, they preferred the AI design.
🔑Key Takeaways
This study shows that AI can accurately replicate expert-level socket design:
Automates the design process without compromising precision.
Enables rapid 3D printing—potentially cutting fitting time.
Improves access in low-resource or rural areas lacking experienced prosthetists.
🔗 Evaluating the Effectiveness of Transtibial Prosthetic Socket Shape Design Using Artificial Intelligence: A Clinical Comparison With Traditional Plaster Cast Socket Designs. van der Stelt, Merel et al. Archives of Physical Medicine and Rehabilitation, Volume 106, Issue 2, 239- 246. https://doi.org/10.1016/j.apmr.2024.08.026
This meta-analysis explores if Robotic Exoskeleton Gait Training was superior to Conventional Physical Therapy in spinal cord injury (SCI).
🔬Methods
Type: Meta-analysis of 15 randomized controlled trials
Sample: 579 patients with Spinal cord injury (ASIA A–D), ages 26–71
Time Since Injury: 2 months–15 years
Interventions: 3–20 weeks, 2–5 sessions/week
Assessment:
Walking Speed: 10-Meter Walk Test
Endurance: 6-Minute Walk Test
Balance & Mobility: Timed Up and Go
Pulmonary: Forced Expiratory Volume in 1 Second (FEV1)
📊Results
Walking Speed: No significant difference (p = 0.08)
Walking Distance: No significant difference (WMD = -1.83 meters, p = 0.78)
Balance: Significantly better in Robotic training (p = 0.04)
Functional Scores:
WISCI-II: Significant improvement (p = 0.0001)
LEMS: Significant improvement (p = 0.0005)
Respiratory Function (FEV1): Improved in REGT (p = 0.03)
🔑Key Takeaways
Robotic exoskeleton training significantly improves balance, strength, functional independence, and respiratory capacity.
Robotic exoskeleton gait training doesn't outperform conventional physical therapy in speed or distance. Especially for chronic patients.
In chronic injury, conventional therapy may be better to recover speed.
.🔗 Liu S, Chen F, Yin J, Wang G, Yang L. Comparative efficacy of robotic exoskeleton and conventional gait training in patients with spinal cord injury: a meta-analysis of randomized controlled trials. J NeuroEng Rehabil. 2025;22:121. https://doi.org/10.1186/s12984-025-01212-6
This comprehensive review explores the integration of AI across stroke rehabilitation. From acute management through chronic recovery.
📌 Key Contributions
AI-enhanced imaging (CT/MRI)
Enables early detection of ischemic penumbra, supporting faster, more personalized interventions.
Clinical decision support tools
Optimize acute treatments like thrombolysis and endovascular therapy.
Robotic systems & exoskeletons,
AI exoskeletons and robotic assistive devices enable adaptive motor training.
How? by interpreting patient-specific movement patterns.
Real-time feedback loops improve precision and reduce therapist burden.
AR + AI Integration
Virtual Reality and Augmented Reality environments powered by AI offer:
Task-specific and immersive rehab scenarios.
Platforms personalize difficulty levels based on performance.
Brain–Computer Interfaces (BCIs)
Machine learning models decode neural signals for intention-based movement control in BCI.
Applied to upper limb recovery and attention training post-stroke.
Wearables with AI
Provide continuous monitoring and feedback, extending rehab to home settings.
AI-driven tele-rehabilitation
Bridges geographic gaps for remote care delivery.
Predictive Analytics
AI models can forecast functional recovery. This allows adjustments to therapy and to personalise treatment.
🔑 Key Takeaways
This article outlines a clear vision:
AI is not a replacement for human care.
It is an amplifier of clinical precision, individualized recovery, and healthcare efficiency.
Also flags critical considerations—data privacy, regulatory standards, and ethical implementation—as essential for safe, effective adoption.
🔗 Kopalli SR, Shukla M, Jayaprakash B, et al. Artificial Intelligence in Stroke Rehabilitation: From Acute Care to Long-term Recovery. Neuroscience. 2025 Apr;572:120412. https://doi.org/10.1016/j.neuroscience.2025.03.017.
🦾TechTool
3DGait is an AI enhanced gait analysis system that uses:
Single consumer-grade depth camera
No markers required
No complex setup or calibration
No trained personnel needed
⚙️ Technology:
Advanced machine learning algorithms
Produces 49 gait biomarkers
Angular, spatial, and temporal measurements
🔑 Key takeaways
Clinically acceptable accuracy vs. traditional systems
No markers, calibration, or fixed cameras needed
Practical for non-specialist clinics and home use
Supports patient monitoring and chronic disease management
🔗 L, Chang R, Wang J, et al. Artificial intelligence-enhanced 3D gait analysis with a single consumer-grade camera. J Biomech. 2025 Jun;187:112738. https://doi.org/10.1016/j.jbiomech.2025.112738
A research team from the University of Bristol developed a socially assistive walker. This device delivers both physical and cognitive support to older adults.
Designed to be affordable and user-friendly.
🔬 Methods
A team of geriatric care professionals co-design the walker.
It was a walker frame with integrated sensors, feedback mechanisms, and a user interface for interactive support.
They tested the walker on healthy adults during daily living activities.
Had two distinct modes:
High Engagement: Voice prompts, virtual agent presence, personalized interactions
Low Engagement: Text-only instructions, minimal feedback.
📊Results
78.5% preferred the high-interaction mode (p < 0.05). In embodiment, verbal feedback, and proactive cues.
Users valued verbal praise, clear instructions, and multimodal feedback.
🔑 Key Takeaways
Social-cognitive features improve patient preference and engagement.
🔗 Haque MR, Yang H, Yoshida T, Tsujita T. (2024). Socially assistive walker: user preferences on low and high interaction modes. Frontiers in Robotics and AI, 11:1401663. doi: 10.3389/frobt.2024.1401663
🧬AIMedily Snaps:
🧩TriviaRX
Why can't you tickle yourself?
A) Cerebellum predicts self-touch
B) Spinal gate theory
C) Lack of surprise element
D) Motor cortex inhibition
(The answer will be in the next issue)
That's all for today!
If you have articles on AI and Rehabilitation and want to share them, reply to this email 📤.
If you want me to write about a specific topic, make your request!
And one last favor, please share AIMedily with colleagues. I would be forever grateful 🥰 https://www.aimedily.com/
See you next week,
Itzel Fer, MD PM&R
Forwarded this email? Sign up here
