🤖 AIBytes
This study evaluated whether machine learning can predict upper limb recovery in stroke patients with data from the 1st week post-stroke.
🔬 Methods
Participants: 296 stroke patients evaluated within 2 weeks post-stroke, and at 6 months.
Intervention: Machine learning algorithms.
Predictors used: Action Research Arm Test score (ARAT), shoulder abduction, and finger extension.
📊 Results
This machine learning model XGBoost outperformed other models in predicting recovery.
Early motor assessment scores and patient age significantly contributed to prediction accuracy.
🔑 Key Takeaways
The machine learning XGBoost model offers superior early prediction of stroke upper limb recovery vs. classic models.
Incorporating early motor scores and age improves predictive performance
Models like this could support accurate prognoses, helping clinicians set realistic treatment goals, inform patients, and plan discharge
The code and predictive model are available to use.
🔗 Van der Gun GJ et al. Can machine learning improve on the early prediction of upper limb recovery after stroke? J NeuroEng Rehabil. 2025; 22:223. https://doi.org/10.1186/s12984-025-01743-4
A 2025 network meta-analysis compared 13 types of AI-assisted rehabilitation for patients with musculoskeletal disorders (MSDs).
The goal: identify which AI-based interventions most effectively improved pain, functional recovery, and range of motion (ROM).
🔬 Methods
Design: Systematic review and network meta-analysis.
Sources: 33 randomized controlled trials across 15 countries.
Population: Adults with osteoarthritis, tendinopathies, post-operative conditions, and chronic pain.
Interventions:
AI-Feedback Motion Training
AI-Prescription Apps
Robotic Exoskeletons
Single-Joint Rehabilitation Robots
Therapeutic and Gamified Exergaming
Synchronous / Asynchronous Telerehabilitation
Virtual Reality and Multimodal Digital Platforms
Hybrid Physical Therapy + Exergaming
📊 Results
Pain Relief
Most effective: Therapeutic Exergaming (87.6 %) and Robotic Exoskeletons (86.3 %).
Least effective: Asynchronous Telerehabilitation (4.2 %) and Conventional Care (12 %)
Functional Outcomes
Top: Gamified Exergaming (99.6 %), PT + Exergaming (81.2 %)
Lowest: AI-Feedback Motion Training (17.8 %), Conventional Care (17.1 %)
Range of Motion (ROM)
Most effective: Single-Joint Rehabilitation Robot (84.7 %), AI-Feedback Motion Training (83.7 %)
Least: Gamified Exergaming (31.2 %), Conventional Care (15 %)
Subgroup Insights:
Younger patients and those with mild-to-moderate MSDs showed the greatest benefit.
Long-term efficacy remains uncertain due to short follow-up durations.
🔑 Key Takeaways
AI-assisted rehabilitation consistently outperformed conventional therapy across pain, function, and ROM outcomes.
Gamified and Therapeutic Exergaming improved both engagement and functional recovery.
Robotic Exoskeletons and Single-Joint Robots delivered the best results for pain reduction and mobility gains.
Asynchronous Telerehabilitation was less effective.
🔗 Luo Z et al. Effectiveness of AI-assisted rehabilitation for musculoskeletal disorders: a network meta-analysis. Front Bioeng Biotechnol. 2025;13:1660524. https://doi.org/10.3389/fbioe.2025.1660524
🦾TechTool
A platform for clinic management and exercise prescription.
Can be used for exercise prescription, assestment templates, and outcome tracking designed for rehabilitation.
Contains a library with thousands of planned exercises, customizable for stroke and neurorehabilitation.
Turns complex topics, cases, or research ideas into clear, structured mind maps.
Uses AI to connect related concepts—making patterns easier to spot.
A powerful way to visualize reasoning, design protocols, or organize ideas.
Check a Mind Map of AI in medicine here.
Let’s you ask questions directly to PDFs and get clear, accurate answers in seconds.
Connects instantly to all your tools and turns messy, scattered data into one clean, searchable source with references.
Uses AI to understand your questions and surface the exact data you need.
Now, we will continue learning how to improve your prompts:
Chain-of-Thought Prompting
Is a prompting method where you tell the model to reason step-by-step before giving the final answer.
Improves performance on tasks that need logic, calculations, or multi-step reasoning.
When to use it
Useful for clinical decision trees, protocol steps, and multi-criteria comparisons.
Not necessary for simple fact retrieval. Will make responses longer.
Example prompt: “Think step by step, then give me a short final answer.”
🧬AIMedily Snaps
‘It keeps me awake at night’: machine-learning pioneer on AI’s threat to humanity (Link).
Leading a New Era of AI-Powered Biology to help cure, prevent and manage disease with Meta (Link).
OpenAI is weighing a move into consumer health apps (Link).
25th European Congress of Physical and Rehabilitation Medicine (ESPRM2026) 23-27 March 2026. Krakow, Poland (Link).
Want to keep up with AI in medicine in less than 5 minutes a week?
Subscribe free to AIMedily.Itzel Fer, MD PM&R
🧩TriviaRX
What was the first medical specialty to formally test an AI system in a real clinical setting (with patients) rather than simulations or case vignettes?
A) Radiology
B) Emergency Medicine
C) Dermatology
D) Oncology
Now, the answer from the last TriviaRX.
✅ C) It used nearly 2,000 YouTube surgery videos. Stanford and Harvard created the AVOS (“Annotated Videos of Open Surgery”) dataset with nearly 2,000 YouTube-sourced surgical videos.