PT-CHARLIE
PT-CHARLIE is an early Dutch-first prototype for people exploring low back pain guidance. It feels simple on the surface: you chat or speak with Charlie. Underneath, it explores clinical reasoning patterns: safety-oriented questions, important indicators, patient patterns, and possible next steps.
Plain explanation
PT-CHARLIE is an early AI physiotherapy prototype for low back pain education, exploration, and product development. It is not built to give random chatbot advice. It is built to ask better questions, notice important signals, and explain possible routes forward in language people can understand.
For a patient, that means: “What could be happening, what should I watch out for, and what can I do next?” For a physiotherapist, it means: “Which indicators did the system pick up, and does the suggested pathway make clinical sense?” For a researcher or student, it becomes a visible model of clinical reasoning.
The prototype is especially focused on chronic and persistent low back pain, where the story is rarely only about tissue. Pain, sleep, work, stress, fear of movement, function, and safety signals all matter.
Important boundary: PT-CHARLIE is not MDR or CE registered, is not a medical device, and should not be used for diagnosis, treatment decisions, emergency triage, or replacing a qualified healthcare professional.
What happens inside the app
Charlie may sound conversational, but the prototype is designed around details physiotherapists often need: red flags, pain behavior, function, psychosocial context, and recovery barriers.
How it works
The current repository shows a modular prototype: conversational intake, safety-oriented screening, indicator extraction, and pathway exploration. These layers make the reasoning visible for patients, clinicians, students, and researchers while keeping the boundary clear: this is exploratory software, not a regulated medical device.
Layer 1
The prototype asks plain-language questions about pain duration, function, work, sleep, movement fear, mood, beliefs, support, and previous care. The goal is to make intake feel human while still collecting structured information.
Layer 2
Red flag prompts sit early in the flow. The prototype checks for warning signs such as neurological changes, bladder or bowel symptoms, unexplained weight loss, fever, trauma, cancer history, infection risk, and inflammatory patterns.
Layer 3
User answers can be mapped to a 47-indicator framework across pain, red flags, psychological factors, functioning, sociodemographic context, and somatic information. Some indicators are extracted with confidence levels and may need validation.
Layer 4
The prototype can explore low-risk, moderate-risk, high-risk, red-flag, yellow-flag, and complex pathway patterns. This helps show how clinical reasoning could be structured, but it is not a validated clinical decision engine.
Prototype modules
Behind the interface are separate modules for safety, indicator tracking, assessment logic, exercise support, voice, and summaries. That makes PT-CHARLIE useful as a product and research prototype, while still requiring professional oversight before any clinical use.
Prototype module
A safety layer gives red flag findings priority. If risk signals are present, the prototype is designed to stop normal physiotherapy-style guidance and move toward medical evaluation messaging.
Prototype module
The code includes an indicator framework for chronic low back pain with domains for pain, red flags, psychology, functioning, and context. It helps make reasoning traceable instead of hidden in a black box.
Prototype module
The prototype explores STarT Back, ODI, risk profiles, exercise parameters, behavioral interventions, and shared decision-making prompts. These are implementation ideas, not claims of regulatory validation.
Prototype module
The app also contains walking progression, ergonomic advice, voice interaction, text-to-speech support, and patient summary flows. These show the product direction beyond a single chat screen.
Conversation first
A user explains what happened, where the pain is, how long it has been there, what makes it worse, and what daily life is being affected. The prototype feels like a conversation, while the structure underneath explores clinical reasoning patterns.
Live extraction
Pain duration, location, radiation, work status, sleep, movement fear, medication, mood, and stress can be extracted as indicators while the conversation continues. This is prototype functionality for exploration, not a validated clinical decision tool.
Safety first
Charlie includes questions about serious signs such as cauda equina symptoms, fever, unexplained weight loss, fracture risk, infection, malignancy, and inflammatory patterns. These prompts are educational safety support and do not replace professional triage.
Patient archetypes
The app can match patterns such as persistent pain with fear avoidance, work stress, poor sleep, or high distress. That makes the next step more personal than a generic exercise PDF.
Voice mode
The app supports natural voice conversations. That matters for people who do not want a long form, students who want to hear clinical reasoning unfold, and clinicians testing how an intake could sound.
Guided next step
Charlie can explore pathway guidance, exercise support, ergonomic advice, progress checks, or referral prompts. Any real-world care decision still belongs with the user and an appropriate healthcare professional.
Example conversation
A person might say they have had low back pain for three months after lifting a heavy box at work. Charlie then asks about symptoms that could suggest risk, checks whether pain radiates, asks about sleep, stress, work, and movement confidence, and turns the answers into clinical indicators.
That is the difference between PT-CHARLIE and a generic health chatbot prototype: the useful part is the structured reasoning behind the conversation. The output still needs human interpretation and should not be treated as clinical advice.
Patient says
Charlie recognizes persistent low back pain rather than treating it as a simple acute episode.
Charlie asks
The conversation checks red flags before moving into exercise or reassurance.
System extracts
The important part is not only the pain score. Charlie also tracks the factors that often keep back pain going.
Next step
For example: gradual movement, education about pain, confidence building, ergonomic support, or a prompt to seek professional care when safety or complexity requires it.
Clinical signal map
Low back pain decisions are better when the system understands the whole picture: symptoms, risk, function, confidence, work, and daily life.
Signal domain
Duration, location, intensity, radiation, triggers, medication response, and flare-up pattern as user-reported information.
Signal domain
Safety questions that suggest when self-management content may be inappropriate and professional care should be considered.
Signal domain
Work, sleep, movement, daily activities, avoidance, and the practical effect of pain on normal life.
Signal domain
Fear of movement, stress, mood, confidence, distress, and beliefs about pain and recovery.
Signal domain
Symptoms, body region, physical pattern, and clinical details that shape the pathway.
Signal domain
Age, work type, activity level, health context, and the situation in which the back pain started.
Who it helps
PT-CHARLIE is written so a patient can use it without medical training, but it is structured deeply enough for physiotherapists, researchers, and students to inspect the logic behind the guidance.
That balance matters. If AI in healthcare is only impressive technically, it fails. If it is only simple, it becomes unsafe. PT-CHARLIE sits between those layers as a prototype: understandable on the front end, clinically structured underneath, and still clearly bounded as non-regulated exploratory software.
For patients
Charlie explains back pain concepts in plain language, asks safety-oriented questions, and helps people prepare better questions for a healthcare professional.
For physiotherapists
The product shows how patient stories can be turned into useful indicators, safety checks, and pathway suggestions without pretending to replace clinical judgment.
For researchers
PT-CHARLIE connects guideline logic, patient-reported information, clinical indicators, and interface design in an early prototype environment.
For students
Students can see how questions, red flags, psychosocial factors, and functional limitations change the direction of care.
Safety and boundaries
PT-CHARLIE is an early prototype. It is not MDR registered, not CE marked, and not offered as a regulated medical device.
The red flag prompts are there to make the prototype safer and more educational, not to replace emergency triage, diagnosis, treatment, or professional clinical judgment.
When the story appears suitable for physiotherapy-style support, Charlie can help a person understand possible themes: education, gradual movement, exercise, ergonomic advice, confidence building, progress tracking, or professional care when needed. The final decision should remain with the person and a qualified healthcare professional.
Ecosystem
It is one of the clearest examples of what Physiotherapy.ai and Avivly are exploring: practical AI may support better care workflows when the clinical logic, safety boundaries, and professional responsibility come first.
Early prototype
Dutch-first early prototype for AI-supported low back pain guidance, built around chat, voice, safety-oriented prompts, and clinical indicators.
Early public milestone
An early public version appeared in the January 2024 "Create Your Custom GPTs" hackathon, where G-PT Charlie showed the first direction of the product.
Connected system
Evidence-based practice support for clinical research, reasoning, and physiotherapy education.
Parent layer
The wider home for practical AI systems in physiotherapy, education, research, and care navigation.
Recognition
Recognition for innovation in first-line healthcare.
Care pathway
The prototype explores patient, professional, and referral use cases. It is not a regulated care pathway or a replacement for professional responsibility.
Explore PT-CHARLIE
Charlie can ask questions, check safety-oriented signals, and help explain possible next steps. It is exploratory guidance only: not a diagnosis, not treatment advice, not a medical device, and not a replacement for a doctor or physiotherapist.
PT-CHARLIE is an early prototype. It is not MDR or CE registered, not a medical device, and not intended for diagnosis, treatment decisions, emergency triage, or replacing a qualified healthcare professional.
PT-CHARLIE is being shaped as a practical support layer around triage, education, and next-step guidance. It keeps clinical boundaries visible while helping people move through decisions with less friction.
