Posts

Artificial intelligence and rehabilitation medicine: the revolution has begun

/0 Comments/in /by

Artificial intelligence (AI) is already a concrete reality at ISICO, where it is actively used in daily clinical practice, as a research tool in studies published in scientific journals, and as part of the European PREPARE project, in which our institute is a partner. AI is fundamentally transforming the work of rehabilitation professionals, opening up new possibilities in early diagnosis, patient monitoring, treatment personalisation, and therapeutic education.

The potential of AI is immense. In a context where the volume of clinical, instrumental and functional data is growing exponentially, intelligent algorithms allow for the rapid and efficient analysis of this information, turning it into tangible support for clinical decision-making.

“We are only at the beginning of a transformation that will become increasingly far-reaching,” explains Francesco Negrini, physiatrist at ISICO and Associate Professor at the University of Insubria.
“Just as the first computers revolutionised our work in the 1980s, AI is set to become an indispensable tool in the daily work of clinicians.”

In recent years, research on AI in the medical field has accelerated dramatically. Between 2022 and 2023 alone, funding in this sector increased eightfold, reaching $25 billion. Applications range from imaging diagnostics and outcome prediction to motor rehabilitation support and functional assessment.

ISICO has contributed to this evolution. One example is a 2023 study on the use of AI to improve scoliosis screening. The team developed an algorithm capable of combining multiple variables (ATR, age, sex, BMI, family history, prominence height, curve location) to detect the condition early — going beyond the limitations of traditional tools like the scoliometer.

But AI’s applications in rehabilitation go far beyond scoliosis. Through data collection from wearable sensors, surface electromyography, and robotic devices, it is now possible to analyse patients’ movements in real time, support the monitoring of gait and speech disorders, classify levels of motor dysfunction, and adapt treatment plans in a truly personalised way.

At ISICO, AI is also used to enhance the patient experience through projects such as ISAIA (ISICO AI Assistant) — a virtual assistant designed to improve communication between clinicians and patients, answer frequently asked questions, and support the management of emails and FAQs. This approach helps maintain an informative and educational dialogue beyond the consultation, responding to patients’ ongoing concerns during treatment.

Another promising frontier is the integration of AI into telerehabilitation. Smartwatches and dedicated apps now make it possible to monitor home exercises with precision, optimising therapeutic pathways remotely. One recent study, for instance, demonstrated that AI could automatically recognise post-stroke rehabilitation exercises with over 99% accuracy.

Of course, with opportunity comes challenge. The handling of sensitive data, the lack of transparency in algorithms (often still “black boxes”), the risk of over-reliance on automated solutions, and the quality of training data are all critical issues requiring careful oversight and regulation. A responsible adoption of AI must be grounded in strong clinical supervision, ongoing human oversight, and the scientific quality of the sources used.

In conclusion, artificial intelligence will not replace clinicians — it will support them, helping them become more accurate, more efficient, and better equipped to offer truly personalised care.
At ISICO, that revolution has already begun.

The importance of observational studies

/0 Comments/in /by

Our study, Observational Studies: Specific Considerations for the Physical and Rehabilitation Medicine Physician, was recently published in the American Journal of Physical Medicine & Rehabilitation. 

Experimental and non-experimental designs are used to investigate the effect or association of an intervention and clinical or surrogate outcome. These methods aim to improve knowledge and develop new strategies to manage a disease or condition. 

While experimental research studies entail scrutiny by the scientist and provide results that are less prone to systematic errors, their downside is that they are poorly generalisable. “What all this means in clinical terms,” explains Dr Sabrina Donzelli, physiatrist at Isico and first author of the published research,” is that a treatment that worked fine during a study may in the long term, following its prescription by a hospital or general practitioner, throw up problems that did not emerge in the experimental research”. 

Therefore, to verify what happens in the real world, non-experimental studies, called observational studies, can be carried out, of the kind dealt with by the research we have just published. 

Well-designed observational studies can provide valuable information regarding exposure factors and the event under investigation. 

“Basically, what the researcher does is simply observe data, without having the possibility to manipulate it”, Dr Donzelli goes on. “The researcher’s task is to interpret and contextualise the results, taking into account all potential errors introduced during the selection of the study sample. To eliminate, as far as possible, systematic errors that could lead to incorrect evaluations and interpretations, it is necessary to implement a series of methodological strategies that are not very widespread in the rehabilitation field.” 

In physical and rehabilitation medicine, where complex procedures and multiple risk factors can be involved in the same disease, the use of observational study must be planned in detail and a priori to avoid overestimations. 

“This is why we wrote this article, to offer clear suggestions to researchers in the rehabilitation field who are interested in planning an observational study”, concludes Dr Donzelli. “We give an overview of the methods used for observational design studies and describe when it is appropriate to use them and how to do so in different scenarios”.