8
Personalized AI-based Exercise Therapy for Oncology and Physiotherapy Patients

As the prevalence of cancer and chronic physical conditions continues to rise globally, the demand for personalized, effective and scalable rehabilitation strategies becomes increasingly critical. Exercise therapy has long been recognized as a cornerstone of physiotherapy and supportive cancer care, improving functional capacity, reducing treatment-related side effects and enhancing patients’ quality of life. However, conventional therapy approaches often rely on standardized protocols that lack adaptability to individual patient needs, progress and physiological differences. The integration of artificial intelligence (AI) into exercise therapy presents a transformative solution. By leveraging machine learning algorithms, computer vision, wearable technologies and real-time data analytics, AI can design and deliver highly individualized exercise regimens. These systems dynamically adapt to patient feedback, physiological signals, clinical parameters and personal goals, thereby optimizing therapeutic outcomes. In oncology, AI-based exercise therapy can address cancer-related fatigue, monitor post-treatment recovery and support immune health. In general physiotherapy, AI tools can facilitate rehabilitation from musculoskeletal injuries, neurological conditions and post-surgical recovery through data-driven guidance and real-time monitoring.

This chapter explores the technological architecture, clinical applications, benefits and challenges of AI-powered personalized exercise therapy across oncology and physiotherapy settings. We discuss current innovations, clinical evidence, ethical considerations and future directions, emphasizing the critical role of multidisciplinary collaboration in the development and deployment of these systems. While the technology offers enormous promise, challenges such as data privacy, model bias and system integration must be addressed to ensure equitable and effective care. Personalized AI-based exercise therapy stands as a paradigm shift in patient-centered rehabilitation, offering the potential for scalable, adaptive and evidence-based care that aligns with the precision medicine model of the 21st century.

8.1. Introduction

Cancer and musculoskeletal (MSK) disorders represent two of the most pressing healthcare challenges worldwide, significantly contributing to disability, diminished quality of life and healthcare expenditure. Cancer survivors often experience a multitude of side effects from treatments such as chemotherapy, radiotherapy and surgery, including fatigue, muscle wasting, joint stiffness, neuropathy and reduced cardiorespiratory capacity (Emery et al. 2022). Likewise, individuals with MSK conditions, including arthritis, fractures, spinal cord injuries and post-operative complications, frequently require extended rehabilitation to restore mobility, strength and functional independence.

Rehabilitation for these populations is typically centered on physiotherapy and structured exercise therapy. These interventions have been shown to reduce symptom burden, improve functional status and enhance psychological well-being. However, traditional approaches often adopt standardized, one-size-fits-all models that may not fully consider patient-specific variables such as cancer stage, treatment history, pain levels, comorbidities or progression rates. Such generalization can limit therapeutic effectiveness and patient engagement, particularly in complex or fluctuating conditions.

The rise of artificial intelligence (AI) in healthcare presents a groundbreaking opportunity to transform these limitations into strengths (Aminizadeh et al. 2024). With the ability to process large volumes of multi-dimensional data – including biometric signals, imaging results, behavioral patterns and clinical parameters – AI can facilitate real-time personalization of exercise programs. Through machine learning (ML), computer vision (CV), wearable sensors and decision-support systems, AI can monitor patient responses, predict outcomes and dynamically adjust therapy protocols to meet evolving needs.

This chapter explores the convergence of AI technologies with exercise therapy in the domains of oncology and physiotherapy. Specifically, it examines how personalized AI-based exercise therapy can be designed, implemented and scaled to support individualized rehabilitation plans (Deshmukh 2023). This chapter also investigates current innovations, real-world applications, benefits, limitations and future opportunities, emphasizing the potential of AI to redefine patient-centered care and enhance long-term outcomes.

8.2. Background

Personalized rehabilitation is a cornerstone of modern therapeutic care, particularly in complex domains such as oncology and physiotherapy. As healthcare transitions from generalized treatment approaches to precision medicine, it becomes imperative to understand the diverse needs of individual patients and how evolving technologies – especially AI – can support that transition (Ghanem et al. 2024). This section outlines the clinical necessity for personalized rehabilitation in oncology and MSK care while identifying the limitations of traditional exercise therapy that have paved the way for AI-driven solutions.

8.2.1. Oncology and physiotherapy: the need for personalization

Cancer and physiotherapy patients often present multifaceted and dynamic clinical profiles that demand individualized care strategies. In oncology, patients undergoing chemotherapy, radiotherapy or surgical interventions frequently experience a spectrum of debilitating side effects, including cancer-related fatigue (CRF), muscle wasting (cachexia), neuropathies, lymphedema and cardiopulmonary dysfunction. These effects not only compromise physical function but also reduce psychological well-being and hinder participation in everyday activities. Exercise therapy, when appropriately prescribed, has been shown to mitigate many of these side effects, enhance immune response, reduce the risk of cancer recurrence and improve overall survival and quality of life.

Similarly, individuals suffering from MSK and neurological conditions – such as osteoarthritis, stroke, spinal cord injuries and orthopedic trauma – require long-term rehabilitation to regain motor control, balance and physical independence (Lo et al. 2021). However, functional recovery in such cases is highly variable and influenced by factors such as age, comorbidities, injury severity and baseline fitness. In both patient groups, a one-size-fits-all rehabilitation approach is inadequate.

Personalization in therapy – tailoring exercise intensity, frequency, type and progression to the specific needs and limitations of each patient – is essential for optimizing outcomes. This level of individualized care, while theoretically ideal, is challenging to achieve consistently in conventional clinical settings due to constraints in time, workforce and resources. This gap has stimulated the exploration of AI as a tool to support and automate the delivery of personalized care at scale.

8.2.2. Limitations of traditional exercise therapy

Traditional exercise therapy programs, although grounded in clinical expertise and evidence-based practices, often exhibit systemic limitations that hinder their effectiveness (Chaughule 2024). Among the key shortcomings are:

Lack of real-time feedback: in standard physiotherapy settings, patients typically receive instructions during in-person visits and are expected to continue exercises independently at home. This disconnection results in poor exercise fidelity and an inability to correct improper techniques, which can lead to suboptimal outcomes or even injury.
Generalized treatment protocols: many rehabilitation plans are designed using population-level data and do not account for individual variability in recovery speed, tolerance, preferences or comorbid conditions. Such generalization may overlook nuanced patient needs and result in decreased efficacy or engagement.
Limited monitoring outside clinical settings: once discharged or between sessions, patients often perform exercises unsupervised, and clinicians lack visibility into their progress, adherence or challenges. This absence of remote oversight limits timely interventions and adjustments.
Difficulty in long-term adherence: sustained adherence to rehabilitation programs is notoriously low, especially when patients do not perceive immediate benefits or when they encounter pain, fatigue or logistical difficulties (Shaw 2015). Without continuous support or motivation, long-term compliance suffers.

These limitations create a pressing need for a solution that can bridge the gap between clinical supervision and independent rehabilitation. AI, through its capabilities in sensing, interpreting and responding to data in real time, holds significant promise in transforming how personalized exercise therapy is delivered.

AI systems can track patient performance continuously, provide adaptive recommendations, issue real-time corrections and maintain patient motivation through interactive feedback mechanisms.

8.3. Technological framework

The integration of AI into personalized exercise therapy relies on a robust technological infrastructure that brings together advanced software, multidimensional data inputs and smart hardware (Mikolajewska et al. 2025). AI-based therapeutic systems function as intelligent, adaptive platforms that assess patient-specific data, interpret real-time physiological and biomechanical signals and deliver personalized exercise recommendations dynamically.

This section outlines the foundational AI technologies, essential data inputs and hardware ecosystem that collectively enable the successful deployment of AI-driven exercise therapy in oncology and physiotherapy.

8.3.1. Key AI technologies

8.3.1.1. ML

ML is the backbone of most AI applications in healthcare. It enables systems to analyze large datasets, identify patterns and predict outcomes based on patient-specific variables (Lorenzo et al. 2024). In exercise therapy, ML algorithms are used to:

predict individual recovery trajectories based on treatment history and response;
classify patient types and risk profiles;
continuously optimize exercise prescriptions through ongoing data input.

Supervised learning can be used for initial patient profiling, while unsupervised learning can help cluster similar rehabilitation patterns. Deep learning, a subfield of ML, further enhances the system’s ability to recognize complex patterns in movement and biometric data.

8.3.1.2. CV

CV plays a pivotal role in monitoring and evaluating patient movement. By processing data from cameras or wearable motion sensors, CV algorithms can:

detect and assess posture, joint angles and motion trajectories;
identify deviations from ideal movement patterns;
provide real-time corrective feedback to prevent injury or compensate for impairment.

This technology is especially beneficial in home-based rehabilitation settings, allowing for precise movement assessment without the need for continuous human supervision (Sassi et al. 2024).

8.3.1.3. Natural language processing

Natural language processing (NLP) enables AI systems to understand and respond to spoken or written language, allowing patients to communicate naturally with digital therapeutic assistants. NLP functionalities include:

voice-controlled exercise instructions;
conversational feedback and motivational support;
automatic generation of progress reports based on user input.

This enhances accessibility for elderly patients or those with visual or motor impairments, facilitating a more intuitive and user-friendly interface.

8.3.1.4. Reinforcement learning

Reinforcement learning (RL) allows AI systems to adapt exercise recommendations based on trial-and-error learning (Doherty et al. 2024). The algorithm receives feedback based on patient performance and modifies its strategy to maximize therapeutic benefit. In this context, RL can:

adjust exercise intensity, duration and complexity in response to adherence and outcome data;
reward consistent patient engagement with tailored motivational prompts;
learn from diverse user profiles to enhance personalization across patient populations.

Together, these AI technologies form a multidimensional, responsive system capable of delivering dynamic and personalized rehabilitation programs that evolve in real time.

8.3.2. Data inputs for personalization

The personalization of exercise therapy hinges on the quality, variety and granularity of data fed into the AI system. Key data inputs include:

Patient demographics: age, gender, weight, height and socioeconomic background influence baseline fitness, risk factors and engagement styles (Simmons et al. 2014).
Cancer type/stage or injury type: AI systems tailor therapy differently for patients with breast cancer, colorectal cancer or MSK injuries based on clinical guidelines and individualized risk profiles.
Treatment history: information about past surgeries, chemotherapy, radiation exposure or previous physiotherapy outcomes helps inform exercise tolerances and contraindications.
Functional mobility scores: data from validated tools such as the Timed Up and Go (TUG), 6-Minute Walk Test (6MWT) or manual muscle testing informs baseline function and progress (Fessele and Syrkin 2024).
Pain and fatigue levels: real-time tracking of pain (via numerical rating scales or self-reports) and fatigue (using PROMs or wearable proxies) helps adjust intensity to avoid overexertion.
Wearable sensor data: biometric data such as heart rate variability, respiratory rate, gait symmetry and step count offer continuous insight into physical performance and recovery dynamics.

The integration of these diverse data streams allows AI to make precise, evidence-based decisions in real time, ensuring therapy remains appropriate and effective at every stage of recovery.

8.3.3. Hardware ecosystem

The delivery of AI-based personalized exercise therapy is supported by a diverse range of hardware devices that collect, process and display data (Khalid et al. 2024). Key components include:

8.3.3.1. Wearables

Wearable technologies such as smartwatches, fitness bands, chest straps and sensor-embedded clothing capture real-time physiological data.

Key metrics include:

heart rate and variability;
blood oxygen saturation (SpO₂);
step count and cadence;
sleep quality and activity levels.

These metrics are used to assess physical readiness, recovery status and adherence to prescribed activity levels.

8.3.3.2. Cameras and motion sensors

High-resolution cameras, depth sensors (e.g. Microsoft Kinect) and inertial measurement units (IMUs) capture and interpret body movements, providing:

joint angle measurements;
motion path tracking;
balance and coordination assessments.

This hardware enables real-time movement correction, performance evaluation and long-term progress tracking (Saez et al. 2018).

8.3.3.3. Smartphones and tablets

These devices serve as the central interface for patient interaction with the AI system. Their roles include:

displaying exercise routines via videos or augmented reality (AR);
delivering audio-visual feedback;
hosting virtual consultations with healthcare providers;
collecting self-reported symptoms, pain levels and adherence logs.

8.4. Application in oncology

Oncology patients experience a diverse array of symptoms and functional impairments that vary widely based on cancer type, stage, treatment modality and individual response (Reeve et al. 2014

Only gold members can continue reading. Log In or Register to continue

Stay updated, free articles. Join our Telegram channel

Tags: AI driven Innovations in Physiotherapy and Oncology 1

Mar 15, 2026 | Posted by admin in ONCOLOGY | Comments Off

Oncohema Key

Fastest Oncology & Hematology Insight Engine

Personalized AI-based Exercise Therapy for Oncology and Physiotherapy Patients

8
Personalized AI-based Exercise Therapy for Oncology and Physiotherapy Patients

8.1. Introduction

8.2. Background

8.2.1. Oncology and physiotherapy: the need for personalization

8.2.2. Limitations of traditional exercise therapy