Simone Biles' triumphant return to competitive gymnastics this week is a reminder that even the most accomplished athletes face major setbacks. When asked about the key to a successful comeback, many athletes reflexively say the same thing as startup founders: grit. But faced with debilitating injury, pushing through isn't always advisable. Recuperation and reinvention takes time, expertise, community, and yes, technology.
So to fellow technologists contemplating life after the Olympics: the inspiration doesn't have to stop when the games end. There are at least four ways we can use AI and machine learning right now to transform sports medicine and athletic training—not only for elite athletes like Simone, but for anyone recovering from setbacks and getting back in the game.
1. Real-time performance monitoring—a coach always by your side
Consider the use case of Emily, an amateur triathlete who's fiercely determined, but struggled to find her optimal pace during races. She often pushed too hard early on, only to fade as the finish line approached.
To hit her stride, Emily could try a AI-powered wearable designed specifically for endurance athletes. By combining real-time data with personalized insights and a focus on triathlon-specific performance metrics, an AI-powered wearable could guide Emily through tailored training and apply relevant racing strategies.
She could literally have a portable personal coach on hand, helping her adapt to the unique demands of each triathlon leg—and achieve new personal bests.
Here's how this insightful technology might work in practice.
Relevant wearable technology for performance monitoring:
Heart rate monitoring. A chest strap, Apple Watch, or optical heart rate sensor can continuously measure heart rate, providing real-time feedback on exertion levels.
GPS tracking. Integrated GPS can track pace, distance, and elevation changes during runs and bike rides.
Accelerometer and gyroscope. Sensors can measure movements and body position, enabling analysis of running gait, swimming stroke efficiency, and cycling cadence.
Applicable AI/ML algorithms:
Real-time data streaming & analysis. Data from the wearable sensors can be streamed to a mobile app or cloud platform, where AI algorithms process and analyze it in real time.
Physiological modeling. Heart rate variability (HRV) analysis can estimate fatigue and recovery levels, helping to gauge an athlete's readiness to train or compete.
Biomechanical modeling. Algorithms can assess running gait, swimming stroke, and cycling technique, identifying areas for improvement and real-time feedback to optimize performance before and during a race.
2.Personalized rehabilitation programs—how AI can help you stage a comeback
Let's look at another use case: Kenji, a saber fencer with lightning-fast reflexes—and a nagging wrist injury. The usual physical therapy exercises may feel like a generic training regimen, not the personalized attention he needs to return to championship form.
But what if Kenji could access an AI-powered rehabilitation program designed specifically for fencers? With conversational AI, Kenji could interact with a virtual coach who understands the unique stresses and strains of his sport.
By combining cutting-edge technology with personalized insights and a focus on saber-specific movements, the AI-powered rehabilitation coach can help guide Kenji through a tailored recovery journey. Kenji can strengthen his wrist and even refine his technique, so that he can rejoin his sport at a higher level.
Here's how this cutting-edge rehabilitation technology might work in practice.
For data collection:
Wearable sensors. High-frequency inertial measurement units (IMUs) integrated into a wristband can capture kinematic data (3-axis acceleration, gyroscope, and magnetometer) during sabre-specific movements like flicks, lunges, and parries. This data may be used to quantify wrist range of motion, angular velocity, and acceleration profiles, identifying deviations from normative ranges and subtle compensations in technique.
Electromyography (EMG). Surface EMG sensors can be placed on key muscles of the forearm and wrist (e.g., flexor carpi ulnaris, extensor carpi radialis) to measure muscle activation patterns during sabre actions. This data can help identify muscle imbalances, fatigue, and aberrant activation patterns that could hinder recovery or contribute to re-injury.
Patient-reported outcomes (PROMs). Standardized questionnaires, such as the Patient-Rated Wrist Evaluation (PRWE) and the Disabilities of the Arm, Shoulder, and Hand (DASH) questionnaire, were administered regularly to assess Kenji's perceived pain, function, and quality of life, providing valuable subjective feedback to complement the objective data.
Useful AI/ML algorithms:
Biomechanical modeling. A personalized musculoskeletal model of the wrist can be created using kinematic and EMG data. This model could simulate joint forces, torques, and muscle activations during specific sabre fencing movements (e.g., flèche attack, parry-riposte), enabling the identification of high-stress areas and potential injury mechanisms.
Reinforcement learning (RL). A model-free RL algorithm, Proximal Policy Optimization (PPO), can learn to develop an optimal exercise prescription policy. The algorithm can iteratively refines the plan based on performance and feedback data, maximizing recovery potential while minimizing the risk of re-injury. The reward function was tailored to prioritize sabre-specific movements and functional outcomes.
Clustering. Unsupervised learning techniques such as Gaussian Mixture Models (GMMs) can be applied to cluster data with that of other sabre fencers. This helps to identify subgroups of athletes with similar injury profiles and response patterns, allowing the platform to leverage collective knowledge for personalized recommendations.
3. Predictive injury analytics—using data to prevent injury
Consider the case of Miguel, a high school soccer star prone to hamstring injuries. At this point, his fears of getting sidelined by another injury are holding him back from playing his best game.
An AI system can collect useful insights to help Miguel raise his game without increasing his risk of injury. Different AI algorithms can help analyze his movement on the field, scrutinize training data, and factor in sleep patterns. Sifting through this data, AI may detect a risk: subtle changes in Miguel's gait, an uptick in his training intensity, and insufficient rest could signal heightened risk of a hamstring strain.
Given this AI-generated early warning, coaching staff can proactively adjust Miguel's training plan, incorporating targeted stretches and adding exercises to reinforce his hamstrings. The AI system can help prevent injury, so that Miguel can continue his winning streak.
Let's look at how this cutting-edge predictive technology might work in practice.
For data collection:
Wearable sensors. Inertial measurement units (IMUs) embedded in cleats or shorts can track running mechanics, including stride length, ground contact time, and hip/knee joint angles.
GPS tracking. GPS data can capture overall training load, including distance covered, speed, and acceleration/deceleration patterns.
Heart Rate Variability (HRV). HRV data collected through a heart rate monitor can provide insights into an athlete's autonomic nervous system balance, and track overall recovery status.
Sleep tracking. A wearable device monitoring sleep duration and quality can factor in these essential variables for recovery and injury prevention.
Useful AI/ML algorithms:
Time series analysis. Recurrent Neural Networks (RNNs) or Long Short-Term Memory (LSTM) networks can be used to model the temporal dependencies in biomechanical and physiological data, identifying trends correlated with previous hamstring injuries.
Anomaly detection. Algorithms like Isolation Forest can identify unusual patterns or deviations from baseline metrics, signaling increased injury risk.
4.Virtual coaching and physical therapy—whenever and wherever you need it
Another use case applies here: Adriana is a rising tennis star whose powerful serves are starting to be impacted by persistent shoulder injury. But between matches, she lives with her family in the mountains, far from specialized sports medicine clinics. For Adriana, traditional physical therapy seems out of reach.
Adriana could benefit from an AI-powered virtual physical therapy platform designed specifically for tennis players. With a focus on tennis-specific movements and personalized data insights, this virtual therapy platform could bring tailored rehabilitation to Adriana's living room to strengthen her shoulder and raise her game.
Here's how this portable physical therapy technology might work.
For computer vision:
Motion capture. Using a mobile-device camera to capture movements, an app could run advanced pose estimation algorithms like OpenPose or MediaPipe in real time to suggest posture adjustments.
Landmark identification. The deep learning models within these algorithms can identify and track 2D or 3D coordinates of key anatomical landmarks, such as joints and limbs. They can compile a digital skeletal representation of an athlete, revealing what's happening at the musculoskeletal level as the athlete moves.
Kinematic analysis. Skeletal data can be processed using kinematic analysis to analyze joint angles, velocities, and accelerations. This helps athletes adjust their movements to optimize therapeutic exercises.
AI/ML algorithms:
Form assessment. Extracted pose data can be fed into a multi-layered AI engine trained on extensive datasets of correct tennis form. These learning models can then assess Adriana's movements in real time, comparing her form to reference data and flagging deviations from recommended techniques.
Real-time feedback. An AI engine can generate corrective feedback, highlighting areas for improvement. With visual cues or conversational prompts, AI can guide athletes towards proper technique.
Personalized adaptation. Reinforcement learning algorithms can dynamically adjust exercise regimes, according to an athlete's personal performance and progress. Parameters such as exercise selection, repetitions, intensity, and rest periods may be jointly monitored and continuously optimized by the athlete, their coaches and medical team. This helps maximize therapeutic benefits and minimize re-injury risks.
Simone Biles injured after the Paris 2024 Olympics preliminary round—before winning multiple gold medals
These AI-powered advancements promise to empower athletes of all levels, turning setbacks into comebacks. Recently, a minor hand injury sidelined me from archery for months—but Simone Biles' example helped me to refocus, recover, and return to the archery range. Great athletes like Simone can lift our spirits and change our outlook—and as medical technologists, we can return the favor.
With inspiration from Simone Biles and insights from AI and machine learning, we can unlock new possibilities for personalized care, injury prevention, and enhanced performance monitoring. There will always be unexpected twists along the path to recovery and peak performance—but we can bounce back stronger.