Data-driven approaches for predicting Tommy John surgery risk in major league baseball pitchers

국문 초록 (Abstract)

부상 관리는 모든스포츠에서선수의경기력에 직접적인영향을미치는 핵심소다. 특히 야구는 주당 5˜7경기를 치르기 때문에 선수들의몸에 지속적인부담이가해져부상 위험이높 다. 그중에서도메이저리그(MLB) 투수들이겪는 대표적부상인토미 존수술(Tommy John Surgery, TJS)은심각한위협이된다. 기존의TJS 예측방법은센서나 비디오기반모션 캡처에 의존해실제경기 상황에서적용이어렵고, 부상 발생 30구이내처럼 너무 임박해서야 예측이가능하다는 한계가있었다. 본연구는 이러한문제를 해결하기 위해분류와회귀를 결합한딥러닝(DL) 프레임워크 를 제안한다. 2016-2023년MLB 투구데이터를 활용해, 분류 모델은최대100일전에 TJS 위험을탐지해F1-스코어 0.73의높은성능을보였고, 회귀모델은선수의마지막 수술전 경기까지남은일수를 R² 0.79로예측했다. 더나아가설명 가능한인공지능 기법을적용해 포심패스트볼(직구)의릴리스포인트가낮아질수록척골측측부인대(UCL)가빠르게약화돼 TJS 위험이증가한다는 등주기계적인을밝혀냈다. 이연구결과는 투수의건강을보호하고복잡한외과적처치를 줄이기 위한조기 개입 전략 수립에 실질적기반을제공한다.

다국어 초록 (Multilingual Abstract)

Injury management is critical in all sports, directly impacting player performance.
Baseball players are particularly susceptible to injuries, as players often compete in 5 to 7
games per week, placing continuous strain on their bodies. Among various injuries, Tommy
John Surgery (TJS) poses a notable risk for Major League Baseball (MLB) pitchers.
Traditional TJS prediction methods required sensors or video-based motion capture, which
are impractical during actual games and limited in making predictions too close to the
injuries, such as within 30 pitches. To address these challenges, this study proposes a
deep learning (DL) framework that utilizes both classification and regression tasks. Using
MLB pitching data (2016–2023), the classification model detects injury risk up to 100
days in advance with a high prediction performance of 0.73 F1-score, while the regression
model estimates the time remaining until the player’s last pre-surgery game with R2 of
0.79. In addition, to enhance our model’s applicability, we employ an explainable artificial
intelligence technique to analyze the impacting mechanical features, such as a lowered
four-seam fastball release point, that accelerate UCL deterioration, increasing TJS risk.
These findings provide a practical foundation for early intervention strategies, potentially
preserving pitcher health and reducing the need for complex surgical procedures.

목차 (Table of Contents)

• 1 Introduction 1
• 2 Related Work 5
• 2.1 Traditional Studies about Pitchers’ Elbow Injury 5
• 2.2 ML/DL-Based Studies on Pitchers’ Injuries 7
• 3 Dataset 9
• 3.1 Data Source 9
• 3.2 Preprocessing 10
• 4 Method 13
• 4.1 Classification Framework for TJS Prediction 14
• 4.2 Regression Framework for Predicting Days Until Injury 17
• 5 Experiment 19
• 5.1 Classification Task 19
• 5.1.1 Model Training and Loss Function 19
• 5.1.2 Evaluation Metrics 20
• 5.1.3 Hyperparameters for Classification 21
• 5.2 Regression Task 22
• 5.2.1 Model Training and Loss Function 22
• 5.2.2 Evaluation Metrics on Regression 23
• 5.2.3 Hyperparameters for Regression 24
• 6 Results 25
• 6.1 Results on Classification 25
• 6.2 Results on Regression 28
• 6.3 Interpretation using SHAP values 29
• 6.3.1 The Movement of the Ball 32
• 6.3.2 The Physical Changes in the Pitcher 34
• 6.3.3 Relating to TJS 37
• 7 Discussion and Conclusions 39
• 7.1 Limitations and Future Works 39
• Ethical Considerations 41
• Real-Time Applicability and Future Directions 41
• Clinical Applicability and Complementary Diagnostics 41