1 of 17

A minimal set of screening activities� to characterize anterior cruciate ligament (ACL) injury risk

Kirsten Seagers

Krithika Swaminathan, Julie Kolesar, Sam Hamner, Julie Muccini,

Jennifer Hicks, Gregory Myer, Trevor Hastie, Scott Delp

​

July 31, 2025

2 of 17

ACL injury is highly prevalent and life altering

1

Introduction

7x increase

in risk of

Osteoarthritis4

[1] Samuelson et al., 2017; [2] Niekerk et al., 2023; [3] Webster et al., 2016; [4] Webster et al., 2022

>1.4 million incidents

per year worldwide1

>75% experience PTSD symptoms2

20% risk

of re-injury3

But >70%5 are non-contact injuries & are likely preventable6 with

appropriate screening + training

[5] Della Villa et al., 2020; [6] Myer et al., 2007

3 of 17

Biomechanical Injury Risk Factors and Assessments

2

Introduction

Drop Jump

Hewett et al., 2005

Dynamic Valgus

Midline

Knee Abduction

Ankle Eversion

Femoral Adduction

Weir et al., 2021

Trunk Lean

Stiff Knee

Ankle Angle/

Foot Placement

Dynamic Knee Valgus Posture

Cutting

4 of 17

Biomechanical Injury Risk Factors and Assessments

2

Introduction

Collings 2022

Hopping

Hewett 2005

Jumping

Boden 2009

Cutting

Hewett 2009

Cutting

5 of 17

Identify a minimal set of screening activities that maintain comprehensive characterization of mechanistic risk factors

3

Purpose

6 of 17

Experimental Data Collection

4

Methods

[3] Schweizer et al., 2022

Motion capture + ground reaction force data from 27 adolescent female athletes

force plate

16-camera

motion capture

Accelerate

Accelerate

Jump

Decelerate

Measurements obtained during five commonly-studied activities3

7 of 17

Biomechanical Modeling

5

Methods

[4] Seth et al., 2018

Calculated 35 biomechanical risk factors using musculoskeletal modeling, with 2-9 factors per activity

Drop Jump

Cutting

4

8 of 17

Column Subset Selection5 (CSS)

6

[5] Hastie et al., 2020

Methods

Identify a set of activities that best reconstructs the biomechanical risk features from the left-out activities

Risk Factors from

Left-out Activities

Risk Factors from

Included Activities

Reconstructs

 

 

Data Split

All Activity Risk Factors

 

9 of 17

Column Subset Selection5 (CSS)

6

[5] Hastie et al., 2020

Methods

Identify a set of activities that best reconstructs the biomechanical risk features from the left-out activities

 Normalized 

RMSE (NRMSE) per test subject

Step 1

Model Training:

Leave One Subject Out Cross Validation

Risk Factors from

One Activity

Risk Factors from Left-out Activities

Risk Factors from

Second Activity

Risk Factors from Left-out Activities

 

 

 

 

Step 2

Model Training:

Leave One Subject Out Cross Validation

Risk Factors from

Best One Activity

Average 

NRMSE per test subject

 

 

10 of 17

Optimal Activity Set:

7

Results

11 of 17

Optimal Activity Set:

7

Results

12 of 17

Optimal Activity Set:

7

Results

13 of 17

Optimal Activity Set:

7

Results

14 of 17

Optimal Activity Set: Run Cuts & Single-leg Drop Jumps

7

Results

15 of 17

Kinematics-only Optimal Activity Set:

8

Results

16 of 17

Kinematics-only Optimal Activity Set:

8

Results

17 of 17

Conclusions:

9

​

  • All activities are required if only kinematic risk factors are measured

kseagers@stanford.edu

​

​

  • Double-leg drop jumps provide minimal unique biomechanical risk information
  • Two activities—run cuts & single-leg drop jumps—capture the kinematic & kinetic injury risk information from common screening activities

Accelerate

Decelerate