HearTMD: Detection of Temporomandibular Disorder Using Hearables

The ACM Symposium on User Interface Software and Technology (UIST 2025)

1. HearTMD:
Detectionof Temporomandibular Disorder Using Hearables
01 Introduction
02 Overview
Temporomandibulardisorder(TMD):
dentaldiseasewithpain,jointnoise,
limitedjawmovement
Background Problem Approach
04 Result
AUC: 0.700
Sensitivity: 0.641
Specificity: 0.672
05 Discussion
Explore daily-life screening
for higher usability
06 Future Work
•Refine feature extraction through further analysis and expert
insights
•Validate models with actual patient data as screening tools
•Explore daily-life screening for higher usability
Novalidatedlow-cost,location-independentmethod
•CT/MRIiscostlyandfacility-bound.
•Thermography/EMGrequiresspecialtoolsand
controlledsettings.
DetectingTMDusinghearablesfordailyuse
•IMUsensorstocapturejawmovements
AUC: 0.414
Sensitivity: 0.150
Specificity: 0.908
AUC: 0.457
Sensitivity: 0.275
Specificity: 0.781
AirPods Pro
Participants: 11 with TMD subjective symptoms* / 18 without
Tasks: maximum mouth opening / lateral jaw movement
Data: angular velocity / acceleration / gravitational acceleration
Features
•Statistical : raw data / left-right differences
•Spectral : Peak frequency & amplitude / Top 10 component amplitudes
Maximum mouth opening Lateral jaw movements Both movement
Potential for TMD detection with maximum mouth opening
(Cutoff: Youden Index)
Feature importance
Detection performance for each dataset
*As defined in Background
HiyoriTsuji
1)
, Takashi Amesaka
2)
, Sho Usuda
1)
, Wataru Muraoka
1)
, TaneakiNakagawa
1)
, Yuta Sugiura
1)
1)
Keio University,
2)
The University of Osaka
Contact [email protected]