Ultrasound-based Models for Identifying Sarcopenic Dysphagia: The Key Role of Digastric Muscle.

OBJECTIVES

Sarcopenic dysphagia (SD) is a serious condition in older adults lacking reliable diagnostic tools. This study aimed to develop and validate ultrasound-based diagnostic models for SD by integrating clinical and radiomics features from swallowing muscles (geniohyoid, digastric, tongue) to identify the most informative muscle group for diagnosis.

DESIGN

A diagnostic accuracy study.

SETTING AND

PARTICIPANTS

We recruited 377 nursing home residents aged ≥65 years, with 274 in the training set and 103 in the external validation set.

METHODS

SD was diagnosed using a standardized diagnostic algorithm. Ultrasound images of the geniohyoid (GH), digastric (DM), and tongue (TM) muscles were used to extract 840 ultrasound radiomics features. After feature selection using correlation analysis and least absolute shrinkage and selection operator (LASSO) regression, logistic regression models were built combining clinical and radiomics features. Models were evaluated using the area under the receiver operating characteristic curve (AUC), calibration, and decision curve analysis, with internal (bootstrap) and external validation.

RESULTS

SD was diagnosed in 104 (27.6%) participants. Eight diagnostic models were developed and validated. The model incorporating radiomics features from the digastric muscle (DM model) demonstrated the strongest diagnostic performance, achieving an AUC of 0.78 in internal validation and 0.75 in the external validation set (sensitivity 0.46, specificity 0.87). Adding radiomics features from other muscles (GH or TM) to the DM model did not significantly improve performance (DeLong test, all P > .05).

CONCLUSIONS AND IMPLICATIONS: The DM is the most informative site for ultrasound-based SD diagnosis. A model combining clinical predictors with DM radiomics offers a practical and efficient assessment tool. This approach supports targeted ultrasound use in long-term care to facilitate early detection and intervention for SD.