EUVIMEDEUVIMED entdecken
VergleichsstudieOpen Access · CC BY 4.0

Exploring foundation models for multi-class muscle segmentation in MR images of neuromuscular disorders: A comparative analysis of accuracy and uncertainty

Nicola Casali, Alessandro Brusaferri, Giuseppe Baselli, Marco Moscatelli, Domenico Aquino, Marina Grisoli, Giovanna Rizzo, Alfonso Mastropietro

Computer Methods and Programs in Biomedicine · 2025

DOI: 10.1016/j.cmpb.2025.109035 ↗

Kurz gefasst

Vergleichsstudie in der Zeitschrift Computer Methods and Programs in Biomedicine, 2025. Verfasst von Nicola Casali und 7 weiteren Autor:innen. Beteiligte Einrichtung: Istituto di Sistemi e Tecnologie Industriali Intelligenti per il Manifatturiero Avanzato, Consiglio Nazionale delle Ricerche, Milan, 20133, Italy; Di…. Hauptthemen laut MEDLINE: Magnetic Resonance Imaging, Neuromuscular Diseases, Muscle, Skeletal, Image Processing, Computer-Assisted. Die Arbeit wurde bisher 7-mal zitiert (Crossref) und nennt 60 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Regione Lombardia, Fondazione Cariplo, Ministero dell'Università e della Ricerca.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

BACKGROUND AND OBJECTIVE: Deep learning (DL) models have shown promise for skeletal muscle (SM) segmentation in MR images, which is crucial for extracting biomarkers in neuromuscular disorders (NMDs). However, to ensure safe clinical use, models should provide uncertainty estimates, allowing radiologists to assess predictions and intervene when needed. Foundation Models (FMs) have the potential to play a significant role due to their strong generalization capabilities and well-calibrated predictions. However, their applicability in this context has not yet been explored. This study aims to develop an accurate and trustworthy technique by fine-tuning FMs to delineate fatty-infiltrated SM fascicles in NMD patients. METHODS: We fine-tuned Segment Anything Model (SAM) and MedSAM using two configurations - encoder/decoder and decoder only - and compared their performance against state-of-the-art 2D and 3D nnU-Net using a dataset of thigh MR images from 76 NMD patients, categorized into Early, Moderate, and Severe fatty infiltration groups. Accuracy was evaluated using the Dice Similarity Coefficient (DSC), while Uncertainty Quantification (UQ) was evaluated using the Expected Calibration Error (ECE) and the Negative Log-Likelihood (NLL). Deep Ensembles were used to convey epistemic uncertainty in addition to the aleatoric counterpart. RESULTS: SAM's fine-tuned encoder/decoder outperformed nnU-Net 3D in Moderate and Severe cases (DSC: 0.886 vs 0.883 and 0.857 vs 0.850) and was comparable in Early (DSC: 0.925). MedSAM did not show an advantage over SAM. Regarding UQ, SAM exhibited superior calibration in Moderate and Severe groups (ECE: 3.6% vs. 5.1% and 3.3% vs. 7.1%), CONCLUSIONS:: In conclusion, our findings demonstrate that fine-tuning SAM yields superior performance, considering both accuracy and UQ metrics, highlighting its enhanced reliability in challenging NMD imaging scenarios.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Magnetic Resonance ImagingNeuromuscular DiseasesMuscle, SkeletalImage Processing, Computer-AssistedHumansUncertaintyDeep LearningMaleFemaleAdultReproducibility of ResultsMiddle Aged

Stichwörter der Autor:innen

Deep learningFoundation modelsMRISegmentationSkeletal muscleUncertainty quantification

Autor:innen und Institutionen

  • Istituto di Sistemi e Tecnologie Industriali Intelligenti per il Manifatturiero Avanzato, Consiglio Nazionale delle Ricerche, Milan, 20133, Italy; Dipartimento di Elettronica, Informazione e Bioingegneria, Politecnico d…
  • Istituto di Sistemi e Tecnologie Industriali Intelligenti per il Manifatturiero Avanzato, Consiglio Nazionale delle Ricerche, Milan, 20133, Italy.
  • Dipartimento di Elettronica, Informazione e Bioingegneria, Politecnico di Milano, Milan, 20133, Italy.
  • Fondazione IRCCS Istituto Neurologico C. Besta, Milan, 20133, Italy.
  • Istituto di Sistemi e Tecnologie Industriali Intelligenti per il Manifatturiero Avanzato, Consiglio Nazionale delle Ricerche, Milan, 20133, Italy. Electronic address: alfonso.mastropietro@cnr.it.

Förderung

  • Regione Lombardia
  • Fondazione Cariplo
  • Ministero dell'Università e della Ricerca

Zitieren

APA

Casali, N., Brusaferri, A., Baselli, G., Moscatelli, M., Aquino, D., Grisoli, M., Rizzo, G., & Mastropietro, A. (2025). Exploring foundation models for multi-class muscle segmentation in MR images of neuromuscular disorders: A comparative analysis of accuracy and uncertainty. Computer Methods and Programs in Biomedicine. https://doi.org/10.1016/j.cmpb.2025.109035

Vancouver

Casali N, Brusaferri A, Baselli G, Moscatelli M, Aquino D, Grisoli M, et al. Exploring foundation models for multi-class muscle segmentation in MR images of neuromuscular disorders: A comparative analysis of accuracy and uncertainty. Computer Methods and Programs in Biomedicine. 2025. doi:10.1016/j.cmpb.2025.109035

BibTeX
@article{Casali2025,
  author = {Casali, Nicola and Brusaferri, Alessandro and Baselli, Giuseppe and Moscatelli, Marco and Aquino, Domenico and Grisoli, Marina and Rizzo, Giovanna and Mastropietro, Alfonso},
  title = {Exploring foundation models for multi-class muscle segmentation in MR images of neuromuscular disorders: A comparative analysis of accuracy and uncertainty},
  journal = {Computer Methods and Programs in Biomedicine},
  year = {2025},
  doi = {10.1016/j.cmpb.2025.109035}
}
Kostenlos starten →