EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Multiclass datasets expand neural network utility: an example on ankle radiographs

Suam Kim, Philipp Rebmann, Phuong Hien Tran, Elias Kellner, Marco Reisert, David Steybe, Jörg Bayer, Fabian Bamberg, Elmar Kotter, Maximilian Russe

International Journal of Computer Assisted Radiology and Surgery · 2023

DOI: 10.1007/s11548-023-02839-9 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Computer Assisted Radiology and Surgery, 2023. Verfasst von Suam Kim und 9 weiteren Autor:innen. Beteiligte Einrichtung: Department of Diagnostic and Interventional Radiology, Faculty of Medicine, Medical Center-University of Freiburg, Hugstetter Str. 55, 79106, Freibur…. Hauptthemen laut MEDLINE: Artificial Intelligence, Fractures, Bone. Die Arbeit wurde bisher 4-mal zitiert (Crossref) und nennt 19 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Universitätsklinikum Freiburg.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Purpose

Artificial intelligence in computer vision has been increasingly adapted in clinical application since the implementation of neural networks, potentially providing incremental information beyond the mere detection of pathology. As its algorithmic approach propagates input variation, neural networks could be used to identify and evaluate relevant image features. In this study, we introduce a basic dataset structure and demonstrate a pertaining use case.

Methods

A multidimensional classification of ankle x-rays ( n = 1493) rating a variety of features including fracture certainty was used to confirm its usability for separating input variations. We trained a customized neural network on the task of fracture detection using a state-of-the-art preprocessing and training protocol. By grouping the radiographs into subsets according to their image features, the influence of selected features on model performance was evaluated via selective training.

Results

The models trained on our dataset outperformed most comparable models of current literature with an ROC AUC of 0.943. Excluding ankle x-rays with signs of surgery improved fracture classification performance (AUC 0.955), while limiting the training set to only healthy ankles with and without fracture had no consistent effect.

Conclusion

Using multiclass datasets and comparing model performance, we were able to demonstrate signs of surgery as a confounding factor, which, following elimination, improved our model. Also eliminating pathologies other than fracture in contrast had no effect on model performance, suggesting a beneficial influence of feature variability for robust model training. Thus, multiclass datasets allow for evaluation of distinct image features, deepening our understanding of pathology imaging.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Artificial IntelligenceFractures, BoneHumansAnkleNeural Networks, ComputerRadiographyDiagnostic Imaging

Stichwörter der Autor:innen

Deep learningMusculoskeletalNeural networkRadiographs

Autor:innen und Institutionen

  • Department of Diagnostic and Interventional Radiology, Faculty of Medicine, Medical Center-University of Freiburg, Hugstetter Str. 55, 79106, Freiburg, Germany. suam.kim@uniklinik-freiburg.de.
  • Department of Diagnostic and Interventional Radiology, Faculty of Medicine, Medical Center-University of Freiburg, Hugstetter Str. 55, 79106, Freiburg, Germany.
  • Department of Medical Physics, Faculty of Medicine, Medical Center-University of Freiburg, University of Freiburg, Freiburg, Germany.
  • Department of Oral and Maxillofacial Surgery, Faculty of Medicine, Medical Center-University of Freiburg, Freiburg, Germany.
  • Department of Trauma and Orthopaedic Surgery, Schwarzwald-Baar Hospital, Villingen-Schwenningen, Germany.

Förderung

  • Universitätsklinikum Freiburg

Zitieren

APA

Kim, S., Rebmann, P., Tran, P. H., Kellner, E., Reisert, M., Steybe, D., Bayer, J., Bamberg, F., Kotter, E., & Russe, M. (2023). Multiclass datasets expand neural network utility: an example on ankle radiographs. International Journal of Computer Assisted Radiology and Surgery. https://doi.org/10.1007/s11548-023-02839-9

Vancouver

Kim S, Rebmann P, Tran PH, Kellner E, Reisert M, Steybe D, et al. Multiclass datasets expand neural network utility: an example on ankle radiographs. International Journal of Computer Assisted Radiology and Surgery. 2023. doi:10.1007/s11548-023-02839-9

BibTeX
@article{Kim2023,
  author = {Kim, Suam and Rebmann, Philipp and Tran, Phuong Hien and Kellner, Elias and Reisert, Marco and Steybe, David and Bayer, Jörg and Bamberg, Fabian and Kotter, Elmar and Russe, Maximilian},
  title = {Multiclass datasets expand neural network utility: an example on ankle radiographs},
  journal = {International Journal of Computer Assisted Radiology and Surgery},
  year = {2023},
  doi = {10.1007/s11548-023-02839-9}
}
Kostenlos starten →