EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

MRI-BASED RADIOMICS FOR ASSESSMENT OF THE INFRAPATELLAR FAT PAD'S INFLUENCE ON PATELLOFEMORAL PAIN

I.R. Campos, R.A. van der Heijden, E.H.G. Oei, M. van Middelkoop, S. Klein, J.S. Cardoso, J. Hirvasniemi

Osteoarthritis Imaging · 2023 · 3 · 100134

DOI: 10.1016/j.ostima.2023.100134 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Osteoarthritis Imaging, 2023. Verfasst von I.R. Campos und 6 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref). Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The infrapatellar fat pad (IPFP) has been proposed as a source of knee pain in patients suffering from either knee OA or its supposed precursor, patellofemoral pain (PFP). While radiomics has been successfully applied to different MRI data, there is a lack of studies using radiomic features from the IPFP for the assessment of PFP. The aims of this study were to automatically extract radiomic features from the IPFP and assess the ability of these radiomic features to distinguish between patients with PFP and healthy controls. Sagittal 3D SPGR sequences with and without fat saturation (FS) from the TripleP study (n=64 PFP patients, n=70 controls) were used in the analyses. The IPFP of the 134 knees was automatically segmented using a nnU-Net with a 3D U-Net configuration. The nnU-Net was trained using 30 manually segmented knees and 5-fold cross-validation. Medial and lateral borders of the patella bone were used as anatomical landmarks for the manual IPFP segmentation. Radiomic features related to orientation (n=3), histogram (n=13), shape (n=21), original phase (n=39) and texture (n=376) were extracted (n=452), using the open-source Workflow for Optimal Radiomics Classification toolbox (v. 3.6.0). Elastic Net from the Scikit-learn Python module was used for dimensionality reduction and classification. The hyperparameters were optimized using 5-fold stratified cross-validation with a grid search and 10 repetitions. Three different models were evaluated: 1) basic model consisting of the patients’ information (age, sex and BMI), 2) radiomic features model (n=452), and 3) combination of models 1 and 2. The classification performance of the aforementioned models to distinguish between patients with PFP and controls was assessed using the area under the receiver operating characteristic (ROC AUC) and precision-recall curves (PR AUC). The mean Dice similarity coefficient for the IPFP automatic segmentation was 0.94 (0.02). Visual evaluation showed that up to 15 of the most medial and lateral slices were not correctly segmented. Thus, 15 slices per side were removed from each segmentation mask and the features were extracted from the most central slices of the IPFP. Regarding the results for distinguishing between patients with and without PFP, the basic model resulted in an ROC AUC of 0.67 (SD: 0.08) and an PR AUC of 0.62 (0.14). The radiomic features extracted from the non-FS sequence resulted in an ROC AUC of 0.57 (0.05) and an PR AUC of 0.73 (0.05) for model 2 and in an ROC AUC of 0.57 (0.05) and an PR AUC of 0.71 (0.07) for model 3. The performances of models 2 and 3 were slightly worse when using radiomic features from the FS sequence. However, when using only texture features from the FS sequence in the classification, an ROC AUC of 0.67 (0.05) and an PR AUC of 0.70 (0.10) was obtained. The model with all radiomic features combined was not able to distinguish between patients with and without PFP. However, better performance was obtained when using a subgroup of texture features. The results of this study indicate that texture features extracted automatically from the central part of the IPFP from FS SPGR sequence are promising quantitative imaging biomarkers for the assessment of PFP. None None CORRESPONDENCE ADDRESS: [email protected]

Zitieren

APA

Campos, I., Heijden, R. V. D., Oei, E., Middelkoop, M. V., Klein, S., Cardoso, J., & Hirvasniemi, J. (2023). MRI-BASED RADIOMICS FOR ASSESSMENT OF THE INFRAPATELLAR FAT PAD'S INFLUENCE ON PATELLOFEMORAL PAIN. Osteoarthritis Imaging, 3, 100134. https://doi.org/10.1016/j.ostima.2023.100134

Vancouver

Campos I, Heijden RVD, Oei E, Middelkoop MV, Klein S, Cardoso J, et al. MRI-BASED RADIOMICS FOR ASSESSMENT OF THE INFRAPATELLAR FAT PAD'S INFLUENCE ON PATELLOFEMORAL PAIN. Osteoarthritis Imaging. 2023;3:100134. doi:10.1016/j.ostima.2023.100134

BibTeX
@article{Campos2023,
  author = {Campos, I.R. and Heijden, R.A. van der and Oei, E.H.G. and Middelkoop, M. van and Klein, S. and Cardoso, J.S. and Hirvasniemi, J.},
  title = {MRI-BASED RADIOMICS FOR ASSESSMENT OF THE INFRAPATELLAR FAT PAD'S INFLUENCE ON PATELLOFEMORAL PAIN},
  journal = {Osteoarthritis Imaging},
  year = {2023},
  volume = {3},
  pages = {100134},
  doi = {10.1016/j.ostima.2023.100134}
}
Kostenlos starten →