The use of radiomics and machine learning for the differentiation of chondrosarcoma from enchondroma

Abstract

Purpose To construct and compare machine learning models for differentiating chondrosarcoma from enchondroma using radiomic features from T1 and fat suppressed Proton density (PD) magnetic resonance imaging (MRI). Methods Eighty‐eight patients (57 with enchondroma, 31 with chondrosarcoma) were retrospectively included. Histogram matching and N4ITK MRI bias correction filters were applied. An experienced musculoskeletal radiologist and a senior resident in radiology performed manual segmentation. Voxel sizes were resampled. Laplacian of Gaussian filter and wavelet‐based features were used. One thousand eight hundred eighty‐eight features were obtained for each patient, with 944 from T1 and 944 from PD images. Sixty‐four unstable features were removed. Seven machine learning models were used for classification. Results Classification with all features showed neural network was the best model for both readers' datasets with area under the curve (AUC), classification accuracy (CA), and F1 score of 0.979, 0.984; 0.920, 0.932; and 0.889, 0.903, respectively. Four features, including one common to both readers, were selected using fast correlation based filter. The best performing models with selected features were gradient boosting for Fatih Erdem's dataset and neural network for Gülen Demirpolat's dataset with AUC, CA, and F1 score of 0.990, 0.979; 0.943, 0.955; 0.921, 0.933, respectively. Neural Network was the second‐best model for FE's dataset based on AUC (0.984). Conclusion Using pathology as a gold standard, this study defined and compared seven well‐performing models to distinguish enchondromas from chondrosarcomas and provided radiomic feature stability and reproducibility among the readers.

Enchondroma Detection from Hand Radiographs with an Interactive Deep Learning Segmentation Tool—A Feasibility Study

“…It is imperative to ascertain these to fully evaluate the model's real-world utility. Further DL model development plans from radiographs could be the differential diagnosis of enchondroma and chondrosarcoma from plain radiographs, as previously performed from MRI and CT with ML, texture analysis, and radiomics [17][18][19]. Due to the challenges and constraints associated with the histological diagnosis of peripheral chondrosarcoma and enchondroma, the DL model could offer an additional valuable assessment to assist in determining the most suitable treatment for these patients.…”

“…Additionally, another study focused on utilizing MRI radiomics for distinguishing chondrosarcoma from enchondroma. In this research, pathology served as the gold standard for comparison, and different models were evaluated, all of which demonstrated strong performance in this diagnostic task [19]. These imaging modalities are usually used for treatment planning after suspicion has been raised with a traditional radiograph and is thus not available for diagnostic aid.…”

SPECT/CT Radiomics for Differentiating between Enchondroma and Grade I Chondrosarcoma

“…Several other studies have investigated the value of MR image-based radiomics in chondroid tumors [9,10,21]. These studies have shown promising results in the discrimination of enchondroma and chondrosarcoma.…”

“…Studies have suggested that certain texture parameters that reflect intratumoral heterogeneity can serve as valuable prognostic markers [8]. Texture analysis with MR images have shown the potential for differentiating chondrosarcoma and enchondromas [9,10]. While there have been several attempts to use radiomics from anatomic imaging, the value of SPECT/CT radiomics have not yet been evaluated in bone tumors.…”