POPurevdorj Otgonbayar

Purevdorj Otgonbayar

*This is an algorithmically generated author record based on publicly available information.

Purevdorj Otgonbayar has a total of 6 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    CT-Based Radiomics Combined with Deep Learning for Preoperative Differentiation of Clear Cell and Non–Clear Cell Renal Cell Carcinoma: Internal and External Validation Study

    by Enkhbold Sereejav, Purevdorj Otgonbayar, Telmen Khuvituguldur, Sodgerel Bold, Shiirenyamba Avirmed, Gonchigsuren Dagvasumberel, Oyunbold Lamid-Ochir

    Objective: Accurate preoperative differentiation of renal cell carcinoma (RCC) subtypes remains challenging with conventional computed tomography (CT) owing to overlapping features. This study evaluated CT-based radiomics combined with deep learning for differentiation of clear cell RCC (ccRCC) from non–clear cell RCC (non-ccRCC), using internal and external validation. Methods: This retrospective, single-center study included an internal cohort of 121 patients with histopathologically confirmed RCC (84 ccRCC, 37 non-ccRCC) and an independent external cohort of 73 patients. All underwent preoperative multiphasic contrast-enhanced CT (128-slice Ingenuity). Corticomedullary-phase tumors were segmented in 3D Slicer by two radiologists; inter-observer reliability used the intraclass correlation coefficient (ICC). In total, 122 radiomic features were extracted (first-order, GLCM, GLRLM, GLSZM, GLDM, shape). Least absolute shrinkage and selection operator (LASSO) regression selected features for a deep learning–based classification model. Diagnostic efficacy was assessed by receiver operating characteristic (ROC) curve analysis. Results: Inter-observer agreement was excellent (ICC = 0.91; 95% CI, 0.87–0.94). In the internal cohort, the model achieved an area under the curve (AUC) of 0.92, sensitivity 83.3%, specificity 92.3%, and accuracy 86.0%. GLRLM features were most discriminative (AUC = 0.91, p < .001), followed by first-order uniformity (AUC = 0.88) and variance (AUC = 0.86). In the external cohort, performance remained robust (AUC = 0.86, sensitivity 82.0%, specificity 80.0%, accuracy 80.8%), confirming generalizability. Conclusion: CT-based radiomics combined with deep learning demonstrates high diagnostic performance for differentiating ccRCC from non-ccRCC, generalizing across internal and external datasets. This non-invasive approach may serve as a valuable imaging biomarker supporting clinical decision-making and personalized treatment planning.