Jing Yang ¹ Yang Liu ² Xiao long Liu ³ Yao xin Wang ⁴ Xian hong Wang ⁵ Conghui Ai ⁶ Qiu Bi ^1*

• ¹ Department of MRI, The First People’s Hospital of Yunnan Province, the Affiliated Hospital of Kunming University of Science and Technology, Kunming, Yunnan Province, China
• ² Department of Radiology, the First Affiliated Hospital of Chongqing Medical University, Chongqing, China
• ³ Department of Medical Imaging, the People’s Hospital of Puer, the Affiliated Hospital of Kunming University of Science and Technology, Puer, China
• ⁴ Department of Radiology, Shenzhen Traditional Chinese Medicine Hospital, The Fourth Clinical Medical College of Guangzhou University of Chinese Medicine, Shenzhen, China
• ⁵ The Affiliated Hospital of Kunming University of Science and Technology, Kunming, China
• ⁶ Department of Radiology, Yunnan Cancer Hospital, The Third Affiliated Hospital of Kunming Medical University，Peking University Cancer Hospital Yunnan, Kunming, China

Purpose: To evaluate the effectiveness of MRI-based intratumoral and peritumoral radiomics models for predicting deep myometrial invasion (DMI) of early-stage endometrioid adenocarcinoma (EAC). Methods: The data of 459 EAC patients from three centers were retrospectively collected. Radiomics features were extracted separately from the intratumoral and peritumoral regions expanded by 0mm, 5mm, and 10mm on unimodal and multimodal MRI. Then various radiomics models were developed and validated and the optimal model was confirmed. Integrated models were constructed by ensemble and stacking algorithms respectively based on the above radiomic models. The models' performance was evaluated using the area under the curve (AUC). Results: The multimodal MRI-based radiomics model, which included both intratumoral and peritumoral regions expanded by 5mm was the optimal radiomics model, with an AUC of 0.74 in the validation group. When the same integrated algorithm was utilized, the integrated models with 5mm expansion presented higher AUCs than those with 0mm and 10mm expansion in the validation group. The performance of stacking model and ensemble model with 5mm expansion was similar, and the AUCs were 0.74 and 0.75, respectively. Conclusion: The multimodal radiomic model from the intratumoral and peritumoral regions expanded by 5mm has the potential to improve the performance for detecting DMI of early-stage EAC. The integrated models are of little value in increasing the prediction.