Applying Multi-Modal Magnetic Resonance Imaging Radiomics in Differentiating Benign and Malignant Ovarian Tumors

Fengzhi Cui,Jianhua Liu,Nan Zhang,Junfeng Lv

DOI: https://doi.org/10.1117/12.3006200

2023-01-01

Abstract:OBJECTIVE: To explore the clinical application valuation of prediction model on the basis of multi-modal magnetic resonance imaging (MRI) radiomics for discriminating ovarian malignant and benign tumors. METHODS: Totally 93 patients with 110 ovarian tumors confirmed by operation and histopathology were collected, including the training cohort (n=76) and testing cohort (n=34). Radiomics features are extracted from the DWI, axial T2WI lipid pressure, T1 LAVA enhancement, sagittal T2WI of MRI images by A.K. software. Based on the above features, a Logistic model was constructed to classify benign and malignant ovarian tumors. Additionally, diagnostic performance of the prediction model is assessed and compared. RESULTS: The multi-modal MRI radiomics prediction model exhibits excellent diagnostic performance. In the training cohorts, the accuracy, AUC, sensitivity and specificity of the multimodal MRI radiomic prediction model in differentiating benign and malignant ovarian tumors were 0.816, 0.887, 0.744 and 0.909 respectively, while in the testing cohorts were 0.794, 0.839, 0.684 and 0.933. CONCLUSION: The Logistic model MRIbased DWI, axial T2WI lipid pressure, T1 LAVA enhancement, sagittal T2WI achieved promising diagnostic performance, expected to be regarded as a preoperative imaging tool in classifying benign and malignant ovarian tumors.