Objective: Our primary aim is to develop and validate ultrasound-based radiomics models to discriminate between benign and malignant adnexal masses. The secondary aim is to develop and validate models to discriminate between early-stage ovarian cancers at high- and low-risk of lymph node dissemination. Methods: This is a retrospective study including patients prospectively collected at Fondazione Policlinico Universitario A. Gemelli, IRCCS and included in the ongoing International Ovarian Tumor Analysis (IOTA) studies 5-7 between 2014 and 2024. Patients were examined by transvaginal ultrasound and selected for surgery. The patient cohort was split into training and validation sets. Variables used in model building were age, CA125 and/or those radiomics features that were statistically significantly different between groups (Wilcoxon-Mann-Whitney Test with Benjamini-Hochberg correction for multiple comparisons) and assessed as not redundant based on the Pearson correlation coefficient. We described discriminative ability as area under the receiver operating characteristics curve (AUC). The performance of models in predicting malignancy was compared to that of the IOTA-ADNEX. Results: 2073 images were analyzed. For the primary objective, 14 non redundant radiomics features differed statistically significantly between benign and malignant tumors in the training set. In the validation set, the radiomics model including age, CA125 and radiomics features has a discriminative ability equivalent to that of the IOTA-ADNEX model in distinguishing between benign and malignant adnexal masses (AUC 0.89 vs 0.91, p=0.15). For the secondary objective, 4 non redundant radiomics features differed statistically significantly between tumors at high- and low-risk of lymph node metastases in the training set. In the validation set, the best performing model was the model including age, CA125 and radiomics features with AUC 0.79. Conclusions: Ultrasound-based radiomics models can be used in clinical practice to better personalize surgical treatment in patients with an adnexal mass.

Preoperative value of radiomics applied to ultrasound imaging to identify adnexal masses

CIANCIA, MARIANNA
2022/2023

Abstract

Objective: Our primary aim is to develop and validate ultrasound-based radiomics models to discriminate between benign and malignant adnexal masses. The secondary aim is to develop and validate models to discriminate between early-stage ovarian cancers at high- and low-risk of lymph node dissemination. Methods: This is a retrospective study including patients prospectively collected at Fondazione Policlinico Universitario A. Gemelli, IRCCS and included in the ongoing International Ovarian Tumor Analysis (IOTA) studies 5-7 between 2014 and 2024. Patients were examined by transvaginal ultrasound and selected for surgery. The patient cohort was split into training and validation sets. Variables used in model building were age, CA125 and/or those radiomics features that were statistically significantly different between groups (Wilcoxon-Mann-Whitney Test with Benjamini-Hochberg correction for multiple comparisons) and assessed as not redundant based on the Pearson correlation coefficient. We described discriminative ability as area under the receiver operating characteristics curve (AUC). The performance of models in predicting malignancy was compared to that of the IOTA-ADNEX. Results: 2073 images were analyzed. For the primary objective, 14 non redundant radiomics features differed statistically significantly between benign and malignant tumors in the training set. In the validation set, the radiomics model including age, CA125 and radiomics features has a discriminative ability equivalent to that of the IOTA-ADNEX model in distinguishing between benign and malignant adnexal masses (AUC 0.89 vs 0.91, p=0.15). For the secondary objective, 4 non redundant radiomics features differed statistically significantly between tumors at high- and low-risk of lymph node metastases in the training set. In the validation set, the best performing model was the model including age, CA125 and radiomics features with AUC 0.79. Conclusions: Ultrasound-based radiomics models can be used in clinical practice to better personalize surgical treatment in patients with an adnexal mass.
2022
Preoperative value of radiomics applied to ultrasound imaging to identify adnexal masses
Machine Learning
Ultrasonography
Ovarian masses
File in questo prodotto:
File Dimensione Formato  
TESI - Preoperative value of radiomics applied to ultrasound imaging to identify adnexal masses_def_pdfA.pdf

accesso riservato

Dimensione 1.02 MB
Formato Adobe PDF
1.02 MB Adobe PDF

The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.12608/76229