MRI‐Based Machine Learning for Differentiating Borderline From Malignant Epithelial Ovarian Tumors: A Multicenter Study

Abstract

BackgroundPreoperative differentiation of borderline from malignant epithelial ovarian tumors (BEOT from MEOT) can impact surgical management. MRI has improved this assessment but subjective interpretation by radiologists may lead to inconsistent results.PurposeTo develop and validate an objective MRI‐based machine‐learning (ML) assessment model for differentiating BEOT from MEOT, and compare the performance against radiologists' interpretation.Study TypeRetrospective study of eight clinical centers.PopulationIn all, 501 women with histopathologically‐confirmed BEOT (n = 165) or MEOT (n = 336) from 2010 to 2018 were enrolled. Three cohorts were constructed: a training cohort (n = 250), an internal validation cohort (n = 92), and an external validation cohort (n = 159).Field Strength/SequencePreoperative MRI within 2 weeks of surgery. Single‐ and multiparameter (MP) machine‐learning assessment models were built utilizing the following four MRI sequences: T2‐weighted imaging (T2WI), fat saturation (FS), diffusion‐weighted imaging (DWI), apparent diffusion coefficient (ADC), and contrast‐enhanced (CE)‐T1WI.AssessmentDiagnostic performance of the models was assessed for both whole tumor (WT) and solid tumor (ST) components. Assessment of the performance of the model in discriminating BEOT vs. early‐stage MEOT was made. Six radiologists of varying experience also interpreted the MR images.Statistical TestsMann–Whitney U‐test: significance of the clinical characteristics; chi‐square test: difference of label; DeLong test: difference of receiver operating characteristic (ROC).ResultsThe MP‐ST model performed better than the MP‐WT model for both the internal validation cohort (area under the curve [AUC] = 0.932 vs. 0.917) and external validation cohort (AUC = 0.902 vs. 0.767). The model showed capability in discriminating BEOT vs. early‐stage MEOT, with AUCs of 0.909 and 0.920, respectively. Radiologist performance was considerably poorer than both the internal (mean AUC = 0.792; range, 0.679–0.924) and external (mean AUC = 0.797; range, 0.744–0.867) validation cohorts.Data ConclusionPerformance of the MRI‐based ML model was robust and superior to subjective assessment of radiologists. If our approach can be implemented in clinical practice, improved preoperative prediction could potentially lead to preserved ovarian function and fertility for some women.Level of EvidenceLevel 4.Technical EfficacyStage 2. J. Magn. Reson. Imaging 2020;52:897–904.

Multiparametric‐MRI‐Based Radiomics Model for Differentiating Primary Central Nervous System Lymphoma From Glioblastoma: Development and Cross‐Vendor Validation

“…To enable the VOI to be used with images from all MRI sequences, the CE‐T 1 WI, DWI, and ADC images were resampled and aligned to the same resolution, spacing, and position as the FLAIR images using the open‐source Insight Segmentation and Registration Toolkit (ITK, v. 4.7.2; https://itk.org/). 20 To standardize the MR images from all sequences, the mean value and the standard deviation of intensity in the images from each MRI volume were calculated, and each was normalized by the z‐score method, which consisted of subtracting the mean intensity and dividing by the standard deviation of intensity 21–23 …”

“…20 To standardize the MR images from all sequences, the mean value and the standard deviation of intensity in the images from each MRI volume were calculated, and each was normalized by the z-score method, which consisted of subtracting the mean intensity and dividing by the standard deviation of intensity. [21][22][23] MRI feature extraction was conducted using an open-source Python package Pyradiomics (v. 2.1.2; http://www.radiomics.io/ pyradiomics.html). 24 A total of 851 image features were calculated for each MRI volume, including 14 shape-based features, 18 firstorder statistics features, 75 texture features, and 744 wavelet features.…”

Diagnosing Ovarian Cancer on MRI: A Preliminary Study Comparing Deep Learning and Radiologist Assessments

“…Aramendía et al developed a CAD technique for US images that was able to discriminate between malignant and benign adnexal masses based on a texture analysis of 145 patients [ 18 ]. Jian et al and Li et al constructed an MRI-based texture analysis model to distinguish between type I and type II epithelial ovarian cancers and borderline and malignant epithelial ovarian tumors based on T2WI+DWI+ADC and CE-T1WI+T2WI [ 19 , 20 ]. Wang et al developed a CNN that distinguishes benign from malignant ovarian on T2WI, CE-T1WI, and clinical variables [ 21 ].…”

MR-based radiomics-clinical nomogram in epithelial ovarian tumor prognosis prediction: tumor body texture analysis across various acquisition protocols

“…Its essence is to extract high-throughput quantitative features from high-quality medical images and establish a predictive model for diagnosis and prognostic evaluation [ 19 – 23 ]. Previous studies have reported that radiomics has potential in the classification of ovarian cystadenomas and stratification of ovarian cysts [ 24 , 25 ]. A CT-based radiomics study has demonstrated the feasibility of predicting the risk of postoperative recurrence of advanced high-grade serous ovarian cancer [ 26 ].…”