Automated artifact detection in abbreviated dynamic contrast-enhanced (DCE) MRI-derived maximum intensity projections (MIPs) of the breast

Abstract

Objectives To automatically detect MRI artifacts on dynamic contrast-enhanced (DCE) maximum intensity projections (MIPs) of the breast using deep learning. Methods Women who underwent clinically indicated breast MRI between October 2015 and December 2019 were included in this IRB-approved retrospective study. We employed two convolutional neural network architectures (ResNet and DenseNet) to detect the presence of artifacts on DCE MIPs of the left and right breasts. Networks were trained on images acquired up to and including the year 2018 using a 5-fold cross-validation (CV). Ensemble classifiers were built with the resulting CV models and applied to an independent holdout test dataset, which was formed by images acquired in 2019. Results Our study sample contained 2265 examinations from 1794 patients (median age at first acquisition: 50 years [IQR: 17 years]), corresponding to 1827 examinations of 1378 individuals in the training dataset and 438 examinations of 416 individuals in the holdout test dataset with a prevalence of image-level artifacts of 53% (1951/3654 images) and 43% (381/876 images), respectively. On the holdout test dataset, the ResNet and DenseNet ensembles demonstrated an area under the ROC curve of 0.92 and 0.94, respectively. Conclusion Neural networks are able to reliably detect artifacts that may impede the diagnostic assessment of MIPs derived from DCE subtraction series in breast MRI. Future studies need to further explore the potential of such neural networks to complement quality assurance and improve the application of DCE MIPs in a clinical setting, such as abbreviated protocols. Key Points • Deep learning classifiers are able to reliably detect MRI artifacts in dynamic contrast-enhanced protocol-derived maximum intensity projections of the breast. • Automated quality assurance of maximum intensity projections of the breast may be of special relevance for abbreviated breast MRI, e.g., in high-throughput settings, such as cancer screening programs.

Impact of Non-Contrast Enhanced Imaging Input Sequences on the Generation of Virtual Contrast-Enhanced Breast MRI Scans using Neural Networks

“…The independent test set was evaluated in two experiments (Figure 1b). Experiment 1: Quantitative analyses were performed for the full test dataset for a) the entire-breast volume (n=187) and b) using a subset of the examinations in which target findings, either enhancing or non-enhancing, could be identified and subsequently manually segmented (145/187 cases), as described in the Quantitative Analysis subsection. Experiment 2: Qualitative analyses were performed via visual readings of three independent readers using the full independent test set (n=187) and corresponding scores described in the Qualitative Analysis subsection. All cohort examinations (n=973) were previously included in studies focused on detecting artifacts in dynamic contrast-enhanced and DWI derived maximum-intensity projections (27,28).…”

“…All cohort examinations (n=973) were previously included in studies focused on detecting artifacts in dynamic contrast-enhanced and DWI derived maximum-intensity projections (27,28).…”

Prevalence and influencing factors for artifact development in breast MRI-derived maximum intensity projections

“…This retrospective analysis was performed in a previously reported partially overlapping study sample (2265/2524 examinations) (25). The ethics committee of the Friedrich-Alexander-University (FAU) Erlangen-Nürnberg approved this study and waived the need for written informed consent.…”

“…The ethics committee of the Friedrich-Alexander-University (FAU) Erlangen-Nürnberg approved this study and waived the need for written informed consent. In the previous study, we evaluated an automated detection of MRI artifacts on DCE MIPs of the breast by applying deep learning methods (25). By contrast, here, we describe the prevalence of MRI artifacts on DCE MIPs of the breast in more detail and investigate possible associations with patient characteristics and technical (scanner) attributes.…”

“…As previously reported (25), examinations were performed using the routine clinical MRI systems at UHE (1.5 and 3.0 T MRI; Model names: Aera, Avanto, Skyra, Sola and Vida; Siemens Healthineers, Erlangen, Germany) using dedicated breast MRI coils. Likewise, as in our previous study, the present analysis was also performed with the MIPs of the subtraction series between the second postcontrast images (using the in-phase images in case of a DIXON sequence) and the native T1-weighted images generated by the scanner system.…”