Interinstitutional Portability of a Deep Learning Brain MRI Lesion Segmentation Algorithm.

Abstract

To assess how well a brain MRI lesion segmentation algorithm trained at one institution performed at another institution, and to assess the effect of multi-institutional training datasets for mitigating performance loss. In this retrospective study, a three-dimensional U-Net for brain MRI abnormality segmentation was trained on data from 293 patients from one institution (IN1) (median age, 54 years; 165 women; patients treated between 2008 and 2018) and tested on data from 51 patients from a second institution (IN2) (median age, 46 years; 27 women; patients treated between 2003 and 2019). The model was then trained on additional data from various sources: <i>(a)</i> 285 multi-institution brain tumor segmentations, <i>(b)</i> 198 IN2 brain tumor segmentations, and <i>(c)</i> 34 IN2 lesion segmentations from various brain pathologic conditions. All trained models were tested on IN1 and external IN2 test datasets, assessing segmentation performance using Dice coefficients. The U-Net accurately segmented brain MRI lesions across various pathologic conditions. Performance was lower when tested at an external institution (median Dice score, 0.70 [IN2] vs 0.76 [IN1]). Addition of 483 training cases of a single pathologic condition, including from IN2, did not raise performance (median Dice score, 0.72; <i>P</i> = .10). Addition of IN2 training data with heterogeneous pathologic features, representing only 10% (34 of 329) of total training data, increased performance to baseline (Dice score, 0.77; <i>P</i> < .001). This final model produced total lesion volumes with a high correlation to the reference standard (Spearman <i>r</i> = 0.98). For brain MRI lesion segmentation, adding a modest amount of relevant training data from an external institution to a previously trained model supported successful application of the model to this external institution.<b>Keywords:</b> Neural Networks, Brain/Brain Stem, Segmentation <i>Supplemental material is available for this article.</i> © RSNA, 2021.