Fluid and White Matter Suppression Contrasts MRI Improves Deep Learning Detection of Multiple Sclerosis Cortical Lesions

To investigate, for the first time, the efficacy of Fluid and White Matter Suppression (FLAWS) MRI sequence in improving Deep Learning (DL)-based detection and segmentation of cortical lesions in Multiple Sclerosis (MS) patients even, with applicability to clinical settings where only standard T1-weighted images are available. In this retrospective multi-site study, we analyzed 204 MS patients using DL models developed with FLAWS and Magnetization Prepared 2 Rapid Acquisition Gradient Echoes (MP2RAGE) sequences. Reference standard annotations were established through two approaches: (1) consensus of three expert raters across all contrasts, and (2) single-rater annotations for individual modalities. Models were validated on both internal and external datasets, with performance assessed using F1-score for detection and DSC for segmentation accuracy. Models involving FLAWS demonstrated superior performance over MP2RAGE-only models. The combined MP2RAGE+FLAWS model achieved CL detection with median F1-score of 0.667[0.339−0.840] compared to multirater consensus. Models trained on comprehensive consensus annotations outperformed those trained on single-modality annotations. Notably, a model exclusively based on MP2RAGE images and trained with FLAWS-derived annotations showed, showed strong generalization to external Magnetization Prepared Rapid Gradient-Echo (MPRAGE) clinical datasets (median F1-score: 0.55[0.211 − 0.998]). Integration of FLAWS-derived contrasts and annotations significantly improves DL-based CL detection and segmentation. The models demonstrate capability in identifying lesions missed by individual raters and maintain robust performance even without FLAWS sequences in standard clinical settings. This advancement facilitates clinical translation, supported by publicly available inference models on DockerHub. MP2RAGE+FLAWS MRI sequences achieves superior cortical lesion detection performance Quantitative validation shows median F1 of 0.667[0.339 − 0.840] with combined sequences FLAWS-trained model generalizes well to standard MPRAGE images (F1: 0.55[0.211 − 0.998) Model demonstrates immediate clinical applicability on standard MPRAGE without FLAWS input Implementation publicly accessible via DockerHub for widespread clinical adoption