1.Introduction Accurate delineation of chronic stroke lesions is crucial for many research and clinical applications. Indeed, all neuroimaging research after stroke require a segmentation process. Moreover, the segmentation is implemented in algorithms discussing prognostic elements. However, manual segmentation from T1-w MR images is time-consuming and prone to errors. 2.Objectives This work aimed to automate chronic stroke lesion detection and segmentation using the nnU-Net framework and a combination of datasets. The nnU-Net framework is widely recognized for its state-of-the-art performance in medical image segmentation. We have previously successfully adapted the nnU-Net framework to multiple sclerosis lesion segmentation within the Longiseg4ms tool. 3.Materials and Methods We utilized the ATLAS v2.0 dataset* and an in-house dataset of T1-w and FLAIR MRI scans from patients with chronic stroke lesions. We leveraged the nnU-Net framework and incorporated both T1-w and FLAIR MRI modalities 4.Results The target model achieved a mean Dice score of 0.730 and a mean lesion-wise F1 score of 0.684, demonstrating superior performance compared to the baseline model. The high correlation coefficient (Pearson correlation coefficient = 0.930) between the ground truth and model output volumes indicates a strong agreement between the two segmentation methods. This correlation highlights the accuracy of our volume prediction and highlights the reliability of our automated segmentation approach. 5.Conclusion Our nnU-Net-based model, trained on T1-w and FLAIR images, improved the segmentation of chronic stroke lesions, outperforming the baseline model. These findings highlight the potential of automated models for improving chronic stroke lesion segmentation and using FLAIR modality to enhance the results. *Liew S-L et al. (2022). A large, curated, open-source stroke neuroimaging dataset to improve lesion segmentation algorithms. Sci Data 9, 320.