Co-designed Automated Lesion Mapping and Reporting.
This app loads a structural T1 (or a manual MNI-aligned lesion mask) locally, runs the CALMaR lesion mapping chain in the browser, and never uploads patient data.
The pipeline: SynthStrip brain extraction → SynthStroke baseline lesion segmentation → SynthMorph deformable registration to MNI152 (1603 1mm) → affine-aware resample to the selected atlas grid → atlas overlap → group-FC weighted-sum → threshold + cluster cleanup. All stages are exposed as individual buttons under Advanced controls; Run analysis chains them in one click.
Background execution is possible if this site is added under "Always keep these sites active" in your browser settings.
Uploaded images and masks are read into browser memory for local viewing and analysis.
Static app files and public atlas assets may be served from external hosting, but patient-derived files are not sent to those services by the analysis code. The Neurodesk hosting layer records page views only, unless your browser sends Do Not Track or Global Privacy Control. It does not send custom events, patient images, masks, voxel values, screenshots, generated outputs, filenames, or analysis results to Google Analytics.
Please cite the atlas, segmentation, registration, connectome, visualization, and runtime methods used in your analysis.
Yeo BTT, Krienen FM, Sepulcre J, et al. The organization of the human cerebral cortex estimated by intrinsic functional connectivity. Journal of Neurophysiology. 2011;106(3):1125-65.
Schaefer A, Kong R, Gordon EM, et al. Local-global parcellation of the human cerebral cortex from intrinsic functional connectivity MRI. Cerebral Cortex. 2018;28(9):3095-3114.
CALMaR uses manifest asset lnm-synthstrip, runtime file models/synthstrip.onnx, source version SynthStrip-Hoopes2022-Apache-2.0. The ONNX asset was converted from FreeSurfer's official main SynthStrip model v1 checkpoint, synthstrip.1.pt; it is not the no-CSF or pediatric variant. The browser runs the ONNX FP32 export through the fast WASM wrapper with no final dilation.
Hoopes A, Mora JS, Dalca AV, Fischl B, Hoffmann M. SynthStrip: skull-stripping for any brain image. NeuroImage. 2022;260:119474.
https://surfer.nmr.mgh.harvard.edu/docs/synthstrip/ Original FreeSurfer checkpoint Hosted ONNX assetCALMaR uses manifest asset lnm-stroke-lesion, runtime file models/lnm-stroke-lesion.onnx, source version SynthStroke-baseline-MELBA2025-MIT. The ONNX asset was converted from liamchalcroft/synthstroke-baseline config.json and model.safetensors, not SynthStroke SynthPlus. The browser runs 1283 sliding-window patches with no TTA, then collapses background/stroke logits to stroke probability before thresholding.
Chalcroft L, Pereira S, Reinke A, et al. SynthStroke: synthetic chronic stroke lesion segmentation models. MELBA. 2025.
Upstream SynthStroke baseline model Hosted ONNX assetLiew SL, Lo BP, Donnelly MR, et al. A large, curated, open-source stroke neuroimaging dataset to improve lesion segmentation algorithms. Scientific Data. 2022;9:320.
Hoffmann M, Billot B, Greve DN, Iglesias JE, Fischl B, Dalca AV. SynthMorph: learning contrast-invariant registration without acquired images. IEEE Trans Med Imaging. 2022;41(3):543-58.
https://martinos.org/~malte/synthmorph/ADHD-200 Consortium. The ADHD-200 Consortium: a model to advance the translational potential of neuroimaging in clinical neuroscience. Frontiers in Systems Neuroscience. 2012;6:62.
http://fcon_1000.projects.nitrc.org/indi/adhd200/Boes AD, Prasad S, Liu H, et al. Network localization of neurological symptoms from focal brain lesions. Brain. 2015;138(10):3061-75.
NiiVue Contributors. NiiVue: a WebGL2 medical image viewer.
https://github.com/niivue/niivueMicrosoft. ONNX Runtime: cross-platform, high performance inferencing and training accelerator.
https://onnxruntime.aiLi X, Morgan PS, Ashburner J, Smith J, Rorden C. The first step for neuroimaging data analysis: DICOM to NIfTI conversion. J Neurosci Methods. 2016;264:47-56.
https://github.com/rordenlab/dcm2niix