docs(model_card): Added model cards for current models (#111)

* feaT(model_card): Adding base model_card to models

* feat(model_card): Added Model Card for DDIG models

* feat(model_card): Added model card for DeepCSR

* feat(model_card): Added model card to lcn model

* fix(spec): Cleaned spec

* feat(model_card): Added mocel card for ams and braingen

* feat(model_card): Added model card for brainy

* feat(model_card): Added model card for kwyk

* feat(model_card): Added model card for SynthSeg

* feat(model_card): Added model card for SynthSR

* fix(model_card): Updated ams model card

* fix(model_card): Fixed information on card of DDIG

DDIG/SynthMorph/1.0.0/brains/model_card.yaml

@@ -0,0 +1,43 @@
+Model_details:
+  Organization: "DDIG"
+  Model_date: "2020"
+  Model_version: 1.0.0
+  Model_type: "U-Net"
+  More_information: "brains"
+  Citation_details: "Hoffmann, M., Billot, B., Greve, D. N., Iglesias, J. E., Fischl, B., & Dalca, A. V. (2020). SynthMorph: Learning contrast-invariant registration without acquired images. ArXiv. https://doi.org/10.1109/TMI.2021.3116879"
+  Contact_info: "https://github.com/voxelmorph/voxelmorph/issues/new"
+
+Intended_use:
+  Primary_intended_uses: "Learning contrast-invariant registration without relying on acquired imaging data."
+  Primary_intended_users: "Researchers and clinicians in neuroimaging and medical image analysis."
+  Out_of_scope_use_cases: "Not intended for direct clinical diagnosis or treatment planning."
+
+Factors:
+  Relevant_factors: "MRI contrasts, neural network training strategies."
+  Evaluation_factors: "Generalization to a broad array of MRI contrasts, robustness, accuracy."
+  Model_performance_measures: "Registration accuracy, contrast invariance, computational efficiency."
+
+Metrics:
+  Model Performance Measures: "Dice scores, symmetric surface distances, warp folding proportion."
+  Decision Thresholds:
+  Variation Approaches: "Assessment across various MRI contrasts and processing levels."
+
+Evaluation Data:
+  Datasets: "Includes brain MRI datasets from OASIS, HCP-A, BIRN, UK Biobank, and more."
+  Motivation: "To evaluate robustness to contrast variations and generalizability."
+  Preprocessing: "Standard neuroimaging preprocessing steps, including skull-stripping and normalization."
+
+Training Data:
+  Datasets: "Synthetic data generated from label maps using a generative model."
+  Motivation: "To achieve contrast-invariant registration capabilities."
+  Preprocessing: "Synthesis of images from label maps to create training data with wide-ranging variability."
+
+Quantitative Analyses:
+  Unitary Results: "Performance metrics like Dice scores and surface distances."
+  Intersectional Results: "Analysis across different MRI contrasts and datasets."
+
+Ethical Considerations:
+  "Adherence to ethical standards in AI and medical imaging research, particularly regarding data privacy and responsible use."
+
+Caveats and Recommendations:
+  "Recognition of the model's limitations in clinical settings and the importance of validation with real-world data."

DDIG/SynthMorph/1.0.0/brains/spec.yaml

@@ -60,25 +60,3 @@ training_data_info:
         metrics: None
         data_preprocessing: None
         data_augmentation: None
-
-#### model information and help
-model:
-    model_name: "SynthMorph"
-    description: "3D brain registration model"
-    structure: "U-Net"
-    training_mode: "unsupervised"
-    model_url: "https://github.com/voxelmorph/voxelmorph"
-    Zoo_function: "register"
-    example: "nobrainer-zoo register -m DDIG/SynthMorph/1.0.0/brains --model_type brains <path_to_moving_image> <path_to_fixed_image> <path_to_registered_output>/out.npz"
-    note: "Please provide an output file name with an extension."
-    input_file_type: ".npz"
-    model_details: ""
-    intended_use: ""
-    factors: ""
-    metrics: ""
-    eval_data: ""
-    training_data: ""
-    quant_analyses: ""
-    ethical_considerations: ""
-    caveats_recs: ""
-

DDIG/SynthMorph/1.0.0/shapes/model_card.yaml

@@ -0,0 +1,43 @@
+Model_details:
+  Organization: "DDIG"
+  Model_date: "2020"
+  Model_version: 1.0.0
+  Model_type: "U-Net"
+  More_information: "shapes"
+  Citation_details: "Hoffmann, M., Billot, B., Greve, D. N., Iglesias, J. E., Fischl, B., & Dalca, A. V. (2020). SynthMorph: Learning contrast-invariant registration without acquired images. ArXiv. https://doi.org/10.1109/TMI.2021.3116879"
+  Contact_info: "https://github.com/voxelmorph/voxelmorph/issues/new"
+
+Intended_use:
+  Primary_intended_uses: "Learning contrast-invariant registration without relying on acquired imaging data."
+  Primary_intended_users: "Researchers and clinicians in neuroimaging and medical image analysis."
+  Out_of_scope_use_cases: "Not intended for direct clinical diagnosis or treatment planning."
+
+Factors:
+  Relevant_factors: "MRI contrasts, neural network training strategies."
+  Evaluation_factors: "Generalization to a broad array of MRI contrasts, robustness, accuracy."
+  Model_performance_measures: "Registration accuracy, contrast invariance, computational efficiency."
+
+Metrics:
+  Model Performance Measures: "Dice scores, symmetric surface distances, warp folding proportion."
+  Decision Thresholds:
+  Variation Approaches: "Assessment across various MRI contrasts and processing levels."
+
+Evaluation Data:
+  Datasets: "Includes brain MRI datasets from OASIS, HCP-A, BIRN, UK Biobank, and more."
+  Motivation: "To evaluate robustness to contrast variations and generalizability."
+  Preprocessing: "Standard neuroimaging preprocessing steps, including skull-stripping and normalization."
+
+Training Data:
+  Datasets: "Synthetic data generated from label maps using a generative model."
+  Motivation: "To achieve contrast-invariant registration capabilities."
+  Preprocessing: "Synthesis of images from label maps to create training data with wide-ranging variability."
+
+Quantitative Analyses:
+  Unitary Results: "Performance metrics like Dice scores and surface distances."
+  Intersectional Results: "Analysis across different MRI contrasts and datasets."
+
+Ethical Considerations:
+  "Adherence to ethical standards in AI and medical imaging research, particularly regarding data privacy and responsible use."
+
+Caveats and Recommendations:
+  "Recognition of the model's limitations in clinical settings and the importance of validation with real-world data."

DDIG/SynthMorph/1.0.0/shapes/spec.yaml

@@ -60,24 +60,3 @@ training_data_info:
         metrics: None
         data_preprocessing: None
         data_augmentation: None
-
-#### model information and help
-model:
-    model_name: "SynthMorph"
-    description: "3D brain registration model"
-    structure: "U-Net"
-    training_mode: "unsupervised"
-    model_url: "https://github.com/voxelmorph/voxelmorph"
-    Zoo_function: "register"
-    example: "nobrainer-zoo register -m DDIG/SynthMorph/1.0.0 --model_type shapes <path_to_moving_image> <path_to_fixed_image> <path_to_registered_output>/out.npz"
-    note: "Please provide an output file name with an extension."
-    input_file_type: ".npz"
-    model_details: ""
-    intended_use: ""
-    factors: ""
-    metrics: ""
-    eval_data: ""
-    training_data: ""
-    quant_analyses: ""
-    ethical_considerations: ""
-    caveats_recs: ""

DDIG/SynthStrip/1.0.0/model_card.yaml

@@ -0,0 +1,43 @@
+Model_details:
+  Organization: "DDIG"
+  Model_date: "2022"
+  Model_version: 1.0.0
+  Model_type: "U-Net"
+  More_information: "SynthStrip"
+  Citation_details: "Andrew Hoopes, Jocelyn S. Mora, Adrian V. Dalca, Bruce Fischl, Malte Hoffmann, SynthStrip: skull-stripping for any brain image, NeuroImage, Volume 260, 2022, 119474, ISSN 1053-8119, https://doi.org/10.1016/j.neuroimage.2022.119474."
+  Contact_info: "https://github.com/freesurfer/freesurfer/issues/new"
+
+Intended_use:
+  Primary_intended_uses: "Skull-stripping in various imaging modalities, resolutions, and subject populations."
+  Primary_intended_users: "Researchers and practitioners in medical imaging."
+  Out_of_scope_use_cases: "Non-medical image processing, applications outside brain imaging."
+
+Factors:
+  Relevant_factors: "Image quality, resolution, imaging modalities (MRI, CT, PET), subject age and health condition (infants to adults, including glioblastoma patients)"
+  Evaluation_factors: "Robustness to variations in imaging conditions, accuracy of brain voxel extraction."
+  Model_performance_measures: "Accuracy in brain voxel extraction, compatibility with different imaging modalities."
+
+Metrics:
+  Model Performance Measures: "Accuracy in skull-stripping across different imaging modalities and subject conditions."
+  Decision Thresholds: "Thresholds for voxel classification as brain or non-brain tissue."
+  Variation Approaches: "Adaptation to different imaging conditions and subject demographics."
+
+Evaluation Data:
+  Datasets: "622 MRI, CT, and PET scans with corresponding ground-truth brain masks."
+  Motivation: "To evaluate performance across diverse imaging types and subject populations."
+  Preprocessing: "Standardization of image formats and resolution normalization."
+
+Training Data:
+  Datasets: "131 adult MPRAGE scans with FreeSurfer brain labels and additional non-brain labels."
+  Motivation: "To create a robust model capable of accurately segmenting brain tissue in diverse imaging contexts."
+  Preprocessing: "Segmentation map synthesis, resolution normalization, data augmentation for robustness."
+
+Quantitative Analyses:
+  Unitary Results: "Performance metrics for individual modalities and subject conditions."
+  Intersectional Results: "Analysis across combinations of imaging modalities and subject demographics."
+
+Ethical Considerations:
+  "Consideration of patient privacy and data security, especially given the sensitive nature of medical imaging data."
+
+Caveats and Recommendations:
+  "Users should be aware of the model's limitations in extreme imaging conditions. Continuous validation with new data is recommended to maintain performance accuracy."

DDIG/SynthStrip/1.0.0/spec.yaml

@@ -59,24 +59,3 @@ training_data_info:
         metrics: None
         data_preprocessing: None
         data_augmentation: "domain randomization using a generator"
-
-#### model information and help
-model:
-    model_name: "SynthStrip"
-    description: "3D brain extraction model"
-    structure: "U-Net"
-    training_mode: "supervised"
-    model_url: "https://surfer.nmr.mgh.harvard.edu/docs/synthstrip/"
-    Zoo_function: "predict"
-    example: "nobrainer-zoo predict -m DDIG/SynthStrip/1.0.0 <path_to_input> <path_to_output>/out.nii.gz"
-    note: "Please provide an output file name with an extension."
-    input_file_type: "nii.gz"
-    model_details: ""
-    intended_use: ""
-    factors: ""
-    metrics: ""
-    eval_data: ""
-    training_data: ""
-    quant_analyses: ""
-    ethical_considerations: ""
-    caveats_recs: ""

DDIG/VoxelMorph/1.0.0/model_card.yaml

@@ -0,0 +1,43 @@
+Model_details:
+  Organization: "DDIG"
+  Model_date: "2018"
+  Model_version: 1.0.0
+  Model_type: "U-Net"
+  More_information: "VoxelMorph"
+  Citation_details: "Balakrishnan, G., Zhao, A., Sabuncu, M. R., Guttag, J., & Dalca, A. V. (2018). VoxelMorph: A Learning Framework for Deformable Medical Image Registration. ArXiv. https://doi.org/10.1109/TMI.2019.2897538"
+  Contact_info: "https://github.com/voxelmorph/voxelmorph/issues/new?labels=voxelmorph"
+
+Intended_use:
+  Primary_intended_uses: "Deformable medical image registration for various imaging studies."
+  Primary_intended_users: "Researchers and professionals in medical imaging and computational anatomy."
+  Out_of_scope_use_cases: "Non-medical image processing, applications outside deformable registration."
+
+Factors:
+  Relevant_factors: "Image type variability, subject diversity, anatomical variations."
+  Evaluation_factors: "Accuracy of registration, adaptability to different imaging conditions."
+  Model_performance_measures: "Image matching objective functions, overlap of anatomical segmentations."
+
+Metrics:
+  Model Performance Measures: "Dice coefficient for registration accuracy, runtime efficiency."
+  Decision Thresholds: "Thresholds in loss functions for image similarity and deformation smoothness."
+  Variation Approaches: "Adaptation to different datasets, integration of auxiliary data for improved registration."
+
+Evaluation Data:
+  Datasets: "Used for atlas-based and subject-to-subject registration; includes 3731 T1-weighted MRI scans from various datasets."
+  Motivation: "To validate the model's effectiveness across diverse imaging types and subject populations."
+  Preprocessing: "Affine spatial normalization, brain extraction, and anatomical segmentation using FreeSurfer."
+
+Training Data:
+  Datasets: "Training on a large-scale multi-site dataset including T1-weighted MRI scans."
+  Motivation: "To develop a robust and efficient model for deformable image registration."
+  Preprocessing: "Standard preprocessing steps including affine normalization and brain extraction."
+
+Quantitative Analyses:
+  Unitary Results: "Registration accuracy and runtime comparisons with state-of-the-art methods."
+  Intersectional Results: "Performance analysis across different datasets, subject groups, and imaging conditions."
+
+Ethical Considerations:
+  "Ensuring the privacy and security of sensitive medical imaging data."
+
+Caveats and Recommendations:
+  "Users should be aware of the model's limitations in challenging registration scenarios. Regular updates and validations are recommended for maintaining accuracy and efficiency."

DDIG/VoxelMorph/1.0.0/spec.yaml

@@ -61,23 +61,3 @@ training_data_info:
         data_preprocessing: None
         data_augmentation: None
 
-#### model information and help
-model:
-    model_name: "VoxelMorph"
-    description: "3D brain registration model"
-    structure: "U-Net"
-    training_mode: "unsupervised"
-    model_url: "https://github.com/voxelmorph/voxelmorph"
-    Zoo_function: "register"
-    example: "nobrainer-zoo register -m DDIG/VoxelMorph/1.0.0 <path_to_moving_image> <path_to_fixed_image> <path_to_registered_output>/out.npz"
-    note: "Please provide an output file name with an extension."
-    input_file_type: ".npz"
-    model_details: ""
-    intended_use: ""
-    factors: ""
-    metrics: ""
-    eval_data: ""
-    training_data: ""
-    quant_analyses: ""
-    ethical_considerations: ""
-    caveats_recs: ""

DeepCSR/deepcsr/1.0/model_card.yaml

@@ -1,33 +1,43 @@
 Model_details:
-  Organization: DeepCSR
-  Model_date: t
-  Model_version: 1.0
-  Model_type: t
-  More_information: t
-  Citation_details: t
-  Contact_info: t
+  Organization: "DeepCSR"
+  Model_date: "2020"
+  Model_version: 1.0.0
+  Model_type: "Convolutional Neural Network"
+  More_information: "DeepCSR"
+  Citation_details: "Cruz, R. S., Lebrat, L., Bourgeat, P., Fookes, C., Fripp, J., & Salvado, O. (2020). DeepCSR: A 3D Deep Learning Approach for Cortical Surface Reconstruction. ArXiv. /abs/2010.11423"
+  Contact_info: "https://github.com/neuroneural/DeepCSR-fork/issues/new"
+
 Intended_use:
-  Primary_intended_uses: t
-  Primary_intended_users: t
-  Out_of_scope_use_cases: t
+  Primary_intended_uses: "Cortical surface reconstruction from magnetic resonance imaging (MRI)."
+  Primary_intended_users: "Researchers and practitioners in neuroimaging and neurodegenerative disease studies."
+  Out_of_scope_use_cases: "Applications outside of MRI-based brain imaging."
+
 Factors:
-  Relevant_factors: t
-  Evaluation_factors: t
-  Model_performance_measures: t
+  Relevant_factors: "Variability in MRI scans, subject demographics, and cortical surface complexity."
+  Evaluation_factors: "Precision in cortical surface reconstruction, adaptability to different MRI datasets."
+  Model_performance_measures: "Accuracy in cortical surface reconstruction, speed of processing."
+
 Metrics:
-  Model Performance Measures: t
-  Decision Thresholds: t
-  Variation Approaches: t
+  Model Performance Measures: "Reconstruction accuracy, runtime efficiency."
+  Decision Thresholds: "Thresholds in hypercolumn feature extraction and surface representation."
+  Variation Approaches: "Adapting to different MRI resolutions and cortical surface complexities."
+
 Evaluation Data:
-  Datasets: t
-  Motivation: t
-  Preprocessing: t
+  Datasets: "Alzheimer’s Disease Neuroimaging Initiative (ADNI) study dataset."
+  Motivation: "To assess performance in a clinically relevant context with a variety of brain images."
+  Preprocessing: "Affine registration to a brain template, implicit surface representation computation."
+
 Training Data:
-  Datasets: t
-  Motivation: t
-  Preprocessing: t
+  Datasets: "MRI data and corresponding pseudo-ground truth surfaces generated with FreeSurfer V6.0."
+  Motivation: "To develop a model capable of accurate and efficient cortical surface reconstruction."
+  Preprocessing: "Co-registering MR images to a brain template, point sampling near the target surface."
+
 Quantitative Analyses:
-  Unitary Results: t
-  Intersectional Results: t
-Ethical Considerations: t
-Caveats and Recommendations: t
+  Unitary Results: "Comparison with traditional cortical reconstruction methods like FreeSurfer."
+  Intersectional Results: "Performance analysis across different MRI scans and cortical surface types."
+
+Ethical Considerations:
+  "Data privacy and ethical use of MRI scans, especially in the context of neurodegenerative diseases."
+
+Caveats and Recommendations:
+  "Users should be aware of the model's limitations in extremely complex cortical surfaces. Continuous updates and validation with diverse datasets are recommended."

UCL/SynthSR/1.0.0/general/model_card.yaml

@@ -0,0 +1,43 @@
+Model_details:
+  Organization: "UCL"
+  Model_date: "2021"
+  Model_version: 1.0.0
+  Model_type: "U-Net"
+  More_information: "general"
+  Citation_details: "Billot, B., Greve, D. N., Puonti, O., Thielscher, A., Van Leemput, K., Fischl, B., Dalca, A. V., & Iglesias, J. E. (2021). SynthSeg: Segmentation of brain MRI scans of any contrast and resolution without retraining. ArXiv. https://doi.org/10.1016/j.media.2023.102789"
+  Contact_info: "https://github.com/BBillot/SynthSR/issues/new"
+
+Intended_use:
+  Primary_intended_uses: "Super-resolution and synthesis of MRI scans to 1 mm isotropic MP-RAGE volumes, lesion inpainting."
+  Primary_intended_users: "Researchers and clinicians in neuroimaging."
+  Out_of_scope_use_cases: "Non-MRI based imaging applications and non-clinical uses."
+
+Factors:
+  Relevant_factors: "MRI scan orientation, resolution, and contrast; lesion presence."
+  Evaluation_factors: "Accuracy in super-resolution and synthesis, lesion inpainting effectiveness."
+  Model_performance_measures: "Quality of generated 1 mm isotropic MP-RAGE volumes, lesion inpainting accuracy."
+
+Metrics:
+  Model Performance Measures: "Quality assessment of synthesized MP-RAGE volumes, lesion inpainting effectiveness."
+  Decision Thresholds:
+  Variation Approaches: "Assessment across different MRI modalities and resolutions."
+
+Evaluation Data:
+  Datasets: "Clinical MRI and CT scans with different orientations, resolutions, and contrasts."
+  Motivation: "To evaluate the model's ability to generate high-quality MP-RAGE volumes from diverse clinical scans."
+  Preprocessing: "Handling various MRI and CT scan formats and conditions."
+
+Training Data:
+  Datasets: "Synthetic data generated using a generative model based on SynthSeg."
+  Motivation: "To train a network for effective super-resolution and synthesis of MRI scans."
+  Preprocessing: "Synthetic data generation with randomized imaging parameters."
+
+Quantitative Analyses:
+  Unitary Results: "Performance analysis on individual datasets."
+  Intersectional Results: "Analysis across different scan types and conditions."
+
+Ethical Considerations:
+  "General ethical considerations in AI and medical imaging, such as data privacy and responsible use."
+
+Caveats and Recommendations:
+  "Awareness of model's limitations and importance of supplementary clinical evaluation."