Datasets:

pyushkevich
/

crashs_template_package

text stringlengths 0 1.22k

#######################################################################
Please cite the following paper when using nnU-Net:
Isensee, F., Jaeger, P. F., Kohl, S. A., Petersen, J., & Maier-Hein, K. H. (2021). nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nature methods, 18(2), 203-211.
#######################################################################


This is the configuration used by this training:
Configuration name: 3d_fullres
{'data_identifier': 'nnUNetPlans_3d_fullres', 'preprocessor_name': 'DefaultPreprocessor', 'batch_size': 2, 'patch_size': [64, 128, 128], 'median_image_size_in_voxels': [64.0, 113.0, 113.0], 'spacing': [0.9765625, 0.5, 0.48828125], 'normalization_schemes': ['ZScoreNormalization'], 'use_mask_for_norm': [False], 'UNet_class_name': 'PlainConvUNet', 'UNet_base_num_features': 32, 'n_conv_per_stage_encoder': [2, 2, 2, 2, 2, 2], 'n_conv_per_stage_decoder': [2, 2, 2, 2, 2], 'num_pool_per_axis': [4, 5, 5], 'pool_op_kernel_sizes': [[1, 1, 1], [1, 2, 2], [2, 2, 2], [2, 2, 2], [2, 2, 2], [2, 2, 2]], 'conv_kernel_sizes': [[1, 3, 3], [3, 3, 3], [3, 3, 3], [3, 3, 3], [3, 3, 3], [3, 3, 3]], 'unet_max_num_features': 320, 'resampling_fn_data': 'resample_data_or_seg_to_shape', 'resampling_fn_seg': 'resample_data_or_seg_to_shape', 'resampling_fn_data_kwargs': {'is_seg': False, 'order': 3, 'order_z': 0, 'force_separate_z': None}, 'resampling_fn_seg_kwargs': {'is_seg': True, 'order': 1, 'order_z': 0, 'force_separate_z': None}, 'resampling_fn_probabilities': 'resample_data_or_seg_to_shape', 'resampling_fn_probabilities_kwargs': {'is_seg': False, 'order': 1, 'order_z': 0, 'force_separate_z': None}, 'batch_dice': False}

These are the global plan.json settings:
{'dataset_name': 'Dataset303_3TT1WMSegASHSGT', 'plans_name': 'nnUNetPlans', 'original_median_spacing_after_transp': [0.9765625, 0.5, 0.48828125], 'original_median_shape_after_transp': [64, 113, 113], 'image_reader_writer': 'SimpleITKIO', 'transpose_forward': [1, 0, 2], 'transpose_backward': [1, 0, 2], 'experiment_planner_used': 'ExperimentPlanner', 'label_manager': 'LabelManager', 'foreground_intensity_properties_per_channel': {'0': {'max': 685.0, 'mean': 356.69915771484375, 'median': 396.0, 'min': 117.0, 'percentile_00_5': 149.0, 'percentile_99_5': 570.0, 'std': 120.57451629638672}}}

2023-08-02 21:58:34.149126: unpacking dataset...
2023-08-02 21:58:34.473478: unpacking done...
2023-08-02 21:58:34.474119: do_dummy_2d_data_aug: False
2023-08-02 21:58:34.474694: Using splits from existing split file: /data/liyue7/Data/M1_nnUNet/nnUNet_preprocessed/Dataset303_3TT1WMSegASHSGT/splits_final.json
2023-08-02 21:58:34.474854: The split file contains 5 splits.
2023-08-02 21:58:34.474900: Desired fold for training: 0
2023-08-02 21:58:34.474939: This split has 46 training and 12 validation cases.
2023-08-02 21:58:34.480075: Unable to plot network architecture:
2023-08-02 21:58:34.480146: No module named 'hiddenlayer'
2023-08-02 21:58:34.519761:
2023-08-02 21:58:34.519815: Epoch 0
2023-08-02 21:58:34.519891: Current learning rate: 0.01
2023-08-02 22:00:03.004448: train_loss -0.1602
2023-08-02 22:00:03.004621: val_loss -0.6431
2023-08-02 22:00:03.004667: Pseudo dice [0.7123]
2023-08-02 22:00:03.004718: Epoch time: 88.49 s
2023-08-02 22:00:03.004757: Yayy! New best EMA pseudo Dice: 0.7123
2023-08-02 22:00:04.079823:
2023-08-02 22:00:04.079919: Epoch 1
2023-08-02 22:00:04.080002: Current learning rate: 0.00998
2023-08-02 22:01:06.331137: train_loss -0.6851
2023-08-02 22:01:06.331275: val_loss -0.7601
2023-08-02 22:01:06.331314: Pseudo dice [0.8097]
2023-08-02 22:01:06.331359: Epoch time: 62.25 s
2023-08-02 22:01:06.331396: Yayy! New best EMA pseudo Dice: 0.7221
2023-08-02 22:01:08.636403:
2023-08-02 22:01:08.636506: Epoch 2
2023-08-02 22:01:08.636586: Current learning rate: 0.00995
2023-08-02 22:02:11.011815: train_loss -0.7477
2023-08-02 22:02:11.011961: val_loss -0.8005
2023-08-02 22:02:11.012002: Pseudo dice [0.8433]
2023-08-02 22:02:11.012047: Epoch time: 62.38 s
2023-08-02 22:02:11.012084: Yayy! New best EMA pseudo Dice: 0.7342
2023-08-02 22:02:13.056767:
2023-08-02 22:02:13.056870: Epoch 3
2023-08-02 22:02:13.056954: Current learning rate: 0.00993
2023-08-02 22:03:15.491260: train_loss -0.7857
2023-08-02 22:03:15.491742: val_loss -0.8215
2023-08-02 22:03:15.491786: Pseudo dice [0.8605]
2023-08-02 22:03:15.491831: Epoch time: 62.44 s
2023-08-02 22:03:15.491870: Yayy! New best EMA pseudo Dice: 0.7468
2023-08-02 22:03:17.505135:
2023-08-02 22:03:17.505237: Epoch 4
2023-08-02 22:03:17.505322: Current learning rate: 0.00991
2023-08-02 22:04:19.921209: train_loss -0.8016
2023-08-02 22:04:19.921346: val_loss -0.8273
2023-08-02 22:04:19.921388: Pseudo dice [0.8615]
2023-08-02 22:04:19.921433: Epoch time: 62.42 s
2023-08-02 22:04:19.921470: Yayy! New best EMA pseudo Dice: 0.7583
2023-08-02 22:04:21.908744:
2023-08-02 22:04:21.908846: Epoch 5
2023-08-02 22:04:21.908923: Current learning rate: 0.00989
2023-08-02 22:05:24.340911: train_loss -0.8115
2023-08-02 22:05:24.341050: val_loss -0.8428
2023-08-02 22:05:24.341090: Pseudo dice [0.8746]
2023-08-02 22:05:24.341135: Epoch time: 62.43 s
2023-08-02 22:05:24.341172: Yayy! New best EMA pseudo Dice: 0.7699
2023-08-02 22:05:26.416294:
2023-08-02 22:05:26.416399: Epoch 6
2023-08-02 22:05:26.416478: Current learning rate: 0.00986
2023-08-02 22:06:28.831770: train_loss -0.8274
2023-08-02 22:06:28.831909: val_loss -0.8475
2023-08-02 22:06:28.831949: Pseudo dice [0.8801]
2023-08-02 22:06:28.831995: Epoch time: 62.42 s
2023-08-02 22:06:28.832032: Yayy! New best EMA pseudo Dice: 0.7809
2023-08-02 22:06:30.999928:
2023-08-02 22:06:31.000031: Epoch 7
2023-08-02 22:06:31.000112: Current learning rate: 0.00984
2023-08-02 22:07:33.432782: train_loss -0.8394
2023-08-02 22:07:33.432915: val_loss -0.8555
2023-08-02 22:07:33.432956: Pseudo dice [0.8845]
2023-08-02 22:07:33.433001: Epoch time: 62.43 s
2023-08-02 22:07:33.433038: Yayy! New best EMA pseudo Dice: 0.7913
2023-08-02 22:07:35.622762:
2023-08-02 22:07:35.622863: Epoch 8
2023-08-02 22:07:35.622944: Current learning rate: 0.00982
2023-08-02 22:08:38.060140: train_loss -0.838
2023-08-02 22:08:38.060285: val_loss -0.8542
2023-08-02 22:08:38.060333: Pseudo dice [0.883]
2023-08-02 22:08:38.060412: Epoch time: 62.44 s
2023-08-02 22:08:38.060493: Yayy! New best EMA pseudo Dice: 0.8005
2023-08-02 22:08:40.178575:
2023-08-02 22:08:40.178674: Epoch 9
2023-08-02 22:08:40.178754: Current learning rate: 0.0098
2023-08-02 22:09:42.615384: train_loss -0.8516
2023-08-02 22:09:42.615526: val_loss -0.8548

End of preview. Expand in Data Studio

CRASHS Template and Model Package

This folder contains the templates and deep learning models needed to run CRASHS (cortical reconstruction for automated segmentation of hippocampal subfields). Please see CRASHS github page for details on using this dataset.

This package is compatible with CRASHS version 0.2.10 and later

Downloads last month: 94

Size of downloaded dataset files:

713 kB

Size of the auto-converted Parquet files:

252 kB

Number of rows:

14,948