Finished manual tutorial

README.md

@@ -21,7 +21,7 @@ sudo cp /usr/include/cudnn*.h /usr/local/cuda-11.2/include/
 sudo cp -P /usr/lib/x86_64-linux-gnu/libcudnn*.so* /usr/local/cuda-11.2/lib64/
 sudo ldconfig
 ```
-Finally, we must configure the system paths (or add these lines to ~/.bashrc:
+Finally, we must configure the system paths in .bashrc:
 ```
 echo 'export CUDA_HOME=/usr/local/cuda-11.2' >> ~/.bashrc
 echo 'export PATH=$CUDA_HOME/bin:$PATH' >> ~/.bashrc
@@ -135,8 +135,9 @@ After preprocessing, we can generate predictions for the IXI dataset using the s
 ```
 eval $(poetry env activate)
 ```
-Next, run the prediction-script:
+Next, make an output-folder for the predictions and run the prediction-script:
 ```
-python scripts/predict_from_fastsurfer_folder.py
+mkdir ~/data/ixi/outputs
+python scripts/predict_from_fastsurfer_folder.py ~/data/ixi/preprocessed -d ~/data/ixi/outputs/predictions.csv
 ```
 </details>

poetry.lock

@@ -2714,6 +2714,74 @@ files = [
 [package.extras]
 diagrams = ["jinja2", "railroad-diagrams"]
 
+[[package]]
+name = "pyqt5"
+version = "5.15.11"
+description = "Python bindings for the Qt cross platform application toolkit"
+optional = false
+python-versions = ">=3.8"
+groups = ["main"]
+files = [
+    {file = "PyQt5-5.15.11-cp38-abi3-macosx_11_0_arm64.whl", hash = "sha256:c8b03dd9380bb13c804f0bdb0f4956067f281785b5e12303d529f0462f9afdc2"},
+    {file = "PyQt5-5.15.11-cp38-abi3-macosx_11_0_x86_64.whl", hash = "sha256:6cd75628f6e732b1ffcfe709ab833a0716c0445d7aec8046a48d5843352becb6"},
+    {file = "PyQt5-5.15.11-cp38-abi3-manylinux_2_17_x86_64.whl", hash = "sha256:cd672a6738d1ae33ef7d9efa8e6cb0a1525ecf53ec86da80a9e1b6ec38c8d0f1"},
+    {file = "PyQt5-5.15.11-cp38-abi3-win32.whl", hash = "sha256:76be0322ceda5deecd1708a8d628e698089a1cea80d1a49d242a6d579a40babd"},
+    {file = "PyQt5-5.15.11-cp38-abi3-win_amd64.whl", hash = "sha256:bdde598a3bb95022131a5c9ea62e0a96bd6fb28932cc1619fd7ba211531b7517"},
+    {file = "PyQt5-5.15.11.tar.gz", hash = "sha256:fda45743ebb4a27b4b1a51c6d8ef455c4c1b5d610c90d2934c7802b5c1557c52"},
+]
+
+[package.dependencies]
+PyQt5-Qt5 = ">=5.15.2,<5.16.0"
+PyQt5-sip = ">=12.15,<13"
+
+[[package]]
+name = "pyqt5-qt5"
+version = "5.15.17"
+description = "The subset of a Qt installation needed by PyQt5."
+optional = false
+python-versions = "*"
+groups = ["main"]
+files = [
+    {file = "PyQt5_Qt5-5.15.17-py3-none-macosx_10_13_x86_64.whl", hash = "sha256:d8b8094108e748b4bbd315737cfed81291d2d228de43278f0b8bd7d2b808d2b9"},
+    {file = "PyQt5_Qt5-5.15.17-py3-none-macosx_11_0_arm64.whl", hash = "sha256:b68628f9b8261156f91d2f72ebc8dfb28697c4b83549245d9a68195bd2d74f0c"},
+    {file = "PyQt5_Qt5-5.15.17-py3-none-manylinux2014_x86_64.whl", hash = "sha256:b018f75d1cc61146396fa5af14da1db77c5d6318030e5e366f09ffdf7bd358d8"},
+]
+
+[[package]]
+name = "pyqt5-sip"
+version = "12.17.1"
+description = "The sip module support for PyQt5"
+optional = false
+python-versions = ">=3.9"
+groups = ["main"]
+files = [
+    {file = "pyqt5_sip-12.17.1-cp310-cp310-macosx_10_9_universal2.whl", hash = "sha256:bd4f73b1ebd5e0bd8d4539a8e55132318efc70a92f648ef0f9d93329ad50adeb"},
+    {file = "pyqt5_sip-12.17.1-cp310-cp310-manylinux_2_5_x86_64.manylinux1_x86_64.whl", hash = "sha256:b52e85520dbfe5c3d0c0c47aa2c10fc1853d892ae60ebebfe8154b052394da50"},
+    {file = "pyqt5_sip-12.17.1-cp310-cp310-win32.whl", hash = "sha256:71a67e2c9b77a74e943e220db0a341c702fd9bcf83c4a2e07342dfce691742ae"},
+    {file = "pyqt5_sip-12.17.1-cp310-cp310-win_amd64.whl", hash = "sha256:2710effb921bf6955b902779c763d890bb593da6325f0e128a0e3991cc855e9f"},
+    {file = "pyqt5_sip-12.17.1-cp311-cp311-macosx_10_9_universal2.whl", hash = "sha256:5134d637efadd108a70306bab55b3d7feaa951bf6b8162161a67ae847bea9130"},
+    {file = "pyqt5_sip-12.17.1-cp311-cp311-manylinux_2_5_x86_64.manylinux1_x86_64.whl", hash = "sha256:155cf755266c8bf64428916e2ff720d5efa1aec003d4ccc40c003b147dbdac03"},
+    {file = "pyqt5_sip-12.17.1-cp311-cp311-win32.whl", hash = "sha256:9dfa7fe4ac93b60004430699c4bf56fef842a356d64dfea7cbc6d580d0427d6d"},
+    {file = "pyqt5_sip-12.17.1-cp311-cp311-win_amd64.whl", hash = "sha256:2ddd214cf40119b86942a5da2da5a7345334955ab00026d8dcc56326b30e6d3c"},
+    {file = "pyqt5_sip-12.17.1-cp312-cp312-macosx_10_9_universal2.whl", hash = "sha256:c362606de782d2d46374a38523632786f145c517ee62de246a6069e5f2c5f336"},
+    {file = "pyqt5_sip-12.17.1-cp312-cp312-manylinux_2_5_x86_64.manylinux1_x86_64.whl", hash = "sha256:140cc582151456103ebb149fefc678f3cae803e7720733db51212af5219cd45c"},
+    {file = "pyqt5_sip-12.17.1-cp312-cp312-win32.whl", hash = "sha256:9dc1f1525d4d42c080f6cfdfc70d78239f8f67b0a48ea0745497251d8d848b1d"},
+    {file = "pyqt5_sip-12.17.1-cp312-cp312-win_amd64.whl", hash = "sha256:d5e2e9e175559017cd161d661e0ee0b551684f824bb90800c5a8c8a3bea9355e"},
+    {file = "pyqt5_sip-12.17.1-cp313-cp313-macosx_10_13_universal2.whl", hash = "sha256:9ebbd7769ccdaaa6295e9c872553b6cde17f38e171056f17300d8af9a14d1fc8"},
+    {file = "pyqt5_sip-12.17.1-cp313-cp313-manylinux_2_5_x86_64.manylinux1_x86_64.whl", hash = "sha256:b023da906a70af2cf5e6fc1932f441ede07530f3e164dd52c6c2bb5ab7c6f424"},
+    {file = "pyqt5_sip-12.17.1-cp313-cp313-win32.whl", hash = "sha256:36dbef482bd638786b909f3bda65b7b3d5cbd6cbf16797496de38bae542da307"},
+    {file = "pyqt5_sip-12.17.1-cp313-cp313-win_amd64.whl", hash = "sha256:d04e5551bbc3bcec98acc63b3b0618ddcbf31ff107349225b516fe7e7c0a7c8b"},
+    {file = "pyqt5_sip-12.17.1-cp314-cp314-macosx_10_15_universal2.whl", hash = "sha256:c49918287e1ad77956d1589f1d3d432a0be7630c646ea02cf652413a48e14458"},
+    {file = "pyqt5_sip-12.17.1-cp314-cp314-manylinux_2_5_x86_64.manylinux1_x86_64.whl", hash = "sha256:944a4bf1e1ee18ad03a54964c1c6433fb6de582313a1f0b17673e7203e22fc83"},
+    {file = "pyqt5_sip-12.17.1-cp314-cp314-win32.whl", hash = "sha256:99a2935fd662a67748625b1e6ffa0a2d1f2da068b9df6db04fa59a4a5d4ee613"},
+    {file = "pyqt5_sip-12.17.1-cp314-cp314-win_amd64.whl", hash = "sha256:aaa33232cc80793d14fdb3b149b27eec0855612ed66aad480add5ac49b9cee63"},
+    {file = "pyqt5_sip-12.17.1-cp39-cp39-macosx_10_9_universal2.whl", hash = "sha256:6fdc457bd528e5909a5893db0a7dee0066d5f22e08234c9152db0ae6df9a367f"},
+    {file = "pyqt5_sip-12.17.1-cp39-cp39-manylinux_2_5_x86_64.manylinux1_x86_64.whl", hash = "sha256:06ea59741c1bffb198d99b00d26594791f45fb11b10f774c8105aea5962e3835"},
+    {file = "pyqt5_sip-12.17.1-cp39-cp39-win32.whl", hash = "sha256:b9ef23869d35c6740a95fcb1f387f4aea8d8fac80e19096fbaf1a64e18409c4b"},
+    {file = "pyqt5_sip-12.17.1-cp39-cp39-win_amd64.whl", hash = "sha256:90eed15f19557dfab22e68c7763e3690053cc8dd30d93ade2523d1b5a04a87be"},
+    {file = "pyqt5_sip-12.17.1.tar.gz", hash = "sha256:0eab72bcb628f1926bf5b9ac51259d4fa18e8b2a81d199071135458f7d087ea8"},
+]
+
 [[package]]
 name = "pytest"
 version = "8.3.3"
@@ -4088,4 +4156,4 @@ test = ["pytest", "pytest-cov"]
 [metadata]
 lock-version = "2.1"
 python-versions = "3.10.4"
-content-hash = "5bcab6231b21344133a42a97b439b6c847b1ef803c281d2518a40bbfa7d9cef2"
+content-hash = "f4ed40969ecb1dec94a4d0f7f994e17cea56a8e86b72ea95a021f5247942d8a1"

pyment/models/utils/ensure_weights.py

@@ -28,7 +28,8 @@ def _lookup_identifier(identifier: str, local_cache: str) -> str:
                 local_cache,
                 f'{identifier}.data-00000-of-00001'
             ),
-            description=f'Downloading {identifier} data'
+            description=f'Downloading {identifier} data',
+            decode_github=True
         )
         download_file(
             url=BASE_URL + '/' + IDENTIFIERS[identifier]['index'],
@@ -36,7 +37,8 @@ def _lookup_identifier(identifier: str, local_cache: str) -> str:
                 local_cache,
                 f'{identifier}.index'
             ),
-            description=f'Downloading {identifier} index'
+            description=f'Downloading {identifier} index',
+            decode_github=True
         )
 
     return os.path.join(local_cache, identifier)

pyment/preprocessing/conform.py

@@ -3,14 +3,10 @@ import numpy as np
 from typing import Tuple
 
 
-logging.basicConfig(
-    format='%(asctime)s - %(levelname)s - %(name)s: %(message)s',
-    level=logging.DEBUG
-)
 logger = logging.getLogger(__name__)
 
 def _pad_if_necessary(
-    image: np.ndarray, 
+    image: np.ndarray,
     target_shape: Tuple[int, int, int]
 ) -> np.ndarray:
     pad = [(0, 0)] * 3
@@ -25,7 +21,7 @@ def _pad_if_necessary(
     return np.pad(image, tuple(pad), mode='constant', constant_values=0)
 
 def _crop_if_necessary(
-    image: np.ndarray, 
+    image: np.ndarray,
     target_shape: Tuple[int, int, int]
 ) -> np.ndarray:
     nonzero = np.where(image != 0)
@@ -36,7 +32,7 @@ def _crop_if_necessary(
         if image.shape[dim] > target_shape[dim]:
             extrafluous = target_shape[dim] / 2
             center = np.round(np.mean([
-                np.amin(nonzero[dim]), 
+                np.amin(nonzero[dim]),
                 np.amax(nonzero[dim])
             ]))
             min_idx = int(center - extrafluous)
@@ -45,7 +41,7 @@ def _crop_if_necessary(
             if min_idx < 0:
                 max_idx -= min_idx
                 min_idx = 0
-            
+
             if max_idx > image.shape[dim]:
                 diff = max_idx - image.shape[dim]
                 min_idx -= diff
@@ -58,7 +54,7 @@ def _crop_if_necessary(
     return image
 
 def _center_crop_or_pad(
-    image: np.ndarray, 
+    image: np.ndarray,
     target_shape: Tuple[int, int, int]
 ) -> np.ndarray:
     image = _pad_if_necessary(image, target_shape)
@@ -67,22 +63,22 @@ def _center_crop_or_pad(
     return image
 
 def conform(
-    image: np.ndarray, 
+    image: np.ndarray,
     relative_normalization: bool = False
 ) -> np.ndarray:
-    """Conforms an image to the expected format if necessary. The 
+    """Conforms an image to the expected format if necessary. The
     expected format means an image of shape 224x192x224 with voxel
-    values spanning the range [0, 1]. If the image has a redundant 
-    channel-dimension, this is removed. If the image is currently too 
+    values spanning the range [0, 1]. If the image has a redundant
+    channel-dimension, this is removed. If the image is currently too
     large along any dimension, a "central" crop is made by determining
-    the bound of the brain (e.g. non-zero voxels) and retaining 
+    the bound of the brain (e.g. non-zero voxels) and retaining
     equivalent padding on each side. If the image is currently too small
-    along either axis, the image is zero-padded equally on each side. 
+    along either axis, the image is zero-padded equally on each side.
     If the voxel-values does not fall within the expected range, they
-    are normalized. If the relative_normalization-flag is set, the 
+    are normalized. If the relative_normalization-flag is set, the
     values are normalized by dividing by the image max, otherwise they
-    are divided by 255. However, if the largest value is >255, this 
-    indicates that the image has not been processed with FastSurfer, 
+    are divided by 255. However, if the largest value is >255, this
+    indicates that the image has not been processed with FastSurfer,
     and an error is raised.
 
     Parameters
@@ -106,7 +102,7 @@ def conform(
     if len(image.shape) == 4:
         if image.shape[-1] != 1:
             raise ValueError(f'Unable to handle multi-channel images')
-        
+
         image = image[...,0]
 
     if image.shape != (224, 192, 224):
@@ -114,7 +110,7 @@ def conform(
 
     logger.debug('Conformed image shape: %s', str(image.shape))
     logger.debug(
-        'Original image voxel value range: %f-%f', 
+        'Original image voxel value range: %f-%f',
         np.amin(image), np.amax(image)
     )
 
@@ -124,8 +120,8 @@ def conform(
         image *= 255.0
 
     logger.debug(
-        'Conformed image voxel value range: %f-%f', 
+        'Conformed image voxel value range: %f-%f',
         np.amin(image), np.amax(image)
     )
 
-    return image
+    return image

pyment/utils/download_file.py

@@ -1,3 +1,5 @@
+import base64
+import json
 import math
 import requests
 from tqdm import tqdm
@@ -6,7 +8,8 @@ from tqdm import tqdm
 def download_file(
     url: str,
     destination: str,
-    description: str = None
+    description: str = None,
+    decode_github: bool = False
 ) -> str:
     with requests.get(url, stream=True) as response:
         response.raise_for_status()
@@ -28,3 +31,13 @@ def download_file(
         with open(destination, 'wb') as f:
             for chunk in progress_bar:
                 f.write(chunk)
+
+    if decode_github:
+        # Assumes a JSON file downloaded from GitHub
+        with open(destination, 'rb') as f:
+            data = json.load(f)
+
+        data = base64.b64decode(data['content'])
+
+        with open(destination, 'wb') as f:
+            f.write(data)

pyproject.toml

@@ -37,6 +37,7 @@ pytest = "8.3.3"
 scikit-learn = "1.5.1"
 xlrd = "2.0.1"
 pydantic = "2.10"
+pyqt5 = "^5.15.11"
 
 [build-system]
 requires = ["poetry-core>=2.0.0,<3.0.0"]

scripts/predict_from_fastsurfer_folder.py

@@ -13,10 +13,6 @@ from pyment.models.sfcn import sfcn_factory
 from pyment.preprocessing.conform import conform
 
 
-logging.basicConfig(
-    format='%(asctime)s - %(levelname)s - %(name)s: %(message)s',
-    level=logging.DEBUG
-)
 logger = logging.getLogger(__name__)
 
 def _parse_folder_name(name: str) -> Tuple[str, str, str]:
@@ -71,7 +67,10 @@ def predict_from_fastsurfer_folder(
         logger.debug('Conforming image from %s', os.path.join(source, folder))
         image = conform(image)
 
-        predictions = model.predict(np.expand_dims(image, axis=0))[0]
+        predictions = model.predict(
+            np.expand_dims(image, axis=0),
+            verbose=0
+        )[0]
         logger.debug('Predictions for %s: %s', folder, str(predictions))
 
         results.append({

tutorials/evaluate_ixi_predictions.py

@@ -0,0 +1,66 @@
+import argparse
+import matplotlib
+import matplotlib.pyplot as plt
+import numpy as np
+import os
+import pandas as pd
+
+from download_ixi import DEFAULT_DESTINATION
+
+
+def evaluate_ixi_predictions(
+    labels: str,
+    predictions: str
+) -> None:
+    labels = pd.read_excel(labels)
+    print(labels.head())
+    predictions = pd.read_csv(predictions)
+    print(predictions.head())
+    predictions['IXI_ID'] = predictions['source'].apply(
+        lambda path: int(path.split('/')[-1][3:6])
+    )
+    predictions['age_prediction'] = predictions['age']
+    predictions = pd.merge(
+        predictions[['IXI_ID', 'age_prediction']],
+        labels[['IXI_ID', 'AGE']],
+        on='IXI_ID',
+        how='left'
+    )
+
+    mae = np.mean(np.abs(predictions['AGE'] - predictions['age_prediction']))
+    print(f'MAE: {mae}')
+
+    plt.scatter(predictions['AGE'], predictions['age_prediction'])
+    plt.xlabel('True age')
+    plt.ylabel('Predicted age')
+    plt.title('Age prediction')
+    plt.show()
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(
+        'Evaluates predictions for the IXI dataset'
+    )
+    parser.add_argument(
+        '-l', '--labels',
+        required=False,
+        default=os.path.join(DEFAULT_DESTINATION, 'IXI.xls'),
+        help='Path to XLSX containing labels'
+    )
+    parser.add_argument(
+        '-p', '--predictions',
+        required=False,
+        default=os.path.join(
+            DEFAULT_DESTINATION,
+            'outputs',
+            'predictions.csv'
+        ),
+        help='Path to CSV containing predictions'
+    )
+
+    args = parser.parse_args()
+
+    evaluate_ixi_predictions(
+        labels=args.labels,
+        predictions=args.predictions
+    )