OSCR

AI-Based Pipeline for the Segmentation of White Matter Hypoattenuations in CT Scans: A Design-Choice Validation Study

Code ↔ Paper

2 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 2 matches
  1. [1] § Methods and Materials › Image Preprocessing and Registration › Format Conversion and Quality Control ↔ src/nifti_selection.py, lines 81–157 · score 0.77 · gantry tilt, dcm2niix, NIfTI slices, equidistant, DICOM
  2. [2] § Methods and Materials › Image Preprocessing and Registration › Format Conversion and Quality Control ↔ src/datasets/structure.py, lines 18–157 · score 0.69 · dcm2niix, NIfTI, equidistant, tilt, resamples, matched

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The authors' code

Python · 381 lines · 16 KB · no license · 1 match

  1. """Pipeline step summaries"""
  2. import json
  3. from pathlib import Path
  5. import pandas as pd
  7. from src.create_region_of_interest import select_regions_of_interest
  8. from src.datasets.structure import DatasetStructure
  9. from src.determine_patient_orientation import get_patient_orientation, PatientOrientation
  10. from src.read_nifti_information import get_nifti_headers, NIFTI_SELECTION_FILES
  12. NON_INFORMATIVE_NIFTI_COLUMNS = [
  13. 'ConversionSoftware',
  14. 'ConversionSoftwareVersion',
  15. 'SeriesNumber',
  16. 'ImageComments'
  17. ]
  20. def clean_nifti_info(
  21. base_path: Path,
  22. column_missing_data_threshold: int = 5,
  23. ):
  24. """
  25. Clean the nifti info dataframe.
  27. Args:
  28. column_missing_data_threshold: Maximum percentage of missing data in a column before it is rejected.
  29. """
  30. nifti_info_file_path = base_path / NIFTI_SELECTION_FILES[0]
  31. cleaned_nifti_info_file_path = base_path / NIFTI_SELECTION_FILES[1]
  33. all_nifti_info = pd.read_pickle(nifti_info_file_path)
  35. # Convert vector to PatientOrientation
  36. all_nifti_info['ImageOrientationPatientDICOM'] = all_nifti_info['ImageOrientationPatientDICOM'].apply(
  37. get_patient_orientation)
  39. # Drop unnecessary columns
  40. attribute_frequencies = all_nifti_info.isnull().sum()
  41. five_percent_of_data = len(all_nifti_info) / 100 * column_missing_data_threshold
  42. # Keep columns that have less than 5% missing data
  43. usually_available_attributes = all_nifti_info[
  44. attribute_frequencies[attribute_frequencies <= five_percent_of_data].index]
  46. usually_available_attributes = usually_available_attributes.drop(NON_INFORMATIVE_NIFTI_COLUMNS, axis=1, errors='ignore')
  48. usually_available_attributes.to_pickle(cleaned_nifti_info_file_path)
  51. def add_statistics(
  52. summary: dict,
  53. step_df: pd.DataFrame,
  54. step_name: str,
  55. explanation: str
  56. ):
  57. summary['step_name'].append(step_name)
  58. summary['explanation'].append(explanation)
  60. summary['nifti_file_count'].append(len(step_df))
  61. summary['original_version_count'].append(
  62. len(step_df.groupby(['patient_id', 'study_subdir', 'series_subdir']).count()))
  63. summary['patient_count'].append(len(step_df.groupby(['patient_id']).count()))
  65. return summary
  68. def add_roi_statistics(summary_object, step_df, step_name, explanation):
  69. summary_object['step_name'].append(step_name)
  70. summary_object['explanation'].append(explanation)
  72. summary_object['roi_count'].append(len(step_df))
  73. summary_object['nifti_file_count'].append(len(step_df.groupby('registered_file_path').count()))
  74. summary_object['original_version_count'].append(
  75. len(step_df.groupby(['patient_id', 'study_subdir', 'series_subdir']).count()))
  76. summary_object['patient_count'].append(len(step_df.groupby(['patient_id']).count()))
  78. return summary_object
  81. def filter_nifti_files(
  82. base_path: Path,
  83. dataset_folder_structure: DatasetStructure,
  84. ):
  85. cleaned_nifti_info_file_path = base_path / NIFTI_SELECTION_FILES[1]
  86. filtered_nifti_info_file_path = base_path / NIFTI_SELECTION_FILES[2]
  88. summary_statistics = {
  89. 'step_name': [],
  90. 'nifti_file_count': [],
  91. 'original_version_count': [],
  92. 'patient_count': [],
  93. 'explanation': []
  94. }
  96. cleaned_nifti_info: pd.DataFrame = pd.read_pickle(cleaned_nifti_info_file_path)
  97. summary_statistics = add_statistics(
  98. summary_statistics,
  99. cleaned_nifti_info,
  100. 'Original nifti files produced by dcm2niix v1.0.20230411',
  101. 'For a subset of the data, dcm2niix creates multiple nifti files per DICOM. See https://github.com/rordenlab/dcm2niix/blob/master/FILENAMING.md'
  102. )
  104. # Filter for only axial orientations
  105. only_axial = cleaned_nifti_info.loc[cleaned_nifti_info['ImageOrientationPatientDICOM'] == PatientOrientation.AXIAL]
  106. summary_statistics = add_statistics(
  107. summary_statistics,
  108. only_axial,
  109. 'Only axial files',
  110. 'Remove files where the Image Orientation (Patient) is not axial.'
  111. )
  113. # Filter for non-localizers
  114. only_axial = _add_slice_count(
  115. only_axial,
  116. base_path,
  117. dataset_folder_structure
  118. )
  119. no_localizer_axial = only_axial.loc[
  120. (only_axial['ImageType'].apply(lambda x: 'LOCALIZER' not in x if x is not None else False))
  121. & (only_axial['nifti_slice_count'] > 2)
  122. ]
  123. summary_statistics = add_statistics(
  124. summary_statistics,
  125. no_localizer_axial,
  126. 'Without localizer files',
  127. '''Remove files where the Image Type contains "LOCALIZER" and volumes with less than three slices as
  128. indicated in the NIfTI file header.'''
  129. )
  131. # Filter for ROI nifti files
  132. no_roi_no_localizer_axial = no_localizer_axial.loc[no_localizer_axial['nifti_file'].apply(lambda x: 'ROI' not in x)]
  133. summary_statistics = add_statistics(
  134. summary_statistics,
  135. no_roi_no_localizer_axial,
  136. 'Without ROI files',
  137. 'DICOM files can contain binary overlays. dcm2niix creates additional ROI files for these overlays.'
  138. )
  140. # Choose a single nifti if there are multiple ones per directory
  141. unique_version_no_roi_no_localizer_axial = no_roi_no_localizer_axial \
  142. .groupby(['patient_id', 'study_subdir', 'series_subdir']) \
  143. .apply(dataset_folder_structure.choose_nifti_file, include_groups=False) \
  144. .reset_index()
  145. summary_statistics = add_statistics(
  146. summary_statistics,
  147. unique_version_no_roi_no_localizer_axial,
  148. 'Choose unique version from duplicates',
  149. '''dcm2niix creates additional files for gantry tilt correction (Tilt_1), resliced equidistance
  150. (Eq_1) or both (Tilt_Eq_1) in addition to non-corrected files. We choose the corrected files here.'''
  151. )
  153. summary_statistics_df = pd.DataFrame(summary_statistics)
  154. print(summary_statistics_df)
  155. summary_statistics_df.to_csv(cleaned_nifti_info_file_path.parent / '02_summary_statistics.csv')
  157. unique_version_no_roi_no_localizer_axial.to_pickle(filtered_nifti_info_file_path)
  160. def _add_slice_count(df, base_path: Path, dataset_folder_structure: DatasetStructure):
  161. slice_counts = []
  162. nifti_headers = get_nifti_headers(base_path)
  164. with_slice_count = df.copy()
  166. for index, row in with_slice_count.iterrows():
  167. file_path = dataset_folder_structure.create_path_from_csv_row(base_path / 'niftis', row, file_column_name='nifti_file')
  168. slice_count = nifti_headers[str(file_path)].get_data_shape()[2]
  169. slice_counts.append(slice_count)
  171. with_slice_count.loc[:, 'nifti_slice_count'] = slice_counts
  172. return with_slice_count
  175. def check_registration_files(base_path: Path, dataset_folder_structure: DatasetStructure):
  176. filtered_nifti_info_file_path = base_path / NIFTI_SELECTION_FILES[2]
  177. registered_nifti_info_file_path = base_path / NIFTI_SELECTION_FILES[3]
  179. filtered_nifti_info = pd.read_pickle(filtered_nifti_info_file_path)
  180. age_group_info = pd.read_csv(base_path / 'age_group.csv', dtype={'age_group': 'int32', 'patient_id': 'str'})
  182. nifti_files_with_age_groups = filtered_nifti_info.merge(
  183. age_group_info,
  184. on='patient_id',
  185. how='inner',
  186. )
  187. registered_files = []
  189. for index, row in nifti_files_with_age_groups.iterrows():
  190. if index % 100 == 0:
  191. print(f'Processing {index} / {nifti_files_with_age_groups.shape[0]}')
  192. registered_file = Path(
  193. str(
  194. dataset_folder_structure.create_path_from_csv_row(
  195. root_folder_path=base_path / 'registered_niftis',
  196. row=row,
  197. file_column_name='nifti_file'
  198. )
  199. )
  200. )
  201. if registered_file.is_file():
  202. registered_files.append(registered_file.name)
  203. else:
  204. registered_files.append(None)
  205. print(registered_files)
  207. nifti_files_with_age_groups['registered_file'] = registered_files
  209. success_registration_files = nifti_files_with_age_groups.loc[
  210. ~(nifti_files_with_age_groups['registered_file'].isnull())]
  211. success_registration_files.to_pickle(registered_nifti_info_file_path)
  212. write_registered_nifti_files_summary(registered_nifti_info_file_path)
  214. failed_registration_files = nifti_files_with_age_groups.loc[nifti_files_with_age_groups['registered_file'].isnull()]
  215. failed_registration_files.to_csv(
  216. registered_nifti_info_file_path.parent / '03_failed_registration_files.csv',
  217. index=False
  218. )
  221. def write_registered_nifti_files_summary(registered_nifti_info_file_path):
  222. registered_nifti_files_info = pd.read_pickle(registered_nifti_info_file_path)
  223. summary_statistics = {
  224. 'step_name': [],
  225. 'nifti_file_count': [],
  226. 'original_version_count': [],
  227. 'patient_count': [],
  228. 'explanation': []
  229. }
  230. add_statistics(
  231. summary_statistics,
  232. registered_nifti_files_info,
  233. 'Co-registration with FLIRT v. 6.0',
  234. '''Registration seems to fail especially for the Siemens (Manufacturer) SOMATOM PLUS 4
  235. (Manufacturers Model Name) with Software Version VC10C.'''
  236. )
  237. summary_statistics_df = pd.DataFrame(summary_statistics)
  238. print(summary_statistics_df)
  239. summary_statistics_df.to_csv(registered_nifti_info_file_path.parent / '03_summary_statistics.csv', index=False)
  242. def manual_low_similarity_check_results(
  243. base_path: Path,
  244. dataset_structure: DatasetStructure,
  245. ):
  246. registered_nifti_info = pd.read_pickle(base_path / NIFTI_SELECTION_FILES[3])
  248. registration_failures_path = base_path / dataset_structure.REGISTRATION_FAILURES_CSV
  249. if registration_failures_path.exists():
  250. manual_exclusions = pd.read_csv(registration_failures_path)
  251. else:
  252. manual_exclusions = pd.DataFrame(columns=['image_path'])
  254. registered_nifti_info = dataset_structure.add_relative_path_to_dataframe(
  255. df=registered_nifti_info,
  256. relative_path_column_name='registered_file_path',
  257. file_column_name='registered_file'
  258. )
  260. accurate_registrations = registered_nifti_info.loc[~registered_nifti_info['registered_file_path'].isin(manual_exclusions['image_path'])]
  261. inaccurate_registrations = registered_nifti_info.loc[registered_nifti_info['registered_file_path'].isin(manual_exclusions['image_path'])]
  263. accurate_registration_path = base_path / NIFTI_SELECTION_FILES[4]
  264. accurate_registrations.to_pickle(accurate_registration_path)
  266. # List of accurate registrations for superimpose QC.
  267. full_path = accurate_registrations['registered_file_path'].apply(lambda relative_path: str(base_path / 'registered_niftis' / relative_path))
  268. with open(accurate_registration_path.parent / '04_accurate_registration_files.json', 'w') as file:
  269. json.dump(full_path.to_list(), file, indent=4)
  271. inaccurate_registrations.to_csv(
  272. accurate_registration_path.parent / '04_inaccurate_registration_files.csv',
  273. index=False
  274. )
  275. write_ssim_qc_nifti_files_summary(accurate_registration_path)
  278. def write_ssim_qc_nifti_files_summary(accurate_registration_path):
  279. registered_nifti_files_info = pd.read_pickle(accurate_registration_path)
  280. summary_statistics = {
  281. 'step_name': [],
  282. 'nifti_file_count': [],
  283. 'original_version_count': [],
  284. 'patient_count': [],
  285. 'explanation': []
  286. }
  287. add_statistics(
  288. summary_statistics,
  289. registered_nifti_files_info,
  290. 'Manual check of low SSIM score registration files.',
  291. '''The co-registration can produce inaccurate co-registrations even if the FLIRT call was successful.
  292. This step identifies five groups per age group: complete, skull_base, skull_vault, middle, and incomplete to
  293. group similar scans together (done by a calculation of the information available across the x-axis for the
  294. co-registered scans). Afterwards the SSIM score to measure similarity between CT scan and MRI template is
  295. calculated and low similarity score registrations were examined manually.'''
  296. )
  297. summary_statistics_df = pd.DataFrame(summary_statistics)
  298. print(summary_statistics_df)
  299. summary_statistics_df.to_csv(accurate_registration_path.parent / '04_summary_statistics.csv', index=False)
  302. def manual_superimpose_check_results(
  303. base_path: Path,
  304. ):
  305. registered_nifti_info = pd.read_pickle(base_path / NIFTI_SELECTION_FILES[4])
  307. blacklist_path = base_path / 'quality_control/superimpose_information/blacklist.json'
  308. if blacklist_path.exists():
  309. with open(blacklist_path, 'r') as file:
  310. blacklist_information = json.load(file)
  311. manual_exclusions = [entry['file'] for entry in blacklist_information]
  312. else:
  313. manual_exclusions = []
  315. accurate_registrations = registered_nifti_info.loc[~registered_nifti_info['registered_file_path'].isin(manual_exclusions)]
  316. inaccurate_registrations = registered_nifti_info.loc[registered_nifti_info['registered_file_path'].isin(manual_exclusions)]
  318. accurate_registrations_path = base_path / NIFTI_SELECTION_FILES[5]
  319. accurate_registrations.to_pickle(accurate_registrations_path)
  320. inaccurate_registrations.to_csv(
  321. accurate_registrations_path.parent / '05_inaccurate_registration_files.csv',
  322. index=False
  323. )
  325. write_superimpose_qc_nifti_files_summary(base_path)
  328. def write_superimpose_qc_nifti_files_summary(base_path: Path):
  329. registered_nifti_files_info = pd.read_pickle(base_path / NIFTI_SELECTION_FILES[5])
  330. summary_statistics = {
  331. 'step_name': [],
  332. 'nifti_file_count': [],
  333. 'original_version_count': [],
  334. 'patient_count': [],
  335. 'explanation': []
  336. }
  337. add_statistics(
  338. summary_statistics,
  339. registered_nifti_files_info,
  340. 'Manual check superimposed dataset of registration files.',
  341. '''This is an additional quality control step. The registered CT scans were superimposed and
  342. outliers where identified and inspected manually to find inaccurate registrations. Inaccurate registrations
  343. have regions of interest for the arteries that are not at the actual position of the arteries. These were
  344. excluded.'''
  345. )
  346. summary_statistics_df = pd.DataFrame(summary_statistics)
  347. print(summary_statistics_df)
  348. summary_statistics_df.to_csv((base_path / NIFTI_SELECTION_FILES[5]).parent / '05_summary_statistics.csv', index=False)
  351. def select_scans_with_at_least_one_roi(base_path: Path, dataset_structure: DatasetStructure):
  352. registered_nifti_info = pd.read_pickle(base_path / NIFTI_SELECTION_FILES[5])
  353. selected_rois = select_regions_of_interest(registered_nifti_info, base_path, dataset_structure)
  354. selected_rois.to_pickle(base_path / NIFTI_SELECTION_FILES[6])
  356. write_at_least_one_roi_nifti_files_summary(base_path)
  359. def write_at_least_one_roi_nifti_files_summary(base_path: Path):
  360. registered_nifti_files_info = pd.read_pickle(base_path / NIFTI_SELECTION_FILES[6])
  361. summary_statistics = {
  362. 'step_name': [],
  363. 'roi_count': [],
  364. 'nifti_file_count': [],
  365. 'original_version_count': [],
  366. 'patient_count': [],
  367. 'explanation': []
  368. }
  369. add_roi_statistics(
  370. summary_statistics,
  371. registered_nifti_files_info,
  372. 'Check if at least half of one ROI is present in the scans.',
  373. '''Not all scans contain the ROIs we defined in the MRI template. This step filters scans where
  374. there is no ROI present, e.g. skull vault scans, that don't reach low enough. This step filters by calculating
  375. the size of the ROI mask in the native CT space in absolute (mm^2) values compared to the original size of the
  376. ROI. The condition of inclusion for CT scans is at least one ROI with a size of half of the original
  377. template ROI.'''
  378. )
  379. summary_statistics_df = pd.DataFrame(summary_statistics)
  380. print(summary_statistics_df)
  381. summary_statistics_df.to_csv((base_path / NIFTI_SELECTION_FILES[6]).parent / '06_summary_statistics.csv', index=False)

nifti_selection.py at commit db2d7de, no license · at the source

Overview

Authors: Nada Alamoudi1,2, Maria Del C. Valdés Hernández3, Sohan Seth1, Benjamin Jin3, Eleni Sakka3, Carmen Arteaga-Reyes3, Daniela Jaime García3, Yajun Cheng4, Angela C.C. Jochems3, Grant Mair3, Joanna M. Wardlaw3, Miguel O. Bernabeu5
ORCID iDs: Maria Del C. Valdés Hernández, Benjamin Jin, Carmen Arteaga-Reyes, Grant Mair, Joanna M. Wardlaw
  1. Institute for Adaptive and Neural Computation, Data Science and Artificial Intelligence, School of Informatics, University of Edinburgh, Edinburgh, EH89AB, UK
  2. Computer Science Department, The Applied College, Taibah University, Madinah, 46537, Saudi Arabia
  3. Institute for Neuroscience and Cardiovascular Research, Center for Clinical Brain Sciences, University of Edinburgh, Edinburgh, EH16 4SB, UK
  4. Department of Neurology, West China Hospital, Sichuan University, Chengdu, 610041, China
  5. Usher Institute, Center for Medical Informatics, University of Edinburgh, Edinburgh, EH164UX, UK
Institutions: University of Edinburgh (United Kingdom); Taibah University (Saudi Arabia); West China Hospital of Sichuan University (China); Sichuan University (China)
Journal: medRxiv (preprint)
Dates: published online 11 March 2026
Type: Preprint
License: CC BY
Identifiers: DOI 10.64898/2026.03.10.26348006 · OpenAlex W7135093382
Open access: green, a free copy (OpenAlex)
Status: code verified
Categories: structural MRI / diffusion (modality), other (modality), stroke (population), methods / tools (subfield)
Methods: Connectivity, Statistics, Machine learning, Preprocessing, fMRI & imaging
Keywords: white matter hyperintensities, white matter hypoattenuations, Computed Tomography, Paired CT-MRI Data, Deep Learning, nnU-Net
Topic: Acute Ischemic Stroke Management (Epidemiology, Medicine), according to OpenAlex
Citations: not cited yet (Europe PMC); 41 references in the paper

Abstract

Purpose: White matter hyperintensities are a key imaging marker of vascular pathology, defined on brain magnetic resonance imaging (MRI) and typically manifesting on non-contrast computed tomography (CT) as subtle white matter hypoattenuation (WMH). Accurately segmenting WMH in CT scans remains challenging due to their low contrast with the surrounding tissue. This work presents an end-to-end framework for WMH segmentation in CT scans and validates the design choices in each step of the processing pipeline. We leverage a state-of-the-art deep-learning method combined with manually annotated and pseudo-labelled datasets from paired CT-MRI scans from different clinical scanners to deliver reliable outcomes.

Approach: Our framework includes DICOM data curation, sequence selection, and automatic label generation as preparation steps. Preprocessing includes z-score intensity normalisation, skull stripping, CT windowing and two-step CT-MRI registration to accurately transfer MRI-derived labels into the CT space. Further processing involves the use of a 3D nnU-Net initially trained on CT images with aligned MRI-based WMH manually derived (n=91) and fine-tuned with two additional pseudo-labelled datasets (n=191).

Findings: CT-based WMH volumes showed a near-perfect correlation with ground-truth MRI WMH volumes (r = 0.98), with a systematic overestimation (mean difference = 2.40 mL; 95% limits of agreement: -8.31 to 13.11 mL) that may be adjustable in downstream tasks. This overestimation reflected challenges in the precise delineation of small WMH lesions and confounding from other imaging markers of brain disease. Across the evaluated cohort, ground-truth WMH volumes ranged from 1.02 to 149.34 mL. The best-performing configuration achieved a mean absolute error below 3 mL, corresponding to approximately 17% of the mean WMH volume, and a mean Dice similarity coefficient of 0.57. Segmentation accuracy decreased in the presence of stroke lesions. Models trained on single-pathology datasets, as well as approaches relying on template-based spatial normalisation, did not achieve satisfactory performance despite using the same backbone network configuration.

Conclusion: Using a multi-centre dataset and a multi-modal approach with expert-annotated data combined with pseudo-labelled data for training can substantially narrow the performance gap between CT- and MRI-based WMH segmentation. The framework proposed provides a generalisable solution that underscores the practical viability of CT for evaluating WMH burden in clinical and research scenarios—particularly where MRI is unavailable or contraindicated—thereby broadening access to small-vessel disease assessment.

Reproduced under the paper's license (CC BY), from the paper cited above.

Repository

Its files are read in the Code ↔ Paper reader above, with 2 matches between paragraphs and lines of code.

bjin96/ct-processing

License: none: the authors keep all their rights
State: the link answers, verified on 30 September 2026
Evidence: files inventoried
Commit: db2d7de50a9b480effff9abd857dd64a85383657, 6 July 2026
Languages: Python (23), Shell (1)
Size: 28 files, 24 scripts
Software Heritage: not archived
Found in: the text, “Data Preparation”
Holds: README, environment (requirements.txt), documentation
Not found: license file, CITATION.cff, tests, continuous integration
Tools: pandas (10 files), NiBabel (7 files), Nilearn (6 files), NumPy (6 files), Matplotlib (1 file), pydicom (1 file), scikit-image (1 file), SciPy (1 file)
Availability: 1 check, the latest on 30 September 2026: the link answers
  • 30 September 2026: the link answers
25 files
At the source: github.com/bjin96/ct-processing

The paper's code and data availability statement is in the Data section.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

  • 1 repository of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
  • 24 scripts, each with its path and the digest of its content;
  • 2 matches between paragraphs of the paper and lines of the code (method lexical-v1);
  • neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Data and Code Availability

The framework presented uses software modules and libraries publicly available. The backbone neural network architecture used, namely nnUNet, is also freely available. The imaging data could be available upon request to the principal investigators of the primary studies. The neural network weights of the proposed model can also be made available upon request.

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 30 September 2026: the first record

Recorded: type, journal, dates, 12 authors, 6 keywords, 33 references.

Cite

This paper

Alamoudi, N., Valdés Hernández, M. D. C., Seth, S., Jin, B., Sakka, E., Arteaga-Reyes, C., García, D. J., Cheng, Y., Jochems, A. C., Mair, G., Wardlaw, J. M., & Bernabeu, M. O. (2026). AI-Based Pipeline for the Segmentation of White Matter Hypoattenuations in CT Scans: A Design-Choice Validation Study. medRxiv (preprint). https://doi.org/10.64898/2026.03.10.26348006

BibTeX

@article{alamoudi2026ai,
author = {Alamoudi, Nada and Valdés Hernández, Maria Del C. and Seth, Sohan and Jin, Benjamin and Sakka, Eleni and Arteaga-Reyes, Carmen and García, Daniela Jaime and Cheng, Yajun and Jochems, Angela C.C. and Mair, Grant and Wardlaw, Joanna M. and Bernabeu, Miguel O.},
title = {{AI-Based Pipeline for the Segmentation of White Matter Hypoattenuations in CT Scans: A Design-Choice Validation Study}},
journal = {medRxiv (preprint)},
year = {2026},
month = mar,
publisher = {medRxiv},
doi = {10.64898/2026.03.10.26348006},
url = {https://doi.org/10.64898/2026.03.10.26348006}
}

RIS

TY - JOUR
AU - Alamoudi, Nada
AU - Valdés Hernández, Maria Del C.
AU - Seth, Sohan
AU - Jin, Benjamin
AU - Sakka, Eleni
AU - Arteaga-Reyes, Carmen
AU - García, Daniela Jaime
AU - Cheng, Yajun
AU - Jochems, Angela C.C.
AU - Mair, Grant
AU - Wardlaw, Joanna M.
AU - Bernabeu, Miguel O.
TI - AI-Based Pipeline for the Segmentation of White Matter Hypoattenuations in CT Scans: A Design-Choice Validation Study
T2 - medRxiv (preprint)
J2 - medRxiv
PY - 2026
DA - 2026/03/11
PB - medRxiv
DO - 10.64898/2026.03.10.26348006
UR - https://doi.org/10.64898/2026.03.10.26348006
ER -

CSL-JSON

{
"id": "10.64898/2026.03.10.26348006",
"type": "article",
"title": "AI-Based Pipeline for the Segmentation of White Matter Hypoattenuations in CT Scans: A Design-Choice Validation Study",
"container-title": "medRxiv (preprint)",
"author": [
{
"family": "Alamoudi",
"given": "Nada"
},
{
"family": "Valdés Hernández",
"given": "Maria Del C."
},
{
"family": "Seth",
"given": "Sohan"
},
{
"family": "Jin",
"given": "Benjamin"
},
{
"family": "Sakka",
"given": "Eleni"
},
{
"family": "Arteaga-Reyes",
"given": "Carmen"
},
{
"family": "García",
"given": "Daniela Jaime"
},
{
"family": "Cheng",
"given": "Yajun"
},
{
"family": "Jochems",
"given": "Angela C.C."
},
{
"family": "Mair",
"given": "Grant"
},
{
"family": "Wardlaw",
"given": "Joanna M."
},
{
"family": "Bernabeu",
"given": "Miguel O."
}
],
"container-title-short": "medRxiv",
"DOI": "10.64898/2026.03.10.26348006",
"publisher": "medRxiv",
"URL": "https://doi.org/10.64898/2026.03.10.26348006",
"issued": {
"date-parts": [
[
2026,
3,
11
]
]
}
}

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.1212/wnl.0000000000218502
Retinal Optical Coherence Tomography-Angiography and Longitudinal Changes in Cognition and Cerebral Small Vessel Disease.
Journal: Neurology
In common: stroke, other, structural MRI / diffusion, 3 references, 2 authors
[2] doi:10.1038/s41597-026-07184-5 [code]
A Multiple Sclerosis MRI Dataset with Tri-Mask Annotations for Lesion Segmentation.
Journal: Scientific data
In common: pydicom, scikit-image, NiBabel, 4 other tools, methods / tools, structural MRI / diffusion, 2 references
[3] doi:10.64898/2026.03.09.26347914 [code]
Multimodal Ageing Biomarkers and Plasma Proteomic Signatures Associated with All-Cause Mortality
Journal: medRxiv (preprint)
In common: structural MRI / diffusion, 2 authors
[4] doi:10.1186/s12938-026-01555-0 [code]
Incorporating normal periventricular changes for enhanced pathological white matter hyperintensity segmentation: on multiclass deep learning approaches.
Journal: Biomedical engineering online
In common: scikit-image, pandas, SciPy, 2 other tools, stroke, methods / tools, structural MRI / diffusion, 3 references
[5] doi:10.3389/frai.2026.1771088 [code]
Few-shot deployment of pretrained MRI transformers in brain imaging tasks.
Journal: Frontiers in artificial intelligence
In common: pydicom, scikit-image, NiBabel, 4 other tools, methods / tools, structural MRI / diffusion, 1 reference
[6] doi:10.1080/07853890.2026.2685416 [code]
Pulmonary and cerebral damage in COVID-19 survivors: is there any association?
Journal: Annals of medicine
In common: pydicom, scikit-image, NiBabel, 4 other tools, stroke, other, structural MRI / diffusion
[7] doi:10.1162/imag.a.1262 [code]
Frame-wise multi-echo distortion correction for superior functional MRI.
Journal: Imaging neuroscience (Cambridge, Mass.)
In common: pydicom, Nilearn, scikit-image, 5 other tools
[8] doi:10.1371/journal.pcbi.1014555 [code]
Body surface potential driven personalisation of electrophysiological digital twins in hypertrophic cardiomyopathy.
Journal: PLoS computational biology
In common: pydicom, scikit-image, NiBabel, 4 other tools, other, structural MRI / diffusion
[9] doi:10.1016/j.nicl.2026.104001 [code]
Effect of vascular lesion preprocessing on Brain Intensity AbNormality Classification Algorithm (BIANCA) white matter hyperintensity segmentation.
Journal: NeuroImage. Clinical
In common: NiBabel, pandas, SciPy, 2 other tools, stroke, methods / tools, structural MRI / diffusion, 2 references
[10] doi:10.1212/wnl.0000000000218472 [code]
Lesion-Level Subtypes of White Matter Hyperintensity Evolution Beyond Spatial Location.
Journal: Neurology
In common: Nilearn, NiBabel, pandas, 3 other tools, stroke, structural MRI / diffusion, 1 reference

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.