A pre-trained foundation model framework for multiplanar MRI classification of extramural vascular invasion and mesorectal fascia invasion in rectal cancer.
4 blocks of code here have a permanent identifier (oscr:b:…), listed with the matches below: citable for good, with a page of its own. Select a match and press Cite to copy it. What a block identifier is.
The 12 matches
- [1] § Methods › Evaluation › Evaluation metrics ↔ notebooks/compute_AUC_CI.ipynb, lines 513 to 584 · score 0.83 · confidence intervals, binary predictions, predicted probabilities, balanced accuracy, F1 score, CI · Report a mismatch
- [2] § Methods › Model development › Training strategy ↔ othercode/classification_model/FMmodel.py, lines 231 to 269 · score 0.76 · Cosine annealing, AdamW, decay, UMedPt, scheduling, optimized · block
oscr:b:4DMKBT36HBXEY-M· Report a mismatch - [3] § Methods › Data curation › Dataset ↔ notebooks/export_patient_information.ipynb, lines 87 to 153 · score 0.76 · La Fe, CHU Angers, Centre, Siemens, ULS, manufacturers · Report a mismatch
- [4] § Methods › Model development › Training strategy ↔ othercode/classification_model/segresnet192.py, lines 185 to 229 · score 0.74 · focal loss, Cosine annealing, SGD, decay, scheduling, batch · block
oscr:b:GKPT0FS2D3VZZ-0· Report a mismatch - [5] § Results › Classification results ↔ notebooks/draw_ROC.ipynb, lines 145 to 274 · score 0.74 · ROC curves, SeResnet, axial original, UMedPT_LR, MFI classification, AUC · Report a mismatch
- [6] § Results › Classification results ↔ notebooks/draw_ROC.ipynb, lines 145 to 274 · score 0.64 · axial_harmo, ROC curves, UMedPT_LR, AUC, sagittal, EVI · Report a mismatch
- [7] § Results › Patient characteristics ↔ notebooks/export_patient_information.ipynb, lines 87 to 153 · score 0.61 · CHU Angers, train_val, Siemens, ULS, manufacturers, patient · Report a mismatch
- [8] § Methods › Data curation › Image preprocessing and rectum localization ↔ notebooks/resample_totalmask3.ipynb, lines 83 to 172 · score 0.61 · NIfTI, colon mask, resampled, spacing, TotalSegmentator, rectum · Report a mismatch
- [9] § Results › Classification results ↔ notebooks/draw_ROC.ipynb, lines 1 to 139 · score 0.55 · UMedPT_LR, EVI classification, ROC, curves, AUC, score · Report a mismatch
- [10] § Methods › Data curation › Image preprocessing and rectum localization ↔ othercode/classification_model/centrecrop_correct2.ipynb, lines 26 to 111 · score 0.54 · center crop, voxel, intensity, spacing, TotalSegmentator, mask · Report a mismatch
- [11] § Methods › Model development › Training strategy ↔ othercode/classification_model/FMmodel.py, lines 289 to 345 · score 0.54 · Gaussian noise, affine, intensity, transforms, patch, Training · block
oscr:b:CW10P6FPZGHF1-L· Report a mismatch - [12] § Methods › Model development › Training strategy ↔ othercode/classification_model/embedding_LR.py, lines 277 to 333 · score 0.54 · Gaussian noise, affine, intensity, transforms, patch, Training · block
oscr:b:PW99ZHBJBFPH7-9· Report a mismatch
Paper
Loaded from Europe PMC by your browser; OSCR does not store it: doi.org · Europe PMC
The paper is loaded when this pane is shown.
The authors' code
Jupyter notebook · 278 lines · 9.9 KB · Apache-2.0 · 3 matches
- # %%
- import matplotlib.pyplot as plt
- from sklearn.metrics import roc_curve, auc, roc_auc_score
- from sklearn.utils import resample
- import numpy as np
- import matplotlib as mpl
- import pandas as pd
- #----for evi -----
- df1 = pd.read_csv(r"/path/to/workspace/Complete_set/Experiments/Results/UMedPT_LR/combine_origin.csv")
- df2 = pd.read_csv(r"/path/to/workspace/Complete_set/Experiments/Results/UMedPT_LR/axial_original.csv")
- df3 = pd.read_csv(r"/path/to/workspace/Complete_set/Experiments/Results/UMedPT_LR/sagittal_original.csv")
- df4 = pd.read_csv(r"/path/to/workspace/Complete_set/Experiments/Results/FMmodel192/test/best_FM_192.csv")
- # Extract ground truth and predicted probabilities for EVI and MFI
- y_true = df1['y_test_evi'].values
- y_pred_model1 = df1['y_pred_prob_evi'].values
- y_pred_model2 = df2['y_pred_prob_evi'].values
- y_pred_model3 = df3['y_pred_prob_evi'].values
- y_pred_model4 = df4['predicted_probability_evi'].values
- # Model names and placeholder for AUC formatting
- models = {
- 'UMedPT_LR axial+sagittal': y_pred_model1,
- 'UMedPT_LR axial': y_pred_model2,
- 'UMedPT_LR sagittal': y_pred_model3,
- 'UMedPT axial': y_pred_model4,
- }
- # Soft, elegant colors: light blue, light yellow, light green, light red
- # colors = ['#d62728','#1f77b4', '#ffbb00', '#2ca02c'] # soft red blue, yellow, green
- # colors = ['#e41a1c', '#377eb8', '#4daf4a', '#984ea3']
- # colors = ['#FBDD85', '#80A6E2', '#F46F43', '#403990']
- # colors = ['#403990', '#F46F43','#80A6E2','#FBDD85']
- colors = ['#403990', '#F46F43','#80A6E2','#FBDD85']
- # Set plot style to match high-quality publication standards (e.g., Nature)
- mpl.rcParams['font.family'] = 'sans-serif'
- mpl.rcParams['font.size'] = 10
- #bootstrap compute CI
- def bootstrap_roc_ci(y_true, y_pred, n_bootstraps=1000, alpha=0.95):
- rng = np.random.RandomState(42)
- tprs = []
- base_fpr = np.linspace(0, 1, 101)
- for _ in range(n_bootstraps):
- indices = rng.randint(0, len(y_pred), len(y_pred))
- if len(np.unique(y_true[indices])) < 2:
- continue
- fpr, tpr, _ = roc_curve(y_true[indices], y_pred[indices])
- tpr_interp = np.interp(base_fpr, fpr, tpr)
- tpr_interp[0] = 0.0
- tprs.append(tpr_interp)
- tprs = np.array(tprs)
- mean_tpr = tprs.mean(axis=0)
- std_tpr = tprs.std(axis=0)
- tpr_upper = np.minimum(mean_tpr + 1.96 * std_tpr, 1)
- tpr_lower = np.maximum(mean_tpr - 1.96 * std_tpr, 0)
- return base_fpr, mean_tpr, tpr_lower, tpr_upper
- def bootstrap_auc_ci(y_true, y_pred, n_bootstraps=1000, seed=42, alpha=0.95):
- rng = np.random.RandomState(seed)
- stats = []
- n = len(y_pred)
- y_true = np.asarray(y_true)
- y_pred = np.asarray(y_pred)
- for _ in range(n_bootstraps):
- idx = rng.randint(0, n, n)
- if len(np.unique(y_true[idx])) < 2:
- continue
- stats.append(roc_auc_score(y_true[idx], y_pred[idx]))
- lower = np.percentile(stats, (1 - alpha) / 2 * 100)
- upper = np.percentile(stats, (1 + alpha) / 2 * 100)
- return lower, upper
- # Create the figure
- plt.figure(figsize=(6, 6))
- # Plot ROC curves for each model
- # for (label, y_pred), color in zip(models.items(), colors):
- # fpr, mean_tpr, lower_tpr, upper_tpr = bootstrap_roc_ci(y_true, y_pred)
- # auc_score = auc(fpr, mean_tpr)
- # plt.plot(fpr, mean_tpr, lw=2, label=f"{label} (AUC: {auc_score:.2f})", color=color)
- # # plt.plot(fpr, lower_tpr, color=color, linestyle='--', alpha=0.6, linewidth=1)
- # # plt.plot(fpr, upper_tpr, color=color, linestyle='--', alpha=0.6, linewidth=1)
- # plt.fill_between(fpr, lower_tpr, upper_tpr, color=color, alpha=0.12)
- shade_first_n = 2
- for i, ((model_name, y_pred), color) in enumerate(zip(models.items(), colors)):
- fpr_raw, tpr_raw, _ = roc_curve(y_true, y_pred)
- auc_raw = roc_auc_score(y_true, y_pred)
- # --- AUC bootstrap CI ?? ---
- auc_lo, auc_hi = bootstrap_auc_ci(y_true, y_pred)
- # Legend ??????? or en-dash?
- label_text = f"{model_name} (AUC: {auc_raw:.2f} [{auc_lo:.2f}-{auc_hi:.2f}])"
- # ?????? or ?? plt.plot ???????
- plt.step(fpr_raw, tpr_raw, where='post', lw=2.2, color=color, label=label_text)
- # --- bootstrap CI for shading / dashed lines ---
- # fpr_ci, mean_tpr, lower_tpr, upper_tpr = bootstrap_roc_ci(y_true, y_pred)
- # if i < shade_first_n:
- # plt.fill_between(fpr, lower_tpr, upper_tpr, color=color, alpha=0.1, linewidth=0)
- # else:
- # plt.plot(fpr, lower_tpr, color=color, linestyle='--', linewidth=1.1, alpha=0.7)
- # plt.plot(fpr, upper_tpr, color=color, linestyle='--', linewidth=1.1, alpha=0.7)
- # Plot diagonal line representing random guess
- plt.plot([0, 1], [0, 1], 'k--', lw=1, label='Random Guess')
- # Set titles and labels
- plt.title('Best 4 Models for EVI Classification', fontsize=12, weight='normal')
- plt.xlabel('False Positive Rate', fontsize=12)
- plt.ylabel('True Positive Rate', fontsize=12)
- plt.legend(loc='lower right', frameon=False, fontsize=9)
- plt.grid(True, linestyle='--', alpha=0.3)
- plt.xticks(np.linspace(0, 1, 6))
- plt.yticks(np.linspace(0, 1, 6))
- # Thicker border lines
- for spine in plt.gca().spines.values():
- spine.set_linewidth(1.2)
- output_dir = "/path/to/workspace/Complete_set/Experiments/Results/ROC PHOTO/"
- filename = "roc_curve_evi_new3.png"
- # Optimize layout
- plt.tight_layout()
- plt.savefig(output_dir + filename, dpi=600, bbox_inches='tight')
- plt.show()
- # %%
- # %%
- import matplotlib.pyplot as plt
- from sklearn.metrics import roc_curve, auc
- import numpy as np
- import matplotlib as mpl
- import pandas as pd
- #----for MFI-----
- df1 = pd.read_csv(r"/path/to/workspace/Complete_set/Experiments/Results/axial_harmo_FM192/best_FM_axial_harmo.csv")
- df2 = pd.read_csv(r"/path/to/workspace/Complete_set/Experiments/Results/UMedPT_LR/axial_original.csv")
- df3 = pd.read_csv(r"/path/to/workspace/Complete_set/Experiments/Results/sagittal_harmo/best_scr_segresnet_sagittalharmo.csv")
- df4 = pd.read_csv(r"/path/to/workspace/Complete_set/Experiments/Results/Segresnet/best_segresnet_axial_original.csv")
- # Extract ground truth and predicted probabilities for EVI and MFI
- y_true = df1['ground_truth_mfi'].values
- y_pred_model1 = df1['predicted_probability_mfi'].values
- y_pred_model2 = df2['y_pred_prob_mfi'].values
- y_pred_model3 = df3['predicted_probability_mfi'].values
- y_pred_model4 = df4['predicted_probability_mfi'].values
- # Model names and placeholder for AUC formatting
- models = {
- 'UMedPT axial_harmo': y_pred_model1,
- 'UMedPT_LR axial': y_pred_model2,
- 'SeResnet axial': y_pred_model4,
- 'SeResnet sagittal_harmo': y_pred_model3
- }
- # Soft, elegant colors: light blue, light yellow, light green, light red
- # colors = ['#d62728','#1f77b4', '#ffbb00', '#2ca02c'] # soft red blue, yellow, green
- colors = ['#403990', '#F46F43','#80A6E2','#FBDD85']
- # Set plot style to match high-quality publication standards (e.g., Nature)
- mpl.rcParams['font.family'] = 'sans-serif'
- mpl.rcParams['font.size'] = 10
- def bootstrap_roc_ci(y_true, y_pred, n_bootstraps=1000, alpha=0.95):
- rng = np.random.RandomState(42)
- tprs = []
- base_fpr = np.linspace(0, 1, 101)
- for _ in range(n_bootstraps):
- indices = rng.randint(0, len(y_pred), len(y_pred))
- if len(np.unique(y_true[indices])) < 2:
- continue
- fpr, tpr, _ = roc_curve(y_true[indices], y_pred[indices])
- tpr_interp = np.interp(base_fpr, fpr, tpr)
- tpr_interp[0] = 0.0
- tprs.append(tpr_interp)
- tprs = np.array(tprs)
- mean_tpr = tprs.mean(axis=0)
- std_tpr = tprs.std(axis=0)
- tpr_upper = np.minimum(mean_tpr + 1.96 * std_tpr, 1)
- tpr_lower = np.maximum(mean_tpr - 1.96 * std_tpr, 0)
- return base_fpr, mean_tpr, tpr_lower, tpr_upper
- def bootstrap_auc_ci(y_true, y_pred, n_bootstraps=1000, seed=42, alpha=0.95):
- rng = np.random.RandomState(seed)
- stats = []
- n = len(y_pred)
- y_true = np.asarray(y_true)
- y_pred = np.asarray(y_pred)
- for _ in range(n_bootstraps):
- idx = rng.randint(0, n, n)
- if len(np.unique(y_true[idx])) < 2:
- continue
- stats.append(roc_auc_score(y_true[idx], y_pred[idx]))
- lower = np.percentile(stats, (1 - alpha) / 2 * 100)
- upper = np.percentile(stats, (1 + alpha) / 2 * 100)
- return lower, upper
- # Create the figure
- plt.figure(figsize=(6, 6))
- shade_first_n = 2
- for i, ((model_name, y_pred), color) in enumerate(zip(models.items(), colors)):
- fpr_raw, tpr_raw, _ = roc_curve(y_true, y_pred)
- auc_raw = roc_auc_score(y_true, y_pred)
- # --- AUC bootstrap CI ?? ---
- auc_lo, auc_hi = bootstrap_auc_ci(y_true, y_pred)
- # Legend ??????? or en-dash?
- label_text = f"{model_name} (AUC: {auc_raw:.2f} [{auc_lo:.2f}-{auc_hi:.2f}])"
- # ?????? or ?? plt.plot ???????
- plt.step(fpr_raw, tpr_raw, where='post', lw=2.2, color=color, label=label_text)
- # # --- bootstrap CI for shading / dashed lines ---
- # fpr_ci, mean_tpr, lower_tpr, upper_tpr = bootstrap_roc_ci(y_true, y_pred)
- # if i < shade_first_n:
- # plt.fill_between(fpr, lower_tpr, upper_tpr, color=color, alpha=0.1, linewidth=0)
- # else:
- # plt.plot(fpr, lower_tpr, color=color, linestyle='--', linewidth=1.1, alpha=0.7)
- # plt.plot(fpr, upper_tpr, color=color, linestyle='--', linewidth=1.1, alpha=0.7)
- # Plot diagonal line representing random guess
- plt.plot([0, 1], [0, 1], 'k--', lw=1, label='Random Guess')
- # Set titles and labels
- plt.title('Best 4 Models for MFI Classification', fontsize=12, weight='normal')
- plt.xlabel('False Positive Rate', fontsize=12)
- plt.ylabel('True Positive Rate', fontsize=12)
- # Add legend in bottom right, no frame
- plt.legend(loc='lower right', frameon=False, fontsize=9)
- # Add light grid lines
- plt.grid(True, linestyle='--', alpha=0.3)
- # Customize ticks
- plt.xticks(np.linspace(0, 1, 6))
- plt.yticks(np.linspace(0, 1, 6))
- # Thicker border lines
- for spine in plt.gca().spines.values():
- spine.set_linewidth(1.2)
- output_dir = "/path/to/workspace/Complete_set/Experiments/Results/ROC PHOTO/"
- filename = "roc_curve_mfi_new3.png"
- # Optimize layout
- plt.tight_layout()
- plt.savefig(output_dir + filename, dpi=300, bbox_inches='tight')
- plt.show()
- # %%
draw_ROC.ipynb at commit fa683ec, under Apache-2.0 · at the source
Overview
- The D-Lab, Department of Precision Medicine, GROW - Research Institute for Oncology and Reproduction, Maastricht University,Maastricht, The Netherlands
- Research & Frontiers in AI Department, Quantitative Imaging Biomarkers in Medicine, Quibim SL, Valencia, Spain
- Biomedical Imaging Research Group, La Fe Health Research Institute, Valencia, Spain
- Radiology Unit, Department of Surgical and Medical Sciences and Translational Medicine, Sapienza University of Rome, Sant’Andrea Hospital,Rome, Italy
- Laboratoire HIFIH, Université d’Angers, SFR ICAT 4208,Angers, France
- Department of Radiology, CHU Angers,Angers, France
- Medical Imaging Department, La Fe University and Polytechnic Hospital,Valencia, Spain
- Department of Radiology and Nuclear Medicine, GROW - Research Institute for Oncology and Reproduction, Maastricht University Medical Center+,Maastricht, The Netherlands
Abstract
Objectives: Accurate MRI-based identification of extramural vascular invasion (EVI) and mesorectal fascia invasion (MFI) is crucial for risk-stratified rectal cancer treatment. However, subjective visual assessment and inter-institutional variability limit diagnostic consistency. This study developed and evaluated a multi-center, foundation model-driven framework that automatically classifies EVI and MFI on axial and sagittal MRI.
Materials and methods: A total of 331 pre-treatment rectal cancer T2-weighted MRI scans from three European hospitals were retrospectively recruited. A self-supervised frequency domain harmonization strategy was applied to reduce scanner variability. Three classifiers, SeResNet, the universal biomedical pretrained model (UMedPT) with a multilayer perceptron head, and a logistic-regression variant using frozen UMedPT features (UMedPT_LR), were trained (n = 265) and tested (n = 66). Gradient-weighted class activation mapping (Grad-CAM) visualized model predictions.
Results: UMedPT_LR achieved the best EVI performance with multiplanar fusion (AUC = 0.82, test set). For MFI, UMedPT trained on axial harmonized images yielded the highest performance (AUC = 0.77). Both tasks outperformed the CHAIMELEON 2024 benchmark (EVI: 0.82 vs 0.74; MFI: 0.77 vs 0.75). Harmonization enhanced MFI classification, and multiplanar fusion further boosted EVI performance. Grad-CAM confirmed biologically plausible attention on peritumoral regions (EVI) and mesorectal fascia margins (MFI).
Conclusion: The proposed foundation model-driven framework, leveraging frequency domain harmonization and multiplanar fusion, achieves state-of-the-art performance for automated EVI and MFI classification on MRI, demonstrating strong generalizability across multiple centers.
Critical relevance statement: Addressing inter-center inconsistencies in rectal cancer MRI, a multiplanar foundation model with cross-scanner harmonization significantly improves the detection of EVI and MFI, potentially standardizing staging and guiding therapy.
Key Points: Among the first studies to investigate automated classification of both EVI and MFI using axial and sagittal T2-weighted MRI. Foundation model-derived features outperform conventional convolutional neural networks (CNNs) for EVI and MFI classification. Frequency domain harmonization and multiplanar fusion selectively enhance diagnostic performance. Automated prediction of EVI and MFI may support more consistent staging and clinical decision-making across institutions.
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 12 matches between paragraphs and lines of code.
yumengzhang97/foundation-model-rectal-mri
fa683ec2c4c201aca83cb7070eda9863bbf589d8, 6 July 2026Availability: 1 check, the latest on 28 September 2026: the link answers
- 28 September 2026: the link answers
59 files
- notebooks/
Gender_find.ipynb , Jupyter, 82 lines - notebooks/
Install nnunet.ipynb , Jupyter, 27 lines - notebooks/
Label_count.ipynb , Jupyter, 83 lines - notebooks/
calculate_input_size.ipy , Jupyter, 67 linesnb - notebooks/
calculating_patch_size.i , Jupyter, 157 linespynb - notebooks/
calibration_metrics.ipyn , Jupyter, 213 linesb - notebooks/
centrecrop_axialharmo.ip , Jupyter, 107 linesynb - notebooks/
centrecrop_sagittalharmo , Jupyter, 123 lines.ipynb - notebooks/
clinical_model_new.ipynb , Jupyter, 198 lines - notebooks/
compute_AUC_CI.ipynb , Jupyter, 1,244 lines, 1 match - notebooks/
data_analyze.ipynb , Jupyter, 89 lines - notebooks/
data_analyze2.ipynb , Jupyter, 92 lines - notebooks/
data_resample_centrecrop , Jupyter, 51 lines.ipynb - notebooks/
datamove.ipynb , Jupyter, 66 lines - notebooks/
datamove_classificationm , Jupyter, 62 linesodel.ipynb - notebooks/
draw_ROC.ipynb , Jupyter, 278 lines, 3 matches - notebooks/
error_male_to_female2.ip , Jupyter, 129 linesynb - notebooks/
export_patient_informati , Jupyter, 343 lines, 2 matcheson.ipynb - notebooks/
export_patient_informati , Jupyter, 447 lineson2.ipynb - notebooks/
extract_patch2.ipynb , Jupyter, 141 lines - notebooks/
extract_patch_totalmask2 , Jupyter, 119 lines.ipynb - notebooks/
extract_patches_sagittal , Jupyter, 97 lines2.ipynb - notebooks/
extract_patches_sagittal , Jupyter, 98 lines3.ipynb - notebooks/
find_orientation.ipynb , Jupyter, 86 lines - notebooks/
find_transversal.ipynb , Jupyter, 68 lines - notebooks/
get info.ipynb , Jupyter, 88 lines - notebooks/
label_csv.ipynb , Jupyter, 47 lines - notebooks/
merge_csv_ensemble.ipynb , Jupyter, 294 lines - notebooks/
move_harmonizeimage_csv. , Jupyter, 75 linesipynb - notebooks/
move_train_val_split.ipy , Jupyter, 55 linesnb - notebooks/
nnunet_segmentation.ipyn , Jupyter, 79 linesb - notebooks/
read_dcm.ipynb , Jupyter, 78 lines - notebooks/
rectum_isolate.ipynb , Jupyter, 96 lines - notebooks/
rectum_patches_sagittal. , Jupyter, 160 linesipynb - notebooks/
rename.ipynb , Jupyter, 57 lines - notebooks/
resample_batch_gender.ip , Jupyter, 88 linesynb - notebooks/
resample_batch_gender_pa , Jupyter, 82 linestch.ipynb - notebooks/
resample_totalmask3.ipyn , Jupyter, 175 lines, 1 matchb - notebooks/
sagittal_transpose_coord , Jupyter, 29 linesinate.ipynb - notebooks/
select_axialviews.ipynb , Jupyter, 45 lines - notebooks/
select_sagital.ipynb , Jupyter, 48 lines - notebooks/
select_sagittalviews.ipy , Jupyter, 61 linesnb - notebooks/
split_train_val.ipynb , Jupyter, 17 lines - notebooks/
transfer_dicom_to_nifti. , Jupyter, 78 linesipynb - othercode/
classification_model/ , Python, 762 lines, 2 matchesFMmodel.py - othercode/
classification_model/ , Python, 136 linesLRtrain_S.py - othercode/
classification_model/ , Jupyter, 54 linescentrecrop.ipynb - othercode/
classification_model/ , Jupyter, 117 lines, 1 matchcentrecrop_correct2.ipyn b - othercode/
classification_model/ , Python, 783 lines, 1 matchembedding_LR.py - othercode/
classification_model/ , Python, 800 linesembedding_LR_S.py - othercode/
classification_model/ , Python, 75 linesnormalized_image2.py - othercode/
classification_model/ , Python, 90 linesresample.py - othercode/
classification_model/ , Python, 632 lines, 1 matchsegresnet192.py - othercode/
classification_model/ , Python, 39 linestotal_work.py - othercode/
gradcam/ , Python, 305 linesGRADCAM.py - othercode/
gradcam/ , Python, 757 linesexplainability.py - othercode/
gradcam/ , Python, 308 linesgradcam_new.py - LICENSE, License, 201 lines
- README.md, Text, 102 lines
Code availability
The code will be made publicly available at https://
Reproduced under the paper's license (CC BY), from the paper cited above.
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;
- 57 scripts, each with its path and the digest of its content;
- 12 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 availability
The multicentre MRI datasets used in this study are subject to institutional and EU data protection regulations and therefore cannot be publicly shared. An earlier version of this manuscript was made available as a preprint on arXiv (10.48550/
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, 28 September 2026: the first record
Recorded: type, language, journal, volume, issue, pages, dates, 13 authors, 5 keywords, 13 funders, 40 references.
Cite
This paper
Zhang, Y., Mali, S. A., Khan, D., Amirrajab, S., Ibor-Crespo, E., Jimenez-Pastor, A., Ribas, G., Flor-Arnal, S., Zerunian, M., Aubé, C., Martí-Bonmatí, L., Salahuddin, Z., & Lambin, P. (2026). A pre-trained foundation model framework for multiplanar MRI classification of extramural vascular invasion and mesorectal fascia invasion in rectal cancer. Insights into imaging, 17(1), 141. https://
BibTeX
@article{zhang2026pre,
author = {Zhang, Yumeng and Mali, Shruti Atul and Khan, Danial and Amirrajab, Sina and Ibor-Crespo, Eduardo and Jimenez-Pastor, Ana and Ribas, Gloria and Flor-Arnal, Silvia and Zerunian, Marta and Aubé, Christophe and Martí-Bonmatí, Luis and Salahuddin, Zohaib and Lambin, Philippe},
title = {{A pre-trained foundation model framework for multiplanar MRI classification of extramural vascular invasion and mesorectal fascia invasion in rectal cancer}},
journal = {Insights into imaging},
year = {2026},
month = may,
volume = {17},
number = {1},
pages = {141},
publisher = {Springer},
issn = {1869-4101},
doi = {10.1186/
url = {https://
pmid = {42171860},
pmcid = {PMC13197564}
}
RIS
TY - JOUR
AU - Zhang, Yumeng
AU - Mali, Shruti Atul
AU - Khan, Danial
AU - Amirrajab, Sina
AU - Ibor-Crespo, Eduardo
AU - Jimenez-Pastor, Ana
AU - Ribas, Gloria
AU - Flor-Arnal, Silvia
AU - Zerunian, Marta
AU - Aubé, Christophe
AU - Martí-Bonmatí, Luis
AU - Salahuddin, Zohaib
AU - Lambin, Philippe
TI - A pre-trained foundation model framework for multiplanar MRI classification of extramural vascular invasion and mesorectal fascia invasion in rectal cancer
T2 - Insights into imaging
J2 - Insights Imaging
PY - 2026
DA - 2026/
VL - 17
IS - 1
SP - 141
SN - 1869-4101
PB - Springer
DO - 10.1186/
UR - https://
LA - en
ER -
CSL-JSON
{
"id": "10.1186/
"type": "article-journal",
"title": "A pre-trained foundation model framework for multiplanar MRI classification of extramural vascular invasion and mesorectal fascia invasion in rectal cancer",
"container-title": "Insights into imaging",
"author": [
{
"family": "Zhang",
"given": "Yumeng"
},
{
"family": "Mali",
"given": "Shruti Atul"
},
{
"family": "Khan",
"given": "Danial"
},
{
"family": "Amirrajab",
"given": "Sina"
},
{
"family": "Ibor-Crespo",
"given": "Eduardo"
},
{
"family": "Jimenez-Pastor",
"given": "Ana"
},
{
"family": "Ribas",
"given": "Gloria"
},
{
"family": "Flor-Arnal",
"given": "Silvia"
},
{
"family": "Zerunian",
"given": "Marta"
},
{
"family": "Aubé",
"given": "Christophe"
},
{
"family": "Martí-Bonmatí",
"given": "Luis"
},
{
"family": "Salahuddin",
"given": "Zohaib"
},
{
"family": "Lambin",
"given": "Philippe"
}
],
"container-title-short":
"volume": "17",
"issue": "1",
"page": "141",
"DOI": "10.1186/
"PMID": "42171860",
"PMCID": "PMC13197564",
"ISSN": "1869-4101",
"publisher": "Springer",
"URL": "https://
"language": "en",
"issued": {
"date-parts": [
[
2026,
5,
22
]
]
}
}
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:10.1038/s41467-025-64469-w [code]
- Modality-projection universal model for comprehensive full-body medical imaging segmentation.Journal: Nature communicationsIn common: nnU-Net, MONAI, pydicom, 8 other tools, 1 reference
- [2] doi:10.3389/frai.2026.1771088 [code]
- Few-shot deployment of pretrained MRI transformers in brain imaging tasks.Journal: Frontiers in artificial intelligenceIn common: nnU-Net, MONAI, pydicom, 9 other tools, structural MRI / diffusion
- [3] doi:10.1016/j.neuroimage.2024.120603 [code]
- Fetal brain MRI atlases and datasets: A review.Journal: NeuroImageIn common: nnU-Net, MONAI, pydicom, 9 other tools, structural MRI / diffusion
- [4] doi:10.3389/fonc.2022.819673 [code]
- Multimodal MRI Image Decision Fusion-Based Network for Glioma Classification.Journal: Frontiers in oncologyIn common: nnU-Net, MONAI, pydicom, 8 other tools, structural MRI / diffusion, other condition, 1 reference
- [5] doi:10.1038/s41597-024-03073-x [code]
- Brain tumor segmentation using synthetic MR images - A comparison of GANs and diffusion models.Journal: Scientific dataIn common: nnU-Net, MONAI, pydicom, 9 other tools, other condition
- [6] doi:10.1007/s00701-024-05973-8 [code]
- NnU-Net versus mesh growing algorithm as a tool for the robust and timely segmentation of neurosurgical 3D images in contrast-enhanced T1 MRI scans.Journal: Acta neurochirurgicaIn common: nnU-Net, MONAI, pydicom, 8 other tools, structural MRI / diffusion, other condition
- [7] doi:10.1371/journal.pcbi.1014555 [code]
- Body surface potential driven personalisation of electrophysiological digital twins in hypertrophic cardiomyopathy.Journal: PLoS computational biologyIn common: nnU-Net, MONAI, pydicom, 8 other tools, structural MRI / diffusion
- [8] doi:10.1016/j.isci.2024.109023 [code]
- Deep learning aided preoperative diagnosis of primary central nervous system lymphoma.Journal: iScienceIn common: nnU-Net, MONAI, pydicom, 8 other tools, other condition
- [9] doi:10.3389/fneur.2025.1620469 [code]
- Federated learning for lesion segmentation in multiple sclerosis: a real-world multi-center feasibility study.Journal: Frontiers in neurologyIn common: nnU-Net, MONAI, pydicom, 7 other tools, structural MRI / diffusion
- [10] doi:10.1038/s41593-026-02202-6 [code]
- A generalizable foundation model for analysis of human brain MRI.Journal: Nature neuroscienceIn common: MONAI, SimpleITK, OpenCV, 6 other tools, structural MRI / diffusion, other condition, 2 references
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.
Claim this paper
Correct its record
Say what each link of this record is, remove the ones that are not the paper's, add the ones that are missing. The correction becomes a new version of the record, in its Versions section.
Validate its tracing map
You validate the map as this page shows it: 1 repository of the authors' code, each at its verified commit and with its license, 57 scripts, and 12 matches between paragraphs and code (see the Code and Map sections). It then receives a DOI on Zenodo, with you (your ORCID iD) and OSCR as its creators; the code itself is not deposited.
The map's fingerprint: sha256:b4b893f0c466c5f2…
Add the badge to the code's README
If you are an author or a maintainer of this paper's code, you can add this badge. It says what OSCR knows of the paper (its code, its validated tracing map) and links the code to this page. Paste one of these snippets under the README's title. Keep the two marks around it, so that the badge can be replaced in place and is never added twice. OSCR changes nothing in your repository.
Markdown
<!-- oscr-badge:start -->
[](https://openscicode.org/badge/doi/10.1186/s13244-026-02296-3)
<!-- oscr-badge:end -->reStructuredText
.. oscr-badge:start
.. image:: https://openscicode.org/badge/doi/10.1186/s13244-026-02296-3.svg
:target: https://openscicode.org/badge/doi/10.1186/s13244-026-02296-3
:alt: OSCR
.. oscr-badge:endHTML
<!-- oscr-badge:start -->
<a href="https://openscicode.org/badge/doi/10.1186/s13244-026-02296-3"><img src="https://openscicode.org/badge/doi/10.1186/s13244-026-02296-3.svg" alt="OSCR"></a>
<!-- oscr-badge:end -->On GitHub, open the README, choose Edit (the pencil), paste the snippet under the title, then "Commit changes".
If you choose to, you can have the badge kept up to date for you, with oscr badge --write from the command line or with the GitHub Action in your repository's workflows. Every other form and option (a card, AsciiDoc, LaTeX, shields.io) is on the badge 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: code errors and mismatches between the code and the paper about this paper and its code, from readers and its authors, beside the paper and its tracing map.
Report a mismatch between the code and the paper · Report a code error (signed in)
Reproductions: reports from readers who ran the authors' code and did not get the paper's result, what they ran, with which environment, commit and data.
Report a reproduction failure (signed in)
Versions of its code: the releases of its code tied to a version of this paper (the preprint, the accepted manuscript, the version of record…), each with the tracing map versioned with it.
Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.
