2026
AI-Based Pipeline for the Segmentation of White Matter Hypoattenuations in CT Scans: A Design-Choice Validation Study
Nada Alamoudi, Maria del C. Valdés Hernández, Benjamin Jin, Eleni Sakka, Carmen Arteaga-Reyes, Daniela Jaime García, Yajun Cheng, Angela C.C. Jochems, Grant Mair, Joanna M. Wardlaw, Miguel O. Bernabeu
medRxiv 2026
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 pseudolabelled 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.
AI-Based Pipeline for the Segmentation of White Matter Hypoattenuations in CT Scans: A Design-Choice Validation Study
Nada Alamoudi, Maria del C. Valdés Hernández, Benjamin Jin, Eleni Sakka, Carmen Arteaga-Reyes, Daniela Jaime García, Yajun Cheng, Angela C.C. Jochems, Grant Mair, Joanna M. Wardlaw, Miguel O. Bernabeu
medRxiv 2026
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 pseudolabelled 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.
2025
Calibrated Self-supervised Vision Transformers Improve Intracranial Arterial Calcification Segmentation from Clinical CT Head Scans
Benjamin Jin, Grant Mair, Joanna M. Wardlaw, Maria del C. Valdés Hernández
Data Engineering in Medical Imaging @ Medical Image Computing and Computer Assisted Interventions (MICCAI) 2025
Vision Transformers (ViTs) have gained significant popularity in the natural image domain but have been less successful in 3D medical image segmentation. Nevertheless, 3D ViTs are particularly interesting for large medical imaging volumes due to their efficient self-supervised training within the masked autoencoder (MAE) framework, which enables the use of imaging data without the need for expensive manual annotations. intracranial arterial calcification (IAC) is an imaging biomarker visible on routinely acquired CT scans linked to neurovascular diseases such as stroke and dementia, and automated IAC quantification could enable their large-scale risk assessment. We pre-train ViTs with MAE and fine-tune them for IAC segmentation for the first time. To develop our models, we use highly heterogeneous data from a large clinical trial, the third International Stroke Trial (IST-3). We evaluate key aspects of MAE pre-trained ViTs in IAC segmentation, and analyse the clinical implications. We show: 1) our calibrated self-supervised ViT beats a strong supervised nnU-Net baseline by 3.2 Dice points, 2) low patch sizes are crucial for ViTs for IAC segmentation and interpolation upsampling with regular convolutions is preferable to transposed convolutions for ViT-based models, and 3) our ViTs increase robustness to higher slice thicknesses and improve risk group classification in a clinical scenario by 46%.
Calibrated Self-supervised Vision Transformers Improve Intracranial Arterial Calcification Segmentation from Clinical CT Head Scans
Benjamin Jin, Grant Mair, Joanna M. Wardlaw, Maria del C. Valdés Hernández
Data Engineering in Medical Imaging @ Medical Image Computing and Computer Assisted Interventions (MICCAI) 2025
Vision Transformers (ViTs) have gained significant popularity in the natural image domain but have been less successful in 3D medical image segmentation. Nevertheless, 3D ViTs are particularly interesting for large medical imaging volumes due to their efficient self-supervised training within the masked autoencoder (MAE) framework, which enables the use of imaging data without the need for expensive manual annotations. intracranial arterial calcification (IAC) is an imaging biomarker visible on routinely acquired CT scans linked to neurovascular diseases such as stroke and dementia, and automated IAC quantification could enable their large-scale risk assessment. We pre-train ViTs with MAE and fine-tune them for IAC segmentation for the first time. To develop our models, we use highly heterogeneous data from a large clinical trial, the third International Stroke Trial (IST-3). We evaluate key aspects of MAE pre-trained ViTs in IAC segmentation, and analyse the clinical implications. We show: 1) our calibrated self-supervised ViT beats a strong supervised nnU-Net baseline by 3.2 Dice points, 2) low patch sizes are crucial for ViTs for IAC segmentation and interpolation upsampling with regular convolutions is preferable to transposed convolutions for ViT-based models, and 3) our ViTs increase robustness to higher slice thicknesses and improve risk group classification in a clinical scenario by 46%.
2024
Pre-processing and Quality Control of Large Clinical CT Head Datasets for Intracranial Arterial Calcification Segmentation
Benjamin Jin, Maria del C. Valdés Hernández, Alessandro Fontanellea, Wenwen Li, Eleanor Platt, Paul Armitage, Amos Storkey, Joanna M. Wardlaw, Grant Mair
Data Engineering in Medical Imaging @ Medical Image Computing and Computer Assisted Interventions (MICCAI) 2024
As a potential non-invasive biomarker for ischaemic stroke, intracranial arterial calcification (IAC) could be used for stroke risk assessment on CT head scans routinely acquired for other reasons (e.g. trauma, confusion). Artificial intelligence methods can support IAC scoring, but they have not yet been developed for clinical imaging. Large heterogeneous clinical CT datasets are necessary for the training of such methods, but they exhibit expected and unexpected data anomalies. Using CTs from a large clinical trial, the third International Stroke Trial (IST-3), we propose a pipeline that uses as input non-enhanced CT scans to output regions of interest capturing selected large intracranial arteries for IAC scoring. Our method uses co-registration with templates. We focus on quality control, using information presence along the z-axis of the imaging to group and apply similarity measures (structural similarity index measure) to triage assessment of individual image series. Additionally, we propose superimposing thresholded binary masks of the series to inspect large quantities of data in parallel. We identify and exclude unrecoverable samples and registration failures. In total, our pipeline processes 10,659 CT series, rejecting 4,322 (41%) in the entire process, 1,450 (14% of the total) during quality control, and outputting 6,337 series. Our pipeline enables effective and efficient region of interest localisation for targeted IAC segmentation.
Pre-processing and Quality Control of Large Clinical CT Head Datasets for Intracranial Arterial Calcification Segmentation
Benjamin Jin, Maria del C. Valdés Hernández, Alessandro Fontanellea, Wenwen Li, Eleanor Platt, Paul Armitage, Amos Storkey, Joanna M. Wardlaw, Grant Mair
Data Engineering in Medical Imaging @ Medical Image Computing and Computer Assisted Interventions (MICCAI) 2024
As a potential non-invasive biomarker for ischaemic stroke, intracranial arterial calcification (IAC) could be used for stroke risk assessment on CT head scans routinely acquired for other reasons (e.g. trauma, confusion). Artificial intelligence methods can support IAC scoring, but they have not yet been developed for clinical imaging. Large heterogeneous clinical CT datasets are necessary for the training of such methods, but they exhibit expected and unexpected data anomalies. Using CTs from a large clinical trial, the third International Stroke Trial (IST-3), we propose a pipeline that uses as input non-enhanced CT scans to output regions of interest capturing selected large intracranial arteries for IAC scoring. Our method uses co-registration with templates. We focus on quality control, using information presence along the z-axis of the imaging to group and apply similarity measures (structural similarity index measure) to triage assessment of individual image series. Additionally, we propose superimposing thresholded binary masks of the series to inspect large quantities of data in parallel. We identify and exclude unrecoverable samples and registration failures. In total, our pipeline processes 10,659 CT series, rejecting 4,322 (41%) in the entire process, 1,450 (14% of the total) during quality control, and outputting 6,337 series. Our pipeline enables effective and efficient region of interest localisation for targeted IAC segmentation.
2023
Airborne pollen grain detection from partially labelled data utilising semi-supervised learning
Benjamin Jin, Manuel Milling, Maria Pilar Plaza, Jens O. Brunner, Claudia Traidl-Hoffmann, Björn W. Schuller, Athanasios Damialis
Science of The Total Environment 2023
Airborne pollen monitoring has been conducted for more than a century now, as knowledge of the quantity and periodicity of airborne pollen has diverse use cases, like reconstructing historic climates and tracking current climate change, forensic applications, and up to warning those affected by pollen-induced respiratory allergies. Hence, related work on automation of pollen classification already exists. In contrast, detection of pollen is still conducted manually, and it is the gold standard for accuracy. So, here we used a new-generation, automated, near-real-time pollen monitoring sampler, the BAA500, and we used data consisting of both raw and synthesised microscope images. Apart from the automatically generated, commercially-labelled data of all pollen taxa, we additionally used manual corrections to the pollen taxa, as well as a manually created test set of bounding boxes and pollen taxa, so as to more accurately evaluate the real-life performance. For the pollen detection, we employed two-stage deep neural network object detectors. We explored a semi-supervised training scheme to remedy the partial labelling. Using a teacher-student approach, the model can add pseudo-labels to complete the labelling during training. To evaluate the performance of our deep learning algorithms and to compare them to the commercial algorithm of the BAA500, we created a manual test set, in which an expert aerobiologist corrected automatically annotated labels. For the novel manual test set, both the supervised and semi-supervised approaches clearly outperform the commercial algorithm with an F1 score of up to 76.9 % compared to 61.3 %. On an automatically created and partially labelled test dataset, we obtain a maximum mAP of 92.7 %. Additional experiments on raw microscope images show comparable performance for the best models, which potentially justifies reducing the complexity of the image generation process. Our results bring automatic pollen monitoring a step forward, as they close the gap in pollen detection performance between manual and automated procedure.
Airborne pollen grain detection from partially labelled data utilising semi-supervised learning
Benjamin Jin, Manuel Milling, Maria Pilar Plaza, Jens O. Brunner, Claudia Traidl-Hoffmann, Björn W. Schuller, Athanasios Damialis
Science of The Total Environment 2023
Airborne pollen monitoring has been conducted for more than a century now, as knowledge of the quantity and periodicity of airborne pollen has diverse use cases, like reconstructing historic climates and tracking current climate change, forensic applications, and up to warning those affected by pollen-induced respiratory allergies. Hence, related work on automation of pollen classification already exists. In contrast, detection of pollen is still conducted manually, and it is the gold standard for accuracy. So, here we used a new-generation, automated, near-real-time pollen monitoring sampler, the BAA500, and we used data consisting of both raw and synthesised microscope images. Apart from the automatically generated, commercially-labelled data of all pollen taxa, we additionally used manual corrections to the pollen taxa, as well as a manually created test set of bounding boxes and pollen taxa, so as to more accurately evaluate the real-life performance. For the pollen detection, we employed two-stage deep neural network object detectors. We explored a semi-supervised training scheme to remedy the partial labelling. Using a teacher-student approach, the model can add pseudo-labels to complete the labelling during training. To evaluate the performance of our deep learning algorithms and to compare them to the commercial algorithm of the BAA500, we created a manual test set, in which an expert aerobiologist corrected automatically annotated labels. For the novel manual test set, both the supervised and semi-supervised approaches clearly outperform the commercial algorithm with an F1 score of up to 76.9 % compared to 61.3 %. On an automatically created and partially labelled test dataset, we obtain a maximum mAP of 92.7 %. Additional experiments on raw microscope images show comparable performance for the best models, which potentially justifies reducing the complexity of the image generation process. Our results bring automatic pollen monitoring a step forward, as they close the gap in pollen detection performance between manual and automated procedure.