AutoAFIDs: BIDS App for landmark detection, QC, stereotactic targeting and brain charting
Presented During:
Friday, June 27, 2025: 11:30 AM - 12:45 PM
Brisbane Convention & Exhibition Centre
Room:
M3 (Mezzanine Level)
Poster No:
1839
Submission Type:
Abstract Submission
Authors:
Alaa Taha1,2, Dhananjhay Bansal2, Jason Kai3, Tristan Kuehn2, Arun Thurairajah2, Mohamad Abbass4, Greydon Gilmore5, Ali Khan2,6, Jonathan Lau2,7
Institutions:
1School of Biomedical Engineering, Western University, London, ON, 2Robarts Research Institute, Western University, London, ON, 3Child Mind Institute, New York, NY, 4Division of Neurosurgery, London Health Sciences Center, London , ON, 5Division of Neurosurgery, Emory University, Atlanta, GA, 6Department of Medical Biophysics, Western University, London, ON, 7Division of Neurosurgery, London Health Sciences Center, London, ON
First Author:
Alaa Taha
School of Biomedical Engineering, Western University|Robarts Research Institute, Western University
London, ON|London, ON
School of Biomedical Engineering, Western University|Robarts Research Institute, Western University
London, ON|London, ON
Co-Author(s):
Ali Khan
Robarts Research Institute, Western University|Department of Medical Biophysics, Western University
London, ON|London, ON
Robarts Research Institute, Western University|Department of Medical Biophysics, Western University
London, ON|London, ON
Jonathan Lau
Robarts Research Institute, Western University|Division of Neurosurgery, London Health Sciences Center
London, ON|London, ON
Robarts Research Institute, Western University|Division of Neurosurgery, London Health Sciences Center
London, ON|London, ON
Introduction:
Brain landmark localization is an important step in many neuroimaging and clinical workflows. For instance, identifying salient regions like the anterior (AC) and posterior commissure (PC) is common during brain alignment [1], and the AC-PC line is used as a reference during stereotactic targeting [2].
We extended the search for salient brain regions by creating [3] and validating [4] an open-access protocol for annotating 32 landmarks, called anatomical fiducials (AFIDs). However, manual localization is time intensive, making automation necessary for application in large-scale datasets and clinical use.
In this work, we introduce AutoAFIDs (github.com/afids), a BIDS App for automatic landmark detection using deep learning. We demonstrate its broad utility in three downstream applications: 1) image registration, 2) stereotactic targeting and 3) brain charting.
We extended the search for salient brain regions by creating [3] and validating [4] an open-access protocol for annotating 32 landmarks, called anatomical fiducials (AFIDs). However, manual localization is time intensive, making automation necessary for application in large-scale datasets and clinical use.
In this work, we introduce AutoAFIDs (github.com/afids), a BIDS App for automatic landmark detection using deep learning. We demonstrate its broad utility in three downstream applications: 1) image registration, 2) stereotactic targeting and 3) brain charting.
Methods:
To train/test AutoAFIDs, we leverage 5 MRI datasets (n = 184) sampling field strengths (1.5, 3, and 7T) and disease (healthy, abnormal ventricles, and neurodegenerative) with 20,000+ AFIDs manually localized by expert raters [5]. A nnUNet-like framework [6] adapted for landmark regression was trained via 4-fold CV (n = 112) and tested on a stratified subset (n = 71). We report the Euclidean distance (ED) between AutoAFIDs and averaged expert rater localizations, termed AFID localization error (AFLE). End-users can acquire AFID coordinates or train on their landmarks using BIDS App syntax (Figure 1).
We validated the sensitivity for registration quality control (QC) by comparing AFLEs obtained using AutoAFIDs to those obtained by propagating AFIDs from MNI to native space using fMRIPrep (v21.0.1) [7]. End-users can compute registration QC reports (in *.html format) using the "--regqc" flag. We make use of pre-computed MNI transforms to report AFID registration error (AFRE), a vector describing error in millimeters (mm) at 32 AFIDs which survey the brain (Figure 2A).
As in prior work [8], we used predicted coordinates from AutoAFIDs to localize stereotactic targets (e.g., subthalamic nucleus (STN)) and validate accuracy using annotations by neurosurgeons on 7T MRI. End-users can acquire native space target coordinates (in *.fcsv format) using the "--stereotaxy" flag, which also provides AC-PC transforms (in *.tfm/*.mat format).
Finally, to date, we applied AutoAFIDs on 2,500+ MRI scans across adult lifespan in healthy controls (CN; n = 2,000; 4 datasets [9–12]), Parkinson's disease (PD; n = 650; PPMI [13]), and Alzheimer's disease (AD; n = 200; ADNI [14]). We compute pairwise AFID EDs to assess morphometric changes. End-users can generate stereotactic charting outputs (in *.html format) using the "--charts" flag.
We validated the sensitivity for registration quality control (QC) by comparing AFLEs obtained using AutoAFIDs to those obtained by propagating AFIDs from MNI to native space using fMRIPrep (v21.0.1) [7]. End-users can compute registration QC reports (in *.html format) using the "--regqc" flag. We make use of pre-computed MNI transforms to report AFID registration error (AFRE), a vector describing error in millimeters (mm) at 32 AFIDs which survey the brain (Figure 2A).
As in prior work [8], we used predicted coordinates from AutoAFIDs to localize stereotactic targets (e.g., subthalamic nucleus (STN)) and validate accuracy using annotations by neurosurgeons on 7T MRI. End-users can acquire native space target coordinates (in *.fcsv format) using the "--stereotaxy" flag, which also provides AC-PC transforms (in *.tfm/*.mat format).
Finally, to date, we applied AutoAFIDs on 2,500+ MRI scans across adult lifespan in healthy controls (CN; n = 2,000; 4 datasets [9–12]), Parkinson's disease (PD; n = 650; PPMI [13]), and Alzheimer's disease (AD; n = 200; ADNI [14]). We compute pairwise AFID EDs to assess morphometric changes. End-users can generate stereotactic charting outputs (in *.html format) using the "--charts" flag.
·Figure 1. Overall workflow adopted for automatic anatomical landmark (anatomical fiducials or AFIDs) localization using machine learning.
Results:
AutoAFIDs mean AFLE was 1.81 +/- 4.78 mm, which is competitive with expert human raters [3]. AutoAFIDs outperformed registration-based tools, with a conservative outlier profile (Figure 1D).
STN AFLE using AutoAFIDs was 1.11 ± 0.65 mm, outperforming a conventional non-linear registration approach which failed for 6 patients (Figure 2B). AutoAFIDs accurately predicted the STN independently of field strength/contrast via Wilcoxon signed-rank test that revealed no statistical difference in AFLE at 7T or 1.5T MRI + contrast.
Across all groups (18-100 y.o, F: 54.8%), AC-PC and inter-ventricular (IV) distance increased with age. An age- and sex- matched cross-sectional analysis between disease groups revealed statistically different AC-PC and IV distance via one-way ANOVA.
STN AFLE using AutoAFIDs was 1.11 ± 0.65 mm, outperforming a conventional non-linear registration approach which failed for 6 patients (Figure 2B). AutoAFIDs accurately predicted the STN independently of field strength/contrast via Wilcoxon signed-rank test that revealed no statistical difference in AFLE at 7T or 1.5T MRI + contrast.
Across all groups (18-100 y.o, F: 54.8%), AC-PC and inter-ventricular (IV) distance increased with age. An age- and sex- matched cross-sectional analysis between disease groups revealed statistically different AC-PC and IV distance via one-way ANOVA.
·Figure 2. Three applications of AutoAFIDs for assessing image registration, stereotactic targeting, and brain charting.
Conclusions:
We demonstrate that AutoAFIDs can localize salient landmarks with millimetric accuracy. Our data/code is designed for generalizability, enabling incorporation into other BIDS apps. AutoAFIDs can QC registration, predict challenging DBS targets, and capture morphometric changes in the brain. This work introduces a novel framework for stereotactic brain charting across lifespan, which we leverage to characterize morphometric changes in neurodegenerative disease.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) 2
Modeling and Analysis Methods:
Methods Development
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
Neuroanatomy Other
Neuroinformatics and Data Sharing:
Workflows
Informatics Other 1
Keywords:
Aging
Computational Neuroscience
Informatics
Machine Learning
Movement Disorder
Neurological
NORMAL HUMAN
Open Data
Open-Source Code
Spatial Normalization
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
2. Isaacs, B. R. et al. Methodological considerations for neuroimaging in deep brain stimulation of the subthalamic nucleus in Parkinson’s disease patients. J. Clin. Med. 9, 3124 (2020).
3. Lau, J. C. et al. A framework for evaluating correspondence between brain images using anatomical fiducials. Hum. Brain Mapp. 40, 4163–4179 (2019).
4. Abbass, M. et al. Application of the anatomical fiducials framework to a clinical dataset of patients with Parkinson’s disease. Brain Struct. Funct. 227, 393–405 (2022).
5. Taha, A. et al. Magnetic resonance imaging datasets with anatomical fiducials for quality control and registration. Sci Data 10, 449 (2023).
6. Isensee, F., Jaeger, P. F., Kohl, S. A. A., Petersen, J. & Maier-Hein, K. H. nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nat. Methods 18, 203–211 (2021).
7. Esteban, O. et al. fMRIPrep: a robust preprocessing pipeline for functional MRI. Nat. Methods 16, 111–116 (2019).
8. Taha, A., Gilmore, G., Khan, A. & Lau, J. C. An indirect deep brain stimulation targeting tool using salient anatomical fiducials. Neuromodulation 25, S6–S7 (2022).
9. Snoek, L. et al. The Amsterdam Open MRI Collection, a set of multimodal MRI datasets for individual difference analyses. Sci. Data 8, 85 (2021).
10. Taylor, J. R. et al. The Cambridge Centre for Ageing and Neuroscience (Cam-CAN) data repository: Structural and functional MRI, MEG, and cognitive data from a cross-sectional adult lifespan sample. Neuroimage 144, 262–269 (2017).
11. Van Essen, D. C. et al. The WU-Minn Human Connectome Project: an overview. Neuroimage 80, 62–79 (2013).
12. Spreng, R. N. et al. Neurocognitive aging data release with behavioral, structural and multi-echo functional MRI measures. Sci. Data 9, 119 (2022).
13. Marek, K. et al. The Parkinson’s progression markers initiative (PPMI) - establishing a PD biomarker cohort. Ann. Clin. Transl. Neurol. 5, 1460–1477 (2018).
14. Petersen, R. C. et al. Alzheimer’s Disease Neuroimaging Initiative (ADNI): clinical characterization. Neurology 74, 201–209 (2010).