Back to Projects

AI-enabled Imaging, Emerging Imaging

Machine learning for improved clinical decision making ahead of epilepsy surgery

Project ID: 2022_019

1st Supervisor: Alexander Hammers , King’s College London
2nd Supervisor: Jorge Cardoso , King’s College London
Clinical Supervisor: Alexander Hammers , King’s College London

Aim of the PhD Project:

  • Automate metabolic lesion detection and segmentation on FDG-PET using normal/abnormal weakly-supervised signal decomposition 
  • Differentiating seizure-onset and seizure-spread areas 
  • Harness biological and clinical knowledge for optimised detection of relevant lesions 
  • Stretch goals: 
    • Sandbox clinical tool 
    • Routine measurement of multiple region volumes and characteristics 
    • Incorporate semantic non-imaging information

Project description/background:

Background: Epilepsy is the most common serious neurological condition, with >600,000 people affected in the UK (https://www.epilepsysociety.org.uk/about-epilepsy). Over 30% of people with epilepsy have medication-resistant seizures. Focal seizures start in one part of the brain, and surgery may be an option if the epileptogenic zone can be identified. Pre-surgical assessment can take >1yr and involves a range of diagnostic procedures. Imaging, particularly MRI, has a central role in this process. [18F]fluorodeoxyglucose position emission tomography (FDG-PET) is much more sensitive than MRI and particularly well developed in our Centre. Current clinical evaluation of imaging relies on time-consuming and subjective visual analysis and is challenging even for experts. 

Focal cortical dysplasias (FCDs) are small malformations of the brain and one of the most common substrates underlying refractory epilepsy. They can often be detected by MRI alone, but 15-30% of presurgical patients are “MRI-negative”, with the proportion likely higher in national referral centres. In those, FDG-PET has become much relied upon over the past five years: it is essential to have very good localization hypotheses prior to implantation of intracranial electrodes, as these are invasive, carry a small risk, and can at best sample ~10% of brain, with seizure onset zones potentially undetected if electrodes are placed just a few mm away. 

Deep learning methods have gained traction for detecting and characterising abnormalities and lesions [1-3]. As they are fast once trained, they are more suitable than traditional statistical/machine learning methods [4-10] for implementation on clinical workstations. Deep learning work has exclusively focused on MRI analysis, largely ignoring the high yield of FDG-PET, especially in MRI-negative patients. One ML study has investigated FDG-PET, based on asymmetries alone and ignoring MRI [11]; another combined FDG-PET and MRI but used handcrafted features and Support Vector Machines (SVMs) / patch-based classification rather than deep learning [12]. 

This research will demonstrate the feasibility and usefulness of quantitative joint analysis of FDG PET and MR in the clinical setting. It will support clinicians with patient management and may enable more patients to have surgery, which could represent significant cost savings and a positive impact on patient quality of life. 

Goals: 1) Produce tools for quantitative analysis of FDG-PET scans of patients with epilepsy. 2) Combine with MR-based tools 3) Obtain prospective validation of the tools developed and obtain clinician feedback 4) Integrate non-imaging information into the image analysis, e.g. depression scores, EEG data, and semiology (via semantic information or ictal video-EEG directly) 

Prior work: Alexander Hammers has applied quantitative image analysis to epilepsy, including automatic abnormality detection for FDG-PET scans as an objective complement to clinical reporting; automatic segmentation of T1-weighted MRIs allowing access to useful quantitative measures e.g. intracranial or hippocampal volumes; and ML-based attenuation correction for PET-MR. Jorge Cardoso is a recognized expert in AI/machine learning techniques and big data management and has already worked on FCD detection. 

Link to other work in the School: EpiNav depth electrode implantation and analysis software (Rachel Sparks), AI Ethics (Robin Carpenter), AIDE/DICOMserver (Jorge Cardoso), optimized PET reconstruction using MR information (Andrew Reader), DL-based brain lesion detection and segmentation (Tom Booth), whole-body PET/MRI PET/CT abnormality detection for cancer patients (Ashay Patel / Jorge Cardoso), etc. 

The project will suit candidates from engineering, computer science, physics or related areas, or alternatively clinically trained candidates with a strong science/computing background.

Illustration of medical images and how an artificial intelligence algorithm can help detect and segment lesions ahead of epilepsy surgery. Left, PET image on its own where the relevant lesion can not be seen (similarly invisible on MRI alone). Middle, example of a Deep Learning model. Right, PET image overlaid on MR showing the relevant lesion

References

  1. 1. Adler, S., et al., Novel surface features for automated detection of focal cortical dysplasias in paediatric epilepsy. Neuroimage Clin, 2017. 14: p. 18-27. 
  2. Bernasconi, A., et al.,Recommendations for the use of structural magnetic resonance imaging in the care of patients with epilepsy: A consensus report from the International League Against Epilepsy Neuroimaging Task Force.Epilepsia, 2019. 60(6): p. 1054-1068. 
  3. Gill, R.S., et al.,Multicenter Validation of a Deep Learning Detection Algorithm for Focal Cortical Dysplasia.Neurology, 2021. 
  4. Antel, S.B., et al.,Automated detection of focal cortical dysplasia lesions using computational models of their MRI characteristics and texture analysis.Neuroimage, 2003. 19(4): p. 1748-59. 
  5. Huppertz, H.J., et al.,Enhanced visualization of blurred gray-white matter junctions in focal cortical dysplasia by voxel-based 3D MRI analysis.Epilepsy Res, 2005. 67(1-2): p. 35-50. 
  6. Huppertz, H.J., M. Kurthen, and J. Kassubek,Voxel-based 3D MRI analysis for the detection of epileptogenic lesions at single subject level.Epilepsia, 2009. 50(1): p. 155-6. 
  7. Hong, S.J., et al.,Automated detection of cortical dysplasia type II in MRI-negative epilepsy.Neurology, 2014. 83(1): p. 48-55. 
  8. El Azami, M., et al.,Detection of Lesions Underlying Intractable Epilepsy on T1-Weighted MRI as an Outlier Detection Problem.PLoS One, 2016. 11(9): p. e0161498. 
  9. Keihaninejad, S., et al.,Classification and lateralization of temporal lobe epilepsies with and without hippocampal atrophy based on whole-brain automatic MRI segmentation.PLoS One, 2012. 7(4): p. e33096. 
  10. Hammers, A., et al.,Automatic detection and quantification of hippocampal atrophy on MRI in temporal lobe epilepsy: a proof-of-principle study.Neuroimage, 2007. 36(1): p. 38-47. 
  11. Zhang, Q., et al.,A deep learning framework for (18)F-FDG PET imaging diagnosis in pediatric patients with temporal lobe epilepsy.Eur J Nucl Med Mol Imaging, 2021. 
  12. Tan, Y.L., et al.,Quantitative surface analysis of combined MRI and PET enhances detection of focal cortical dysplasias.Neuroimage, 2018. 166: p. 10-18. 

Back to Projects