PhD Students

Contact

Contact image
"This email address is being protected from spambots. You need JavaScript enabled to view it.
Download information as: vCard

Miscellaneous Information

Research interests

Statistical analysis, machine learning, cross validation, diagnostic algorithms, Alzheimer's disease

Background

My PhD research is concerned with image based diagnosis and outcome prediction in dementia using techniques from machine learning. By learning from a set of labelled 'training' examples, supervised algorithms may be able to recognise subtle and distributed changes across the brain that may not be apparent to a radiologist. We hope that these methods will allow us to recognise Alzheimer's disease in its earlier phases, allowing future treatments to be applied before too much damage has occurred.

Typical AD classification workflow

After nearly a decade of research on this topic, a lot of work has gone into the development and refinement of methods to best identify this neurological disease and others. There are already hundreds of diagnostic pipelines for Alzheimer's disease, and it is difficult to know which of them are truly the best. Part of the reason for this is that the need for training examples make it necessary to use cross validation to assess the accuracy of pipelines, and there are still many gaps in our knowledge about cross validation. I have tried to fill these in. In particular, I have tried to answer the following questions:

  1. What is the best cross validation strategy to use when assessing diagnostic algorithms?
  2. Given the unknown dependencies that can exist between component cross validation results, what are the best statistical tests to use? and
  3. How much have we biased our accuracy results by optimising our diagnostic pipelines on limited data?

 

An illustration of the origin of the bias associated with pipeline optimisation. 

 

Publications

  1. A. F. Mendelson, M. A. Zuluaga, M. Lorenzi, B. F. Hutton, and S. Ourselin.
    Selection bias in AD classification
    Submitted to NeuroImage: Clinical, 2016.
  2. M. Lorenzi, I. J. Simpson, A. F. Mendelson, S. B. Vos, M. J. Cardoso, M. Modat, J. M. Schott, and S. Ourselin.
    Multimodal image analysis in Alzheimer’s disease via statistical modelling of non-local intensity correlations.
    Scientific reports, 6, 2016.
  3. A. F. Mendelson, M. A. Zuluaga, B. F. Hutton, and S. Ourselin.
    What is the distribution of the number of unique original items in a bootstrap sample?
    arXiv preprint arXiv:1602.05822, 2016.
  4. M. A. Zuluaga, N. Burgos, A. F. Mendelson, A. M. Taylor, and S. Ourselin.
    Voxelwise atlas rating for computer assisted diagnosis: Application to congenital heart diseases of the great arteries.
    Medical image analysis, 26(1):185-194, 2015.
  5. J. Young, A. F. Mendelson, M. J. Cardoso, M. Modat, J. Ashburner, and S. Ourselin.
    Improving MRI brain image classification with anatomical regional kernels.
    In Machine Learning Meets Medical Imaging, pages 45-53. Springer International Publishing, 2015.
  6. A. F. Mendelson, M. A. Zuluaga, B. F. Hutton, and S. Ourselin.
    Bolstering heuristics for statistical validation of prediction algorithms.
    In Pattern Recognition in NeuroImaging (PRNI), 2015 International Workshop on, pages 77-80. IEEE, 2015.
  7. M. Lorenzi, I. J. Simpson, A. F. Mendelson, J. M. Cardoso, M. Modat, and S. Ourselin.
    Voxel-based statistical multimodal model of brain atrophy and hypometabolism in Alzheimer's disease.
    Alzheimer's & Dementia: The Journal of the Alzheimer's Association, 11(7):P73-P74, 2015.
  8. S. Ferraris, A. F. Mendelson, G. Ballesio, and T. Vercauteren.
    Counting sub-multisets of fixed cardinality.
    arXiv preprint arXiv:1511.06142, 2015.
  9. A. K. H. Duc, G. Eminowicz, R. Mendes, S.-L. Wong, J. McClelland, M. Modat, M. J. Cardoso, A. F. Mendelson, C. Veiga, T. Kadir, et al.
    Validation of clinical acceptability of an atlas-based segmentation algorithm for the delineation of organs at risk in head and neck cancer.
    Medical physics, 42(9):5027-5034, 2015.
  10. N. Burgos, M. J. Cardoso, A. F. Mendelson, J. M. Schott, D. Atkinson, S. R. Arridge, B. F. Hutton, and S. Ourselin.
    Subject-specific models for the analysis of pathological FDG PET data.
    In International Conference on Medical Image Computing and Computer-Assisted Intervention, pages 651-658. Springer International Publishing, 2015.
  11. M. A. Zuluaga, R. Rodionov, M. Nowell, S. Achhala, G. Zombori, A. F. Mendelson, M. J. Cardoso, A. Miserocchi, A. W. McEvoy, J. S. Duncan, et al.
    Stability, structure and scale: Improvements in multi-modal vessel extraction for SEEG trajectory planning.
    International journal of computer assisted radiology and surgery, 10(8):1227-1237, 2015.
  12. M. A. Zuluaga, A. Mendelson, M. J. Cardoso, A. M. Taylor, and S. Ourselin.
    Multi-atlas based pathological stratification of d-TGA congenital heart disease.
    In 2014 IEEE 11th International Symposium on Biomedical Imaging (ISBI), pages 109-112. IEEE, 2014.
  13. A. F. Mendelson, M. A. Zuluaga, L. Thurfjell, B. F. Hutton, and S. Ourselin.
    The empirical variance estimator for computer aided diagnosis: Lessons for algorithm validation.
    In International Conference on Medical Image Computing and Computer-Assisted Intervention, pages 236-243. Springer International Publishing, 2014.
  14. J. Young, M. Modat, M. J. Cardoso, A. F. Mendelson, D. Cash, S. Ourselin 
    Accurate multimodal probabilistic prediction of conversion to Alzheimer's disease in patients with mild cognitive impairment.
    NeuroImage: Clinical, 2:735-745, 2013.