Andrew Doyle

I'm a Research Scientist at Elekta where I work on new technologies to deliver radiotherapy to cancer patients as safely as possible. Previously, I was a software developer/researcher working at the Montreal Neurological Institute.

I'm very interested in deep learning, and organized a Deep Learning for Human Brain Mapping course for the Organization for Human Brain Mapping to teach neuroscientists how deep learning can work for them. I really like visualizations, and some of my brainart can be found in the in the brains section.


Neural Power Spectra
Time-domain analysis using features of models of power spectra and other bases.

Parametric models of EEG power spectra can be used to design digital filters that select different components of the neural signal. Hypotheses based on these models can be validated using time-domain deep learning decoding experiments.

Related pre-prints:

  1. Doyle, JA "Amplifying the Neural Power Spectrum." bioRxiv, 2019.
  2. Doyle, JA "What Colour is Neural Noise?" arXiv, 2018.

Related posters:

  1. "Amplifying the Neural Power Spectrum." OHBM, 2019.

Infant Brain MRI Segmentation
MICCAI 2017 Grand Challenge on 6-month infant brain MRI Segmentation

Obtaining accurate white/gray matter segmentations is difficult because the brain's myelin is still growing. Accurate longitudinal segmentations can be used to characterize brain development.

Related publications:

  1. Wang, Li et al. Benchmark on Automatic 6-month-old Infant Brain Segmentation Algorithms: The iSeg-2017 Challenge." IEEE, 2019.
  2. Fonov, Vladimir, et al. "NeuroMTL iSEG challenge methods." bioRxiv (2018): 278465.

Challenge website

AutoQC of T1w MRI
Motion detection and other artifact Quality Control for MRI

This project uses CNNs to automatically predict whether scans will pass quality control and be useful for follow-up study.

Related posters:

  1. Doyle, Andrew, et al. "Automatic Quality Control of Infant MRI." Poster presentation at Montreal AI Symposium, 2017.
  2. Doyle, Andrew, et al. "Automatic Quality Control of Human Brain T1w MRI Scans." Poster presentation at Medical Imaging meets NIPS Workshop of NIPS, 2017.

Neuroinformatics platform for distributed computation

In CBRAIN I worked mainly on implementing the RESTful API and describing it in the OpenAPI specification, as well as created the user registration system.

Related poster:

  1. Doyle, Andrew, et al. "CBRAIN as a Platform for Open Science." Poster presentation at Human Brain Mapping Conference, 2017.


Here are some examples of the 3D printing I've done of my own brain and the BigBrain, as well as some artwork I've generated using style transfer of Nina Cherney's artwork on CIVET surfaces and slices extracted from my own brain MRI.



You can reach me on Twitter or send me an email.

© Andrew Doyle 2020