Explainable AI for Visual Snow Syndrome

Primary supervisor

Leimin Tian

Co-supervisors

Visual Snow Syndrome (VSS) is a recently identified neurological disorder that the medical community still does not know much about. Diagnosing VSS is challenging and mostly reliant on subjective reports from patients. Existing studies demonstrated that VSS patients exhibit different behaviours compared to healthy controls in visual attention tasks, specifically pro-saccade, anti-saccade, and pro/anti switching tests. Using data collected at the Neurology department, pilot deep learning experiments have been conducted using the LSTM model with attention mechanism. These pilot experiments yield promising results indicating that automatic diagnosis of VSS based on a person’s visual attention task performance is possible, and the classifier can achieve performance on-par or even more accurate than human doctors.

    Student cohort

    Double Semester

    Aim/outline

    In this project, you will extend the pilot experiments with two objectives:

    1. Use explainable AI tools, such as LIME or InterpretML, to investigate the pilot LSTM model and understand what makes it so effective. This may lead to discoveries on the causes or treatments of VSS.
    2. Improve the pilot LSTM model if possible, and test other deep learning or machine learning models that may yield reliable classification performance.

    URLs/references

    [1] Marcel Arnold. 2018. Headache classification committee of the international headache society (IHS) the international classification of headache disorders. Cephalalgia 38, 1 (2018), 1–211.

    [2] HR Jäger, NJ Giffin, and PJ Goadsby. 2005. Diffusion-and perfusion-weighted MR imaging in persistent migrainous visual disturbances. Cephalalgia 25, 5 (2005), 323–332.

    [3] Jenny L Lauschke, Gordon T Plant, and Clare L Fraser. 2016. Visual snow: A thalamocortical dysrhythmia of the visual pathway?Journal of ClinicalNeuroscience28 (2016), 123–127.

    [4] GT Liu, NJ Schatz, SL Galetta, NJ Volpe, F Skobieranda, and GS Kosmorsky. 1995. Persistent positive visual phenomena in migraine. Neurology 45, 4(1995), 664–668.

    [5] Abby I Metzler and Carrie E Robertson. 2018. Visual snow syndrome: Proposed criteria, clinical implications, and pathophysiology.Current neurology and neuroscience reports 18, 8 (2018), 52.

    [6] Francesca Puledda, Christoph Schankin, Kathleen Digre, and Peter J Goadsby. 2018. Visual snow syndrome: what we know so far.Current opinion in neurology 31, 1 (2018), 52–58.

    [7] Francesca Puledda, Christoph Schankin, and Peter J Goadsby. 2020. Visual snow syndrome: A clinical and phenotypical description of 1,100 cases.Neurology(2020).

    [8] Christoph J Schankin, Farooq H Maniyar, Kathleen B Digre, and Peter J Goadsby. 2014. ’Visual snow’–a disorder distinct from persistent migraine aura. Brain 137, 5 (2014), 1419–1428.

    [9] Jane C Simpson, Peter J Goadsby, and Prab Prabhakar. 2013. Positive persistent visual symptoms (visual snow) presenting as a migraine variant in a12-year-old girl.Pediatric neurology 49, 5 (2013), 361–363.

    [10] Ghislaine L Traber, Marco Piccirelli, and Lars Michels. 2020. Visual snow syndrome: a review on diagnosis, pathophysiology, and treatment.CurrentOpinion in Neurology 33, 1 (2020), 74–78.