The Future of Seizure Forecasting

Prophecies and predictions they are not …


Is seizure prediction possible?

When a distinguished but elderly scientist states that something is possible, she is almost certainly right. When she states that something is impossible, she is very probably wrong

What steps are needed for seizure prediction?

The only way of discovering the limits of the possible is to venture a little way past them into the impossible.

How will seizure prediction work?

Any sufficiently advanced technology is indistinguishable from magic.

Those were the three laws of Arthur C. Clarke (of 2001: A Space Odyssey fame). I’m going to try to answer the questions a little more concretely, but first – a story

Weather Forecasting

Lewis Fry Richardson was an interesting guy. A pacifist and passionate meteorologist, he worked  as an ambulance driver in France during WWI. During this time he produced the calculations that would later to become the first ‘modern-day’ weather forecast.

Lewis Fry Richardson, 1881 – 1953

In 1922, Richardson published a prediction about the state of the atmosphere in two places over central Europe. To make a forecast just six hours into the future took  him six weeks of manual calculation.

He ended up being wrong …

However, the value of Richardson’s work was in his ideas rather than his results. In a sense, Richardson was ahead of his time. Upon discovering his work was relevant to designing chemical weapons, his beliefs led him to drop meteorology altogether. Years later, eb18197c42261941baaf4fb0fb1fe923with the invention of computers and supercomputing, Richardson’s forecasting methods came to fruition. The same basic approach is in use today, and the results can be found on smartphones everywhere.

Lewis Fry Richardson was helped by his contemporary, Admiral Robert FitzRoy, who set the scene. In fact, FitzRoy didn’t just set the scene for weather forecasting, he invented the term.

Admiral Robert Fitzroy, 1805 – 1865

Robert FitzRoy was captain of the HMS Beagle during Charles Darwin’s famous voyage of discovery, during which time he (supposedly) coined the term ‘forecasts’ as applied to the weather

Prophecies and predictions they are not, the term forecast is strictly applicable to such an opinion as is the result of scientific combination and calculation.

During his life, FitzRoy was continually fighting to defend his forecasts, which were published for five years in British papers (1860-1865).  It is probable that the scorn for his life’s work contributed to his depression and eventual suicide. Now, FitzRoy’s contributions are recognized as a crucial stage in the history of weather forecasting.

An Abridged Timeline of Weather Forecasting

  • 1450 – 1650: Measurement Devices invented – Hygrometer (air humidity, Nicholas Cusa), thermometer (temperature, Galileo Galilei), and barometer (atmospheric pressure, Evangelista Torricelli)
  • 1854: Forecasting was invented as a concept for predicting weather patterns
A hand-drawn weather map by Robert FitzRoy
  • 1860: Somewhat Big Data Global weather-observing stations and telegraph enabled analysis and sharing of long-term, widespread data
  • 20th Century: Advancing Technology led to many improvements and additions to the devices used to measure weather data
  • 1922: Numerical Methods combined observation data with mathematical models of the earth’s atmosphere. Models are simulated forward in time to see what conditions we should expect based on the current state. The same basic principle for weather forecasting is still in use today.
  • 1946: Computing  ENIAC arrives on the scene – a computer capable of performing the calculations needed to run numerical models
  • 21st Century: Supercomputers & Smartphones there have been many, many computing & mathematical advances to bring weather forecasting to its current state.

So, we’re looking at about 350 years from the first measurements to accurate, user-friendly weather forecasting apps. Can we relate this process to seizure prediction?

  • Measurement devices (1924): The main measurement device for brain waves (electroencephalography, EEG) was invented almost 100 years ago. Barring unforeseen developments, seizure predictors will be developed for EEG, because EEG is the only practical way we can measure long-term brain signals.
Signals recorded during the first ever EEG


  • Advanced Technology (20th Century+): There have been many breakthroughs in recording technology, a crucial one being the ability to conduct long-term recordings (still via fairly invasive surgical implantation of electrodes; although – watch this space). More breakthroughs are needed. In particular, the ability to record signals with higher density (spatial resolution) from widespread brain regions simultaneously.
Kaggle seizure prediction competition 
  • Somewhat Big Data (2000s): The advent of long-term recording from epilepsy patients and data sharing initiatives is having a huge impact on the field of seizure prediction. (For examples of long-term monitoring data see EPILEPSIAE, SeizureTracker, NeuroVista Advisory System. For more on data sharing initiatives see Kaggle, iEEG portal, my previous blog, these essays – onetwo)

  • Numerical Methods:  To continue with the weather analogy, numerical methods for seizure forecasting require a mathematical model for simulating brain dynamics. Building accurate brain models could prove to be the biggest barrier facing seizure prediction. However, there is some fascinating work using mathematics to analyse seizure dynamics (for instance this work on critical brain dynamics, or our work on inverse estimation of brain activity).


  • Computing: A challenge of computing for seizure prediction is that it may need to be done in fully implantable devices. However, computing and chip technology is already advanced enough to meet this requirement (for example, see IBM’s TrueNorth chip)

  • Supercomputers & Smartphones: The best way to do seizure prediction is to ‘just do it’. We have the means to implement simple but useful predictors for large numbers of patients via external devices and smartphones. The reason implementation is so valuable is because it enables algorithms to learn from past mistakes and incorporate user feedback.


  • Forecasting: I believe we are in the midst of a fundamental shift in what it means to predict patients’ seizures. The essence of this shift is the move from prediction to forecasting. You can read more about that idea in our recent review paper.

As to the future – we still have 250 years to get seizure forecasting to the same stage we’re at with weather forecasting. I think we’ll have no problem making this deadline. I only hope there is more acceptance for early efforts at seizure forecasting than there was for FitzRoy’s weather predictions.


Arthur C. Clarke ‘Hazards of Prophecy: The Failure of Imagination’ in the collection Profiles of the Future: An Enquiry into the Limits of the Possible (1962, rev. 1973), pp. 14, 21, 36.

Seizure prediction: science fiction or soon to become reality?. by Freestone et al. Current Neurology and Neuroscience Reports 15.11 (2015): 1-9.

Weather Forecasting Through the Ages by Steve Graham, Claire Parkinson, and Mous Chahine NASA Earth Observatory (February 2002)

The birth of the weather forecast by Peter Moore, BBC News Magazine (April 2015)

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s