Motion Algorithm of an ESR’s pathway to success – a sci-fi case study

[By: Iara de Almeida]

Although it may not be true that all conscience afflicted beings have wondered “What is life? Why are we here? Where did I come from?” it isn’t hard to imagine that out in the world, hyperventilating at the same frequency, PhD students around the world ask themselves those very questions.

If it is indeed true that the first three years of life have more impact than any other time period [1], can we extrapolate that thought regarding our careers? Are the first three years into an academical research career the highest contributors to its success?

Assuming that they are, the question dangled in my brain as I was asked to write this blog post: How can I take this home and make these three years the jumping point to a successful career? It is said that usually, when an article starts by asking a lot of questions, their answer is often no. Let’s see if I can turn that around, shall we?

As the algorithm developer for this project, my job is to round up all the movements a subject might make during functional Near-Infrared Spectroscopy (fNIRS – whose principles were firstly mentioned by Jöbsis [2] but whose applications were also broadly described by the lovely Aude per last month [3]) measurements and remove them from a signal, showing my colleagues the true signal, the one underlying behind big scary meaningless spikes, that does represent brain functionality. So, it will come as no surprise that I would at least try to solve my problems as rationally as possible, with mathematical model implementations.

The first step into pre-processing is to take into account systemic, cyclic physiological signals, and remove them by looking into their spectral frequency and remove values above and lower the ones where the true signal should be by using simple high and low-pass frequency filters. So doing that to cyclical tasks like a sleep schedule, calorie intake, exercise, study time should average out bad habits and implement systemic good habits in a day-to-day life.

As it comes without saying that a recurring frequency spike on your data around 1 Hz is your heartbeat, so does that to be successful the first step should be to take care of yourself according to standard parameters: exercise, drink water, eat properly.

However, these are insufficient to achieve a perfect model since, as the changes in oxyhaemoglobin and deoxyhaemoglobin in our blood, life is composed of highly non-stationary multicomponent signals. The next step, common in fNIRS and also in fMRI data, would be to apply some sort of General Linear Model, using regressor linearly independent parameters [4]. Which means that, as we need to define motion we would at this point need to define success. By representing parameters as individual contributors towards end-line successful Phd’s. The requirements diverge from institute to institute yet are always dependent on first author publications.

This ultimately implies somethings that we all well knew before I even started writing this: A well thought out plan, that turns into a relevant project with new results for the scientific community will lead into a publication, repeat that x required number of times + adding it all with a pretty cover and you will find yourself the proud owner of a PhD. So what determines a successful one?

Although the output variables of success may come in different forms considering the specific sampled subject in question, success in a professional field as a function should be dependent of your absolute contribution to your field and how that is perceived by you and your close piers.

These will ultimately influence your productivity, your creativity, your drive. Some authors even consider that those will bring you happiness, which will in turn become more success in your future [5].

Coming to the conclusion that this model will be highly related of who the subject is and that who the subject is is also a time dependent variable – and this is where it becomes tricky – the parameters should be data driven. You would need more information to characterize success, as you do for motion. The use of a Inertial Measuring Unit along your experiment gives you the needed information to localize movement [6], yet you are your own IMU – a tracker for the right path.

One cannot differentiate such subjective measures of “how others view you” from subject to subject and from day to day. One must assign weights for these parameters and how they contribute to the function of “success”, for as many sampling feedback cycles as possible.

In this case, estimating should be assessing, evaluating and planning for the next steps, predicting the best pathway to success and inputting this data back into the model, in a manner suspectingly close to the functioning of a Kalman filter [7].

To conclude, and to answer my question: can I model success and turn these 3 years into a successful career? I guess I can’t, not in a specific manner that would work for all students at least, I’m sorry, I couldn’t risk overfitting [8].

Yet, if you got my intentions from the start, I have been trying to explain to you my plans in order to remove MOTION from a noisy signal by reviewing some of the techniques I like and may use in the future. I studied, I wrote, I promise I drank a lot of water in the process and I outlined my analysis plans.

If in the mean time I succeeded in also putting a smile on your face with this silly post, then I guess that I have modelled my own ideal of success, so case and point: my job here is done.



1 –

2 – Jobsis, F. 1977. Noninvasive, infrared monitoring of cerebral and myocardial oxygen sufficiency and circulatory parameters. Science 198: 1264–1267

3 –

4 – Barker, J. W., Aarabi, A., & Huppert, T. J. (2013). Autoregressive model based algorithm for correcting motion and serially correlated errors in fNIRS. Biomedical optics express, 4(8), 1366-1379.

5 – Lyubomirsky, S., King, L., & Diener, E. (2005). The benefits of frequent positive affect: Does happiness lead to success? Psychological Bulletin, 131, 803–855.

6 – Siddiquee, M. R., Marquez, J. S., Atri, R., Ramon, R., Mayrand, R. P., & Bai, O. (2018). Movement artefact removal from NIRS signal using multi-channel IMU data. BioMedical Engineering OnLine, 17(1), 120.

7 – Tak, S., & Ye, J. C. (2014). Statistical analysis of fNIRS data: a comprehensive review. Neuroimage, 85, 72-91.

8 – Hawkins, Douglas M. The problem of overfitting. Journal of chemical information and computer sciences 44.1 (2004): 1-12.