Mohamed Abdelhack @mabdelhack@qoto.org

Just published‼🎇
https://www.nature.com/articles/s41467-023-43737-7
We found contradictions in brain-wide associations of sleep 😴 and depression when participants were doing a task vs. resting state.
What does that mean❓❗

Resting-state brain dynamics were found to change sleep duration increase in a similar fashion as .... (wait for it)... insomnia frequency increase AND depression symptoms increase 🤨🤨
This was very robust observed in over 30k participants from
UK Biobank & Human Connectome Project.

Similarly, we found hyperconnectivity ⏫ associations of insomnia & depression in resting state but hypoconnectivity ⏬ in task
What does that all tell us ⁉

In insomnia and depression, resting-state dynamics are resembling those of rested-wakefulness... Your brain is giving you a signal that you don't need to sleep which looks like a hyperconnected state (hyperarousal?).

BUT in task connectivity drops which could be signifying a "local sleep" phase which decreases your cognitive performance as the brain is unable to recruit its resources to perform the task.

Happy to finally get this paper out. We devised MFCL, a DNN layer with weights that can be modulated. It is a simple all-in-one solution to multiple issues with data quality.

Missing data: by putting missingness flags as input to the weight modulation network, classification models performed better than models with imputation and even XGBoost on various missingness paradigms. When more missingness was introduced at testing, MFCL was the most robust...

We also removed a whole input feature and MFCL did better than all other networks in almost all the tasks. Use case? Think of transferring a model to a low-resource setting where the healthcare facility has less capability to do all measurements...

Data quality and missingness: We cannot even use imputation when input data has both quality measures and missing data. MFCL though can have these as a modulating signal and it outperforms networks with these signals concatenated with input...

Imputation: want to impute the data using DL? We tested adding MFCL at the first layer in an autoencoder-based imputer and it improved its performance in the non-random removal case...

Interestingly, it outperformed XGBoost which is known to perform better on tabular data (check: https://hal.science/hal-03723551), especially with additional removal of data and on larger datasets

Just published‼🎇
https://www.nature.com/articles/s41467-023-43737-7
We found contradictions in brain-wide associations of sleep 😴 and depression when participants were doing a task vs. resting state.
What does that mean❓❗

Resting-state brain dynamics were found to change sleep duration increase in a similar fashion as .... (wait for it)... insomnia frequency increase AND depression symptoms increase 🤨🤨
This was very robust observed in over 30k participants from
UK Biobank & Human Connectome Project.

Similarly, we found hyperconnectivity ⏫ associations of insomnia & depression in resting state but hypoconnectivity ⏬ in task
What does that all tell us ⁉

In insomnia and depression, resting-state dynamics are resembling those of rested-wakefulness... Your brain is giving you a signal that you don't need to sleep which looks like a hyperconnected state (hyperarousal?).

BUT in task connectivity drops which could be signifying a "local sleep" phase which decreases your cognitive performance as the brain is unable to recruit its resources to perform the task.

Happy to finally get this paper out. We devised MFCL, a DNN layer with weights that can be modulated. It is a simple all-in-one solution to multiple issues with data quality.

Missing data: by putting missingness flags as input to the weight modulation network, classification models performed better than models with imputation and even XGBoost on various missingness paradigms. When more missingness was introduced at testing, MFCL was the most robust...

We also removed a whole input feature and MFCL did better than all other networks in almost all the tasks. Use case? Think of transferring a model to a low-resource setting where the healthcare facility has less capability to do all measurements...

Data quality and missingness: We cannot even use imputation when input data has both quality measures and missing data. MFCL though can have these as a modulating signal and it outperforms networks with these signals concatenated with input...

Imputation: want to impute the data using DL? We tested adding MFCL at the first layer in an autoencoder-based imputer and it improved its performance in the non-random removal case...

Interestingly, it outperformed XGBoost which is known to perform better on tabular data (check: https://hal.science/hal-03723551), especially with additional removal of data and on larger datasets

Hello world!
Since I'm still discovering this Mastodon thing, I will be posting sporadically (well... that's not much different from my Twitter anyway). I will be mainly sharing interesting science stuff and job/educational opportunities that I stumble upon.
Happy Holidays!