12/ In summary using an unusual dataset of direct neural recordings, our findings suggest that several highly interconnected circuit activity patterns are involved in depression. We hope that they may aid in the development of novel personalized treatment approaches.

11/ But what about connectivity – how does this relate? In a Connectivity Analysis looking at inter- and intra-modular connectivity, we found weaker connectivity throughout in our first biotype and increased left frontal connectivity in the second.

10/ We might have stopped there – but wanted the biomarkers to be useful clinically and address heterogeneity in depression. Using hierarchical clustering we found common features clustered participants into 2 biotypes each with their own network expression pattern.

9/ Model Utilization: In an Activity Analysis we defined spectral power features across network modules and applied supervised ML to identify a group-level biomarker. Mean classification accuracy of 82.4% on the test set (sensitivity=0.833, positive-predictive value = 0.87).

8/ We then took a 2-step approach:.Model Building: iEEG from 54 patients to build whole-brain model of neural activity (LFP for 4,244 elecs/pt). Parcellation into functional network modules using community detection - boundaries based on functional (not structural) organization.

7/ The resulting model provided a highly detailed representation of naturalistic brain activity across space and time, and its inherent organization into functional brain networks and allowed us to see patterns of rhythmic activity underlying depressed brain networks.

6/ To overcome this challenge, we utilized graph theory and a novel mathematical approach that allowed us to create a model of multiregional iEEG activity for each individual (SuperEEG: .@ContextMemLab, @LucyLongOwen, @PatrickMDaly). Code:

5/ Yet, despite high rates of depression in epilepsy (where iEEG recordings possible), population level analyses have not been performed. This is due to the heterogenous sampling of disease-relevant networks across subjects (electrodes are placed clinically).

4/ Intracranial EEG (iEEG) enables 1) direct sampling from brain regions (both cortical and subcortical) and 2) recordings over multi-day periods -- both major advantages over non-invasive methods.

3/ Our rationale is that quantitative biomarkers of depression are critical for advancing rational treatment design – but remain elusive. We think this is due to heterogeneity within depression and inadequate resolution of current measurements tools.

2/ Great collaborative effort between @ANKhambhati, @PatrickMDaly @LucyLongOwen, @ContextMemLab and others.

1/ Excited to share our latest study identifying neural network biomarkers for major depression from intracranial recordings! Novel way of understanding psychiatric conditions.

5/ Yet, despite high rates of depression in epilepsy (where iEEG recordings possible), population level analyses have not been performed. This is due to the heterogenous sampling of disease-relevant networks across subjects (electrodes are placed clinically).

4/ Intracranial EEG (iEEG) enables 1) direct sampling from brain regions (both cortical and subcortical) and 2) recordings over multi-day periods -- both major advantages over non-invasive methods.

3/ Our rationale is that quantitative biomarkers of depression are critical for advancing rational treatment design – but remain elusive. We think this is due to heterogeneity within depression and inadequate resolution of current measurements tools.

2/ Great collaborative effort between @ANKhambhati, @PatrickMDaly @LucyLongOwen, @ContextMemLab and others.