Prognosis prediction of patients with disorders of consciousness based on digital twin brain models.
Shaoting Yan, Qing Li, Ruiqi Li and 8 others
PMID 41656193WHAT IT FOUND
An EEG-based brain model predicted six-month consciousness outcome in 121 patients better than baseline diagnosis and traditional EEG features.
It is not a treatment test and was not validated outside one coma centre.
Key findings
01In the final analysis of 121 patients, the EEG-based brain model had the best six-month prognosis prediction performance, with AUC 90.22%, accuracy 88.43%, sensitivity 84.31%, specificity 91.43%, and F-score 86.00%.
02The EEG-based brain model predicted six-month CRS-R scores with r=0.52, and absolute error was less than 5 in 74% of UWS patients and 70% of MCS patients.
03Positive outcomes were associated with higher excitatory gain in frontal and parietal lobes and higher modal controllability in frontal, parietal, and temporal lobes; negative outcomes were associated with higher inhibitory gain in frontal and parietal lobes.
STILL TO COME
How it was doneWhat they foundWhat it means for SLPs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The study was done in one coma awakening centre, and the authors say external validation in other settings is still needed. Performance was evaluated only by holding out each patient in turn within the same centre, not by a separate external test group. EEG was recorded with 32 channels, and individual head models were not available, which may reduce accuracy of inferred brain source activity. Of 169 initially enrolled patients, 30 were lost to follow-up and 17 were excluded for poor EEG quality, leaving 121 analysed. After the EEG, patients received different care, including multidisciplinary rehabilitation and some individualized amantadine, transcranial direct current stimulation, or transcranial magnetic stimulation, and post-discharge settings varied, so the contribution of any single treatment cannot be separated.
The easy way to misread this
Do not read the 90.22% AUC as proof this model can guide patient care now. It was tested only by holding out each patient in turn in one centre, and it did not test a treatment.
Summarised by AI from the full paper, without a clinician reviewing it. Check it against the source before it changes what you do. Read it on PubMed →