Tenho um aluno que é professor em uma universidade inglesa. Ele e seus colegas agora aplicam exames orais, já que os alunos desandaram a entregar trabalhos produzidos por AI.
Saída interessante.
Glad to join the SIRS Global Representation Committee, connecting colleagues across regions to strengthen inclusion, visibility, and collaboration in schizophrenia research. Global perspectives are essential for a more equitable field. #SIRSGlobal#SchizophreniaResearch
AI should not replace clinical assessment.
But brief, low-cost clinical videos may add an objective behavioral layer to multimodal risk stratification in early psychosis.
Exploratory findings; external validation is needed.
#EarlyPsychosis
How can a clinical video become data?
We extracted 68 facial landmarks frame by frame with OpenFace 2.0, focusing on expressive regions: eyebrows, eyes, and mouth.
These dynamics were summarized with PCA.
#DigitalPhenotyping
The key signal was not in obvious facial expressivity.
Predictive information emerged from subtle facial-motor variation captured by higher-order PCA components — patterns hard to detect by visual inspection alone.
#Psychosis
Brief clinical videos may help predict later psychiatric outcomes in clinical high-risk individuals.
Best model: PCA-SO
F1=72%, sensitivity=66%, specificity=76%, balanced accuracy=71%.
Not “reading faces”: quantifying subtle facial dynamics.
#Psychosis
Can subtle facial dynamics extracted from brief clinical videos help predict later psychiatric outcomes in individuals at clinical high risk for psychosis?
In our new paper in Schizophrenia Research, we tested this question using machine learning and facial features.
One key finding: the predictive signal was not primarily in gross facial expressivity.
It emerged from higher-order PCA components, suggesting that clinically relevant information may lie in subtle facial-motor variation that is difficult to detect by visual inspection alone.
Clima de saudosismo após este jogo melancólico de estréia do Brasil na Copa de 2026. Muitos vídeos relembrando os “4 erres” (sem o dizer) da seleção campeã de 2002. Mas genialidade não escolhe geração, ocorre ao acaso. #selecaobrasileira#braxmar#copa2026
What can biomarkers teach us about compulsive sexual behavior disorder — without reducing suffering to biology alone?
This was one of the key questions behind Bianca de Oliveira Ferreira’s PhD thesis, defended at FMUSP.
It was an honor to join the examining committee and discuss a thesis that connects psychiatry, neuroscience, mental health, and stigma reduction.
Congratulations to Bianca for an important contribution.