Deep Learning-Based Differential Diagnosis of Major Depression and Bipolar Disorder Using Microglia-Cellular Sensors and Patient-Derived Small Extracellular Vesicles

This work presents a deep learning–based diagnostic framework that employs microglial cells as biosensors to distinguish between major depressive disorder (MDD), bipolar disorder (BD), and healthy controls. Patient-derived plasma small extracellular vesicles (sEVs) are applied to microglial cultures, and the resulting morphological changes are captured via fluorescence microscopy. Individual cell images are organized into M×M arrays and processed through a DenseNet121 CNN. Classification robustness is enhanced by generating multiple arrays per subject using random cell permutations and affine transformations, with final diagnoses assigned via weighted voting. Evaluated under repeated subject-disjoint cross-validation, the best model correctly classified 44 out of 45 individuals, supporting the proof-of-concept for a novel differential diagnosis platform for mood disorders.

How to cite

@article{zambrano2026mddbd,
  author    = {Zambrano, Jorge and Luarte, Alejandro and Contreras, Julian and Perez, Juan P. and Yant{\'e}n-Fuentes, Liliana and Prieto, Miguel and Lazcano, Pablo and Wyneken, Ursula and Perez, Claudio A.},
  journal   = {Scientific Reports},
  title     = {Deep Learning-Based Differential Diagnosis of Major Depression and Bipolar Disorder Using Microglia-Cellular Sensors and Patient-Derived Small Extracellular Vesicles},
  year      = {2026},
  volume    = {16},
  number    = {1},
  pages     = {11679},
  doi       = {10.1038/s41598-026-47476-9}
}

Recommended citation: J. E. Zambrano, A. Luarte, J. Contreras, J. P. Perez, L. Yantén-Fuentes, M. Prieto, P. Lazcano, U. Wyneken and C. A. Perez, "Deep Learning-Based Differential Diagnosis of Major Depression and Bipolar Disorder Using Microglia-Cellular Sensors and Patient-Derived Small Extracellular Vesicles," in Scientific Reports, vol. 16, no. 1, p. 11679, 2026, doi: 10.1038/s41598-026-47476-9.
Download Paper | Download Bibtex