Machine learning based detection and morphological analysis of dendritic spines

Encadrants : Suvadip Mukherjee
Disponible : OUI

Nombre d'étudiants : 1
Description :

Problem statementDendritic spines are small protrusions from a neuron’s dendrite which are responsible forreceiving signals from other axons. The structural appearance of a dendritic spine couldchange over time, and quantification of this morphological change could reveal informationabout key biological activities in the brain.

Travail demandé : The primary objective of this project would be to develop a machine learning based al-gorithm to automatically detect dendritic spines from two dimensional (maximum intensityprojection) fluorescence microscopy images of neurons. Since dendritic spines often appearas small protrusions from a primary dendrite, automated identification using traditionaltechniques is challenging. We would investigate recent advances in the deep learning com-munity such as the family of deep R-CNN’s [1, 2] to robustly identify the dendritic spines.Once the spines are identified to an acceptable confidence level, we would use image analysistechniques to extract relevant morphological features from the detected objects.

[1] Kaiming He, Georgia Gkioxari, Piotr Doll ́ar, and Ross Girshick. Mask r-cnn. In2017IEEE International Conference on Computer Vision (ICCV), pages 2980–2988. IEEE,2017.[2] Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster r-cnn: Towards real-time object detection with region proposal networks. InAdvances in neural informationprocessing systems (NIPS), pages 91–99, 2015.