Springe direkt zu Inhalt

AI for Health

Thema: AI for Health, 2 Projekte, s.u.

DozentIn(en): Prof. Roland Eils, Julius Upmeier zu Belzen, Thore Buergel

Maximale Teilnehmerzahl: 6, je 3 Teilnehmer*innen pro Projekt

Zeitraum/Vorbesprechungstermin: 01.03.2020 – 15.04.2020

Ort: Kapelle-Ufer 2, 10117 Berlin

Kurze inhaltliche Beschreibung:

Projekt A: Learning Patient representations for biomarker identification in Delir prevention

  • Apply and develop machine learning methods to obtain patient representations using multi-modal data sources
  • Apply the learned representations to identify patients at high risk of experiencing postoperative Delir
  • Statistically analyse RNA-Seq, Demographical, Clinical, Metabolomic and Genetic data

Projekt B: Deciphering Angiography – Learning Latent medical features using Autoencoders

  • Apply variational autoencoder (or similar) setups on angiography images based on existing literature
  • Develop data pre-processing and model optimization pipelines
  • Analysis of the latent spaces of the aforementioned variational autoencoder for generative purposes and medical insights (degree of artery obstruction, shape of arteries, flow, etc.)

Quantitative Aufteilung: (in %)

Praktische Programmierarbeit: 2/3
Soft Skills: 1/3

Verwendete Programmiersprache(n): Python, optional R

Schwierigkeitsgrad (Acht Sterne verteilt auf drei Bereiche):

A Programmieren ****
B Biologie/Chemie *
C Projektmanagement ***

Erforderliche Vorkenntnisse:

  • Experience with the Python programming language
  • Preferably understanding of neural networks and variational autoencoders specifically
  • Preferably prior experience with PyTorch or other DL-Libraries

Kontaktadresse, Webseite/Link:

bianca.hennig@charite.de

Präsentation:
https://charitede-my.sharepoint.com/:b:/g/personal/thore_buergel_charite_de/EdPJtOyzxnBOhJPNc1f_yS8BF7jj00yip1846qMGIWaUiQ?e=FoKacy

eVV WiSe 20/21