Kaapana/Neurodegenerative disease Nexus

Kaapana/Neurodegenerative disease Nexus

Quantitative measures of the brain, such as atrophy patterns related to neurodegenerative diseases, may provide complementary information to images for radiologists and clinicians. Recent advances in image analysis allow automated and rapid processing of medical image data, while new statistical methods enable collection of large harmonized reference datasets to perform comparative analyses of individual measures.  In order to translate such image analytics and machine learning methods to clinic, we have implemented a workflow on Kaapana, an open-source framework for medical data analysis, developed for translating laboratory-validated, quantitative measures into next-generation, precise biomarkers towards personalized diagnostics. The pipeline includes pretrained deep learning models for segmentation of brain anatomy, data harmonization to a large cognitively normal reference population, and calculation of machine-learning based individualized summary scores that quantify neurodegeneration and brain aging. The pipeline inputs raw DICOM images and generates segmentation maps and a patient-specific summary report to visualize measures in comparison to expected distributions based on reference populations.

automated pipeline

Key personnel:

  • Ashish Singh, PhD
  • Vikas Bommineni
  • Guray Erus, PhD
  • Ilya Nasrallah, PhD