Applying-dimensionality-reducion-techniques-for-measured-caridac-blood-flow-

About our Dataset :

• we have patient-specific blood flow information measured using four-dimensional phase-contrast magnetic resonance imaging, also known as 4D-PC-MRI i.e a 3D blood flow information.
• 197 attributes and 122 rows, each row represents characteristics of a patient
• Attributes can be divided into 5 types
  • Pressure related attributes
  • Velocity related attributes
  • Diameter related attributes
  • Vortex related attributes
  • Flow jet related attributes

Task :

• Cluster patients into 3 cohorts
  • Healthy patients
  • Patients suffering from Bicuspid aortic valve disease
  • Tetralogy of fallot.

Methods used :

1. Principal component analysis
2. T-distributed stochastic neighboring embedding
3. Uniform manifold approximation and projection
4. Multi dimensional scaling

Evaluation :

• Qualitative Evaluation :
  • Silhouette coefficient
  • Correlation coefficient
  • Connectivity validity measure
• Quantitative:
  • User study on 40 samples

additionally we have implemented our pre trained parameters (parameters of dimensionality techniques used on blood flow dataset) on a mice-protietn dataset, to check the cohort seperability