Feature extraction is the core aspect of pattern recognition. In the specific domain of NeuroImaging, researchers in the past have formulated uncountable ways to analyse the data and came up with thousands of inherent spatial and temporal features related to brain activities. However, these feature extraction methods are widely scattered in the web, making it arduous for researchers to focus on the problem statement.

Hctsa tool tried solving this problem by creating a zoo of features (>7000) extraction for time series in general. However, adapting hctsa can be computationally expensive and requires licensing to run, limiting widespread adoption for medical and research applications.

The idea here is to provide a single robust and portable open-source platform inclined to NeuroImaging domain and pipelined through catch22, wherein all the relevant features (distilled from large literature into a small subset with minimal loss in performance) are natively coded and wrapped up as a library which can be used as an automatic feature extractor modules for independent projects. This will also enable the usability of this system in time and memory-constrained environment.



Imran Alam


  • Joseph Lizier
  • benfulcher