Quantitative Image Feature Engine (QIFE): an Open-Source, Modular Engine for 3D Quantitative Feature Extraction from...

Area of Science:

• Medical Imaging
• Radiomics
• Computational Biology

Background:

• Radiomics enables quantitative analysis of medical images.
• Extracting radiomic features is crucial for building predictive models.
• Existing tools may lack modularity, integration capabilities, or efficient parallelization.

Purpose of the Study:

• To develop an open-source, modular system for 3D radiomics feature computation.
• To create a versatile tool deployable on various computer systems and integrable into existing workflows.
• To enhance the understanding of associations between image features and clinical data, such as patient survival.

Main Methods:

• Development of the Quantitative Image Feature Engine (QIFE) framework.
• Implementation of a modular design with swappable components for input, pre-processing, feature computation, and output.
• Exploitation of various levels of parallelization for multiprocessor systems.
• Benchmarking on a cohort of 108 lung tumor CT scans using different parallelization levels.

Main Results:

• The QIFE was successfully developed as an open-source MATLAB code and a Docker container.
• Computational efficiency was demonstrated, processing 108 tumors in 2:12 hours (1 core) and 1:04 hours (4 cores).
• The framework supports integration with existing segmentation and imaging workflows.

Conclusions:

• The Quantitative Image Feature Engine (QIFE) provides a flexible and efficient solution for 3D radiomics feature extraction.
• Its modularity and parallelization capabilities facilitate integration and improve computational performance.
• QIFE supports researchers in developing predictive models by linking image features with clinical data.