Knowledge Based Versus Data Based: A Historical Perspective on a Continuum of Methodologies for Medical Image...

Area of Science:

• Medical image analysis
• Artificial Intelligence
• Computer-Aided Diagnosis

Background:

• Medical image analysis has evolved significantly, with recent emphasis on big data and deep learning.
• Historically, the field has utilized both knowledge-driven and data-driven methodologies.
• Traditional artificial intelligence (AI) approaches have a long-standing presence in medical imaging.

Purpose of the Study:

• To provide a historical review of medical image analysis methods.
• To position these methods along a knowledge-driven versus data-driven continuum.
• To underscore the relevance of traditional AI alongside deep learning.

Main Methods:

• Historical review of selected medical image analysis techniques.
• Categorization of methods based on a knowledge-driven vs. data-driven spectrum.
• Analysis of the interplay between different methodological paradigms.

Main Results:

• Medical image analysis encompasses a spectrum from purely knowledge-driven to fully data-driven methods.
• Traditional knowledge-based AI techniques retain historical importance and current utility.
• Data-driven techniques, including deep learning, represent a recent but significant advancement.

Conclusions:

• Both knowledge-driven and data-driven approaches are integral to medical image analysis.
• There is a complementarity between traditional AI and modern deep learning methods.
• Understanding this continuum is crucial for advancing the field of medical image analysis.