EASL: A Framework for Designing, Implementing, and Evaluating ML Solutions in Clinical Healthcare Settings.

Area of Science:

• Clinical Informatics
• Machine Learning in Healthcare
• Medical Imaging Analysis

Background:

• Clinical machine learning (ML) tools require integrated development and evaluation environments.
• Existing solutions often lack a holistic approach, hindering the transition from development to clinical application.

Purpose of the Study:

• Introduce the Explainable Analytical Systems Lab (EASL) framework.
• Provide a versatile, end-to-end solution for clinical ML tool lifecycle management.
• Demonstrate EASL's utility through a deep learning classifier case study for medical image diagnosis.

Main Methods:

• The EASL framework comprises three modules: Workbench (data management, model development), Canvas (imaging viewer, UI development), and Studio (model hosting, analytics, user studies).
• A deep learning classifier for medical image diagnosis was developed and evaluated using the EASL framework.
• The framework integrates data management, ML development, visualization, hosting, and analytics.

Main Results:

• The EASL framework successfully facilitated the development and evaluation of a clinical ML tool.
• The case study demonstrated the framework's capability in designing and assessing a deep learning diagnostic classifier.
• EASL promotes a holistic approach, ensuring clinical ML tools are both effective and reliable.

Conclusions:

• The EASL framework offers a comprehensive solution for developing and evaluating clinical ML tools.
• It simplifies the process of creating reliable and effective ML applications in healthcare settings.
• EASL advances the understanding and application of ML in clinical practice.