Advancing homebrew AI in diagnostic practice: opportunities and barriers^†

Area of Science:

• Pathology
• Medical Informatics
• Artificial Intelligence

Background:

• The Journal of Pathology recently published an argument for integrating homebrew artificial intelligence (AI) models in diagnostic pathology.
• This approach is presented as a method for democratizing digital diagnostics through local model development within pathology departments.

Purpose of the Study:

• To critically examine the real-world implications, practical limitations, and unmet needs associated with adopting homebrew AI in pathology practice.
• To outline the opportunities and challenges for the widespread implementation of homebrew AI in diagnostic pathology.

Main Methods:

• Commentary expanding on a previously published argument for homebrew AI in pathology.
• Critical examination of practical aspects and pathologist needs.

Main Results:

• Homebrew AI presents opportunities for advancing diagnostic pathology.
• Significant challenges exist, including the need for institutional backing, robust digital infrastructure, and sustained training efforts.

Conclusions:

• The successful integration of homebrew AI in pathology is contingent upon addressing practical limitations.
• Without adequate institutional support and infrastructure, the potential benefits of homebrew AI may not be realized.