A brief review of some artificial intelligence methods in nephrology

Area of Science:

• Nephrology
• Artificial Intelligence
• Medical Informatics

Background:

• Artificial intelligence (AI) is increasingly utilized in nephrology research.
• Two prominent AI methods are machine vision (CNNs) and large language models (LLMs).
• Understanding their applications and limitations is crucial for nephrologists.

Purpose of the Study:

• To provide a non-technical overview of AI methods in nephrology.
• To explain the uses and limitations of CNNs and LLMs.
• To inform researchers about current AI tools in kidney disease studies.

Main Methods:

• Review of artificial intelligence applications in nephrology literature.
• Focus on machine vision using convolutional neural networks (CNNs).
• Analysis of large language models (LLMs) like ChatGPT.

Main Results:

• CNNs are established for segmenting and classifying kidney biopsy images, predicting outcomes.
• LLMs are newer, offering natural language interaction but exhibiting performance inconsistencies and hallucinations.
• LLMs may have a higher carbon footprint compared to CNN-based models.

Conclusions:

• CNNs offer robust tools for quantitative analysis of kidney pathology and prognostication.
• Current LLM chatbots in medicine require careful validation due to reliability concerns.
• Further research is needed to optimize LLM performance and assess their long-term impact in nephrology.