AI-based carcinoma detection and classification using histopathological images: A systematic review

Area of Science:

• Digital pathology
• Computational oncology
• Medical image analysis

Background:

• Histopathological image analysis is crucial for cancer diagnosis, with manual evaluation being subjective and time-consuming.
• Carcinoma, a major cancer subtype, requires accurate diagnosis through microscopic examination of biopsy slides.
• Automated methods, particularly using artificial intelligence (AI), are increasingly explored to improve carcinoma detection and classification.

Purpose of the Study:

• To conduct a systematic literature review of state-of-the-art AI approaches for carcinoma diagnosis using histopathological images.
• To categorize and summarize existing methods based on carcinoma origin and AI techniques.
• To identify challenges, limitations, and future research directions in automated carcinoma diagnosis.

Main Methods:

• Systematic literature review of studies from major databases using strict inclusion/exclusion criteria.
• Categorization of selected articles based on carcinoma origin and AI methodologies.
• Summarization of AI methods, challenges, and future research directions.

Main Results:

• 101 articles were selected, with most studies using private datasets and reporting accuracies ranging from 63% to 100%.
• A significant trend towards using deep network models in AI-driven carcinoma diagnosis was observed.
• Varied image sizes and private datasets were common, indicating a lack of standardization.

Conclusions:

• There is a clear need for generalized AI-based systems for carcinoma diagnosis.
• Future research should focus on developing accountable AI approaches that mimic pathologists' evaluations across multiple magnifications.
• Standardization of datasets and methods is essential for advancing automated histopathological analysis.