Artificial intelligence-augmented histopathologic review using image analysis to optimize DNA yield from...

Area of Science:

• Digital Pathology
• Artificial Intelligence in Medicine
• Genomic Sequencing

Background:

• Pathologists currently estimate tissue requirements for DNA extraction visually, leading to potential tissue waste and increased costs.
• Accurate DNA input is crucial for the success of next-generation sequencing (NGS).

Purpose of the Study:

• To develop and validate an artificial intelligence (AI)-augmented system (SmartPath) for quantitative assessment of tissue for DNA extraction.
• To improve the accuracy of determining tissue extraction parameters, thereby reducing waste and costs.

Main Methods:

• SmartPath utilizes deep learning (U-Net and convolutional networks) for cell and tumor segmentation on digitized H&E-stained slides.
• Quantitative features are extracted to predict DNA yield per slide using a regularized linear model.
• The system calculates the number of slides needed for scraping based on pathologist-defined target yields.

Main Results:

• The SmartPath cohort achieved 25% more DNA yields within the target range (100-2000 ng) compared to traditional review.
• The number of extraction attempts remained statistically unchanged between the AI-augmented and traditional review groups.
• SmartPath optimized slide recommendations, saving tissue for large sections and preventing re-extractions for scant or degraded DNA samples.

Conclusions:

• AI-augmented histopathologic review, exemplified by SmartPath, can significantly decrease tissue waste, sequencing time, and laboratory costs.
• SmartPath provides quantitative metrics to guide pathologists, optimizing DNA yields, particularly for challenging samples.
• This technology has the potential to enhance efficiency and reduce costs in molecular sequencing laboratories.