7. AI-Based Algorithms for neoplastic metaphase cells boost efficiencies in the cytogenetics laboratory

• Published in: Cancer genetics Vol. 278-279; pp. 2 - 3
• Main Authors: Hong, Bo, Fedderson, Brian, Zhao, Jian, Andersen, Erica
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01.11.2023
• ISSN: 2210-7762
• DOI: 10.1016/j.cancergen.2023.08.015

Traditionally, clinical cytogenetics laboratories analyzing digitized images of metaphase spreads have relied upon manual keystrokes and mouse clicks within the software to separate and classify individual chromosomes for review. These are time-consuming, error-prone tasks that add labor costs and risks to quality. We implemented artificial intelligence (AI)-assisted digital image analysis of metaphase chromosomes in our laboratory for hematologic oncology cases to improve overall turnaround time and quality. A comparison dataset collected prior to -and post-implementation is summarized. Metaphase cells from a total of 2000 images were included in the comparison study. Using the manufacturer's deep neural networks (DNN)-based algorithms, we compared 1000 images of newly processed oncology cases to 1000 images processed prior to implementation of the DNN software. Data collected included total processing steps, karyogram preparation time, and error rates. The results showed a 76% reduction of processing steps and 52% karyogram preparation time reduction. In addition, the laboratory's routine analysis error rate decreased from 0.15% to 0.03% after the implementation of AI software, showing marked improvement of 0.12%. Overall, our comparison study demonstrates use of AI-based algorithms can significantly improve time savings and accuracy in the clinical cytogenetics laboratory setting. By automating routine tasks, technologists can devote more time to higher scope activities such as banding analysis, resulting in better distribution of time management in the cytogenetics laboratory. These findings have important implications for the future of clinical laboratory practice, where AI is expected to play an increasingly important role in improving workflows and ultimately, patient care.