Skip to main content

Conference Program

Subpage Hero

Loading

Enhancing Interoperability to Enable Broader Adoption of Artificial Intelligence in Chromosome Analysis and Karyotyping: A Pilot Evaluation

Laboratory Genetics and Genomics
  • Primary Categories:
    • Clinical Genetics
  • Secondary Categories:
    • Clinical Genetics
Introduction:
Interoperability plays a growing role in image analysis, facilitating the adoption of advanced image processing software. This is particularly relevant for cytogenetics laboratories that wish to integrate artificial intelligence (AI) in their chromosome analysis and karyotyping process without needing to replace their existing digital slide scanner systems. The objective of this pilot evaluation was to assess the feasibility of converting and importing images obtained from a commercially available digital slide scanner system for analysis using a third-party software. Additionally, this initial evaluation aimed to identify necessary modifications to optimize "out-of-the-box" AI-powered karyotyping models for analyzing the imported images.

Methods:
G-banded bone marrow specimens were processed and stained following standard laboratory protocols. The slides were scanned using the CytoVision imaging system (Leica Biomedical), and the resulting images were converted and imported into the GenASIs platform (Applied Spectral Imaging, ASI) through an automated, unsupervised process. The ASI AI-based computer-aided analysis software had not been trained on metaphase images acquired from third-party scanners. The AI analysis algorithms were deployed in their default, unmodified configuration trained on ASI acquired metaphase images. During the import process, karyograms were automatically generated, ready for analysis using ASI’s HiBand analysis software. The number of imported cells and the processing time were recorded for each case, and a preliminary assessment of the AI algorithms utilized in their standard configuration was conducted by cytogeneticists.

Results:
A total of 268 cells from eight G-banded slides prepared from two bone marrow specimens were included in this pilot evaluation. All cells were successfully converted, imported, and processed, with an average import and processing time of 8.7 seconds per cell. A preliminary karyogram, ready for review, was automatically generated during the importing procedure for 244 out of the 268 imported cells (91%). Cytogeneticists selected 20 cells per case for review and analysis. These cells were initially analyzed using the laboratory-installed software, yielding an average analysis time of 1.9±0.7 minutes per metaphase (ranging from 53 to 255 seconds per cell). The same cells were subsequently analyzed using ASI’s AI computer-aided software, with analysis times ranging from 22 to 213 seconds (p>0.05, paired t-test). Following segmentation corrections, 0 or 1 placement error was reported in 91% of the metaphases karyotyped using the AI classification models in their standard configuration. However, the AI segmentation models were notably affected by the specific enhancement schemes and image properties of the imported scans, and will require optimization.

Conclusion:
This pilot evaluation demonstrates the feasibility of converting, importing, and processing metaphase images from a commercially available scanning system for analysis using third-party software. While AI models, particularly for chromosome segmentation, require further optimization to adapt to the characteristics of the imported images, this initial study highlights the potential of a new automated import and processing utility. This preliminary assessment illustrates how interoperability could facilitate the adoption of AI-based chromosome analysis and karyotyping software in cytogenetics laboratories without necessitating the replacement of costly image acquisition systems.

Agenda

Sponsors