This study developed an artificial intelligence (AI)-based methodology for the quantitative analysis of single-cell morphological data in multiple myeloma (MM). The approach achieves high-precision AI-driven identification and segmentation of myeloma cells, nuclei, cytoplasm, and nucleoli, overcoming the inherent limitations of subjective traditional morphological analysis.

Furthermore, integrating this morphological quantification with cytogenetic abnormality analysis of myeloma cells provides an efficient predictive tool for identifying high-risk cytogenetic abnormalities.

Leveraging AI-guided selection of genetic testing targets, the research applied a rapid genetic abnormality detection technique utilizing first-drop bone marrow aspirate smears. This methodology achieves orders of magnitude improvements in testing cost, sample preprocessing time and detection sensitivity.

Multiple myeloma (MM) is a malignant plasma cell disorder exhibiting a rising global incidence associated with population aging. The significant intraclonal heterogeneity and complexity of clonal evolution inherent in MM genetics contribute to profound inter-individual heterogeneity. This complicates precise prognostic stratification and disease management, rendering MM largely incurable even in the era of novel therapies. In early-stage MM, primary cytogenetic abnormalities predominantly include odd-numbered chromosome trisomies and immunoglobulin heavy chain gene (IGH) rearrangements, such as t(11;14) and t(4;14). Disease progression and terminal stages are frequently characterized by the acquisition of increasingly complex copy number alterations (CNAs) (e.g., 1q gain/amplification [1q+] and 17p deletion [17p-]) and MYC rearrangements. These secondary events further drive cellular proliferation and tumor growth, collectively contributing to MM's high aggressiveness and therapeutic resistance.

Fluorescence in situ hybridization (FISH), particularly when combined with CD138 immunomagnetic bead sorting for plasma cell enrichment, plays a pivotal role in enhancing the detection rate of cytogenetic abnormalities in MM. However, a comprehensive MM genetic assessment necessitates the selection of multiple FISH probes targeting diverse genetic events, including several associated with high-risk progression. This requirement introduces significant complexity in probe selection. The Mayo Clinic utilizes an initial FISH screening panel for newly diagnosed MM targeting loci on chromosomes 1, 3, 7, 8, 9, 11, 13, 14, 15, and 17. Based on the results from the IGH break-apart probe on chromosome 14 (identifying an IGH break in approximately 50% of NDMM cases), positive cases undergo additional FISH testing for specific translocations: t(4;14), t(6;14), t(14;16), and t(14;20). Nevertheless, implementing such a comprehensive FISH approach is often hindered by the large number of required targets, limited plasma cell yields after bead enrichment/purification, and potential FISH assay failure.

While FISH is crucial for detecting MM cytogenetic abnormalities, current National Comprehensive Cancer Network (NCCN) guidelines highlight up to seven specific high-risk cytogenetic abnormalities associated with progression/relapse in NDMM: 1q21 gain (1q21+), 1p32 deletion (1p32-), del(17p)/monosomy 17 (-17)/TP53 mutation, t(4;14), t(14;16), t(14;20), and MYC rearrangement. The practical clinical application and accuracy of FISH are frequently limited by the extensive number of targets required for a complete assessment and the variability/complexity of standard FISH methodologies. An initial FISH screen must include an IGH break-apart probe. Subsequent testing with specific fusion probes for t(11;14), t(4;14), t(6;14), t(14;16), and t(14;20) is performed only if the IGH break-apart probe is positive.

Previous studies have reported on prognostic significance based on expert morphological classification of myeloma cells into plasmablastic, immature, and mature plasma cell types. However, utilizing myeloma cell morphology to predict specific FISH-detectable cytogenetic abnormalities remains unexplored. Artificial intelligence (AI) has demonstrated growing applications in medical image recognition and classification, including high-precision identification of leukemia, lymphoma, and MM cells. Recent research reported AI's ability to identify four specific acute leukemia molecular subtypes (PML::RARA fusion-positive; NPM1-mutated; CBFB::MYH11 fusion-positive without NPM1 mutation; RUNX1::RUNX1T1 fusion-positive) based on leukemia cell morphology in peripheral blood smears. Yet, analogous studies predicting cytogenetic abnormalities from MM cell morphology are currently lacking.

Inclusion Criteria:

• according to the criteria outlined in the Chinese guidelines for MM diagnosis and management (2024 edition)

Exclusion Criteria:

• non-MM

according to the criteria outlined in the Chinese guidelines for MM diagnosis and management (2024 edition)