Timm, Michael

Abstract Background: Flow cytometry (FC) is widely utilized for the identification of measurable residual disease (MRD) in multiple myeloma (MM), providing an essential laboratory indicator for prognostic assessment and therapeutic management. Unfortunately, the high level of expertise required, long manual analysis time, and complexity of FC data files currently limit the offering of this testing to selected reference laboratories. Methods: We developed an artificial intelligence (AI) pipeline to automatically enhance unaltered FC files, minimizing the expertise and manual analysis time needed to detect MM MRD using the 2-tube Euroflow Consortium Panel (Cytognos). Raw FC files corresponding to 150 MRD-positive bone marrow samples (MRD < 5%, median = 0.0022%), and 39 MRD-negative samples were processed through our CCADDAS (Clustering and Classification of All events, Dimensionality reduction, Downsampling and Aberrancy Scaling) pipeline, deployed on a custom clinical-grade cloud environment (Biolegend). Automated processing steps included elimination of acquisition errors (FlowCut), state-of-the-art clustering (PARC), cluster-informed and density-based downsampling with preservation of low-event subsets, dimensionality reduction (UMAP), and cluster-based anomaly detection compared to negative controls (29 bone marrow aspirates). In addition, a deep neural network (DNN) trained on expert-defined subpopulations from negative controls was included for automatic gating of normal subsets (Tensorflow v2). AI-enhanced data files were analyzed using a general-purpose flow cytometry analysis software (Kaluza v 2.3, Beckman Coulter), and %MRD estimates were compared to reported results based on traditional expert analysis of original FC files. Results: Cluster-informed downsampling reduced the number of cells to be manually analyzed from 4.1 million to 196,761 cells per tube, on average (95% cellularity reduction); resulting in a smaller FC data file (from 203.2 MB to 20.5 MB; 90% data reduction). Importantly, low-level MRD events were adequately preserved after downsampling (median 88% and 77% retention, for 0.01-0.001% MRD and < 0.001% MRD, respectively). True number of events were accurately estimated on gated subsets using an “upsampling factor” parameter [observed # events x mean (upsampling factor)]. Gating of normal subsets was completely automated using a DNN classifier parameter. In 96% of positive cases, MRD was rapidly identified as immunoglobulin light chain-restricted plasma cells on a single pre-calculated UMAP plot (tube 2), after dual coloring for kappa and lambda. On the remainder 4% of positive cases, MRD was easily segregated from benign plasma cells using the AI-generated “aberrancy scale” parameter or other strategies. The use of AI-enhanced files reduced manual analysis time from 15.05 minutes (SD = 6.84) to 1.29 minutes (SD = 0.90) per case, on average (p < 0.0001) (91% reduction of manual analysis time). MM MRD was detected in all positive cases above limit of detection for our assay (≥0.0002%), with excellent quantitative correlation with conventional analysis on original FC files (linear regression slope = 0.9841, P < 0.0001). Conclusion: We introduce a largely unsupervised AI pipeline that transforms raw MM MRD FC data from the Euroflow Consortium panel into a markedly smaller, AI-annotated and software-agnostic FC file, including comparison to normal controls. This CCADDAS pipeline simplifies and accelerates detection of MM MRD in clinical diagnostics, reducing the number of cells analyzed by 95% and manual analysis time by 91%, without impacting test performance. Moreover, CCADDAS can be utilized for any laboratory-developed MM MRD assay, and its small-sized export is compatible with any clinical FC software, computing platform, and analysis strategy. Adoption of CCADDAS is likely to facilitate the implementation of MM MRD FC analysis by more clinical laboratories.

Abstract Measurable residual disease (MRD) assessment by flow cytometry (FC) plays an essential role in prognosis and therapy escalation of B-cell acute lymphoblastic leukemia (B-ALL). However, the high degree of expertise and manual analysis time required limits the availability of this assay. To overcome this limitation, we developed a data-enhancing artificial intelligence (AI) pipeline that accelerates and simplifies MRD analysis. Unaltered FC files from 171 B-ALL MRD-positive and 89 MRD-negative cases were processed through an AI pipeline trained with 31 expert-gated negative controls. Cluster-informed downsampling reduced FC files from 1.2 million to 155 884 cells per case, on average, (87% cellularity reduction), whereas preserving small MRD populations (median, 100% retention for MRD of <1%) and allowing for true percentage MRD estimates using a correction factor. A deep neural network cell classifier automatically identified normal hematopoietic subsets (macro-averaged F1 score of 0.86); and an AI measure of anomaly discriminated B-ALL from benign mononuclear (area under the curve [AUC] of 0.98) or B-lymphoid cells (AUC of 0.94). Manual analysis of AI-enhanced files was completed in only 1.01 minutes per case, on average (standard deviation of ±0.57); with 100% positive agreement with conventional analysis (for MRD of ≥0.01%), 100% negative agreement, and excellent quantitative correlation (R2 = 0.92). Our cloud-based AI-enhancement solution accelerates B-ALL MRD identification without compromising test performance and has the potential of facilitating B-ALL MRD analysis by more clinical laboratories.

Background: Flow cytometric detection of minimal residual disease (MRD) in chronic lymphocytic leukemia (CLL) is complex, time-consuming, and subject to inter-operator variability. Deep neural networks (DNNs) offer potential for standardization and efficiency improvement, but require rigorous validation and monitoring for safe clinical implementation. Methods: We evaluated a DNN-assisted human-in-the-loop approach for CLL MRD detection. Initial validation included method comparison against manual analysis (n = 240), precision studies, and analytical sensitivity verification. Post-implementation monitoring comprised four components: daily electronic quality control, input data drift detection, error analysis, and attribute acceptance sampling. Laboratory efficiency was assessed through a timing study of 161 cases analyzed by five technologists. Results: Method comparison demonstrated 97.5% concordance with manual analysis for qualitative classification (sensitivity 100%, specificity 95%) and excellent correlation for quantitative assessment (r = 0.99, Deming slope = 0.99). Precision studies confirmed high repeatability and within-laboratory precision across multiple operators. Analytical sensitivity was verified at 0.002% MRD. Post-implementation monitoring identified 2.97% of cases (26/874) with input data drift, primarily high-burden CLL and non-CLL neoplasms. Error analysis showed the DNN alone achieved 97% sensitivity compared to human-in-the-loop-reviewed results, with 13 missed cases (1.5%) showing atypical immunophenotypes. Attribute acceptance sampling confirmed 98.8% of reported negative cases were true negatives. The DNN-assisted workflow reduced average analysis time by 60.3% compared to manual analysis (4.2 ± 2.3 vs. 10.5 ± 5.8 min). Conclusions: The implementation of a DNN-assisted approach for CLL MRD detection in a clinical laboratory provides diagnostic performance equivalent to expert manual analysis while substantially reducing analysis time. Comprehensive performance monitoring ensures ongoing safety and effectiveness in routine clinical practice. This approach provides a model for responsible AI integration in clinical laboratories, balancing automation benefits with expert oversight.

Background/Objectives: Acute myeloid leukemia (AML) is a biologically heterogeneous malignancy with a variable prognosis. Despite many patients achieving complete remission, relapse remains common, underscoring the need for effective prognostic markers. Measurable residual disease (MRD) has emerged as a critical prognostic indicator, associated with higher relapse risk and shorter survival. This study reports on our initial experience of MRD detection by flow cytometry in 385 bone marrow samples from 126 AML patients. Methods: The flow cytometry MRD assay, validated according to stringent consensus recommendations, consists of a 3-tube, 10-color panel incorporating a broad spectrum of lineage differentiation markers. Analytical specificity, sensitivity, precision, and reproducibility were evaluated, demonstrating the assay’s robustness. Results: The results reveal distinct immunophenotypic aberrancies in all AML cases, with consistent identification of aberrant immunophenotypes in follow-up specimens. AML MRD was detected in 32 out of 126 patients (25%) and in 77 out of 385 analyses (20%), with a median aberrant blast percentage of 1.87% (range, 0.01–12). A change in immunophenotype was documented in 21% of the MRD-positive cases. MRD positivity detected in the first sample studied was associated with reduced overall survival (HR: 5.153; p < 0.0001). Conclusions: Our findings support the integration of flow cytometric MRD analysis into routine clinical practice to enhance risk stratification and treatment planning for AML patients, as currently recommended by professional guidelines.

Background: Flow cytometry (FC) is widely utilized for the identification of minimal residual disease (MRD) in B acute lymphoblastic leukemia (BALL), providing an essential laboratory indicator for prognostic assessment and therapeutic management. Unfortunately, the high level of expertise required, long manual analysis time, and complexity of FC data files limit the offering of this testing to selected reference laboratories. Methods: We developed an artificial intelligence (AI) pipeline to automatically enhance unaltered FC files, minimizing the expertise and manual analysis time needed to detect BALL-MRD on a single-tube 10-color panel (CD10, CD19, CD20, CD22, CD24, CD34, CD38, CD45, CD58, CD66c). Raw FC files corresponding to peripheral blood or bone marrow specimens from 171 BALL MRD positive cases (MRD &lt; 5%, median = 0.11%, 107 patients), and 89 MRD-negative cases (63 patients) were processed through our CCADDAS (Clustering and Classification of All events, Dimensionality reduction, Downsampling and Aberrancy Scaling) pipeline on a cloud environment (Google Vertex AI). Automated processing steps included elimination of acquisition errors (FlowCut), state-of-the-art clustering (PARC), dimensionality reduction (UMAP), cluster-based anomaly detection compared to negative controls (24 bone marrow aspirates and 7 peripheral bloods), and cluster-informed downsampling with preservation of low-event subsets. In addition, a deep neural network (DNN) trained on expert-defined subpopulations from negative controls was included for automatic gating of normal subsets (Tensorflow v2). AI-enhanced data files were analyzed using a general purpose flow cytometry analysis software (Kaluza v 3.5, Beckman Coulter), and %MRD estimates compared to reported results based on expert analysis of original FC files. Results: Cluster-informed downsampling reduced the number of cells per case to be manually analyzed from 1.2 million to 155,884 cells on average (87% cellularity reduction); resulting in a smaller FC data file (from 64.5 MB to 13.7 MB; 79% data reduction). Importantly, low-level MRD events were adequately preserved after downsampling (median 100% retention for both MRD&lt;0.01% and MRD 0.01-1%). True number of events were accurately estimated on gated subsets using an “upsampling factor” parameter [observed # events x mean (upsampling factor)]. Gating of normal subsets was completely automated using a DNN classifier parameter. Identification of MRD was aided by an AI-generated “aberrancy scale” parameter that discriminated BALL (median: 991) from both benign mononuclear cells (median: 486, p &lt; 0.0001) and background B-lymphoid cells (median: 379, p &lt; 0.0001), with a performance (AUC: 0.98 and 0.94 for mononuclear and B-lymphoid cells, respectively) superior to CD20 (0.54 and 0.69), CD38 (0.60 and 0.81), CD45 (0.75 and 0.74), CD58 (0.52 and 0.89), CD66c (0.56 and 0.72) and all other antigens studied. The use of AI-enhanced files reduced manual analysis time from 14.16 minutes (SD = 3.7) to 1.05 minutes (SD = 0.6) per case, on average (p &lt; 0.0001) (93% reduction). BALL MRD was detected in all positive cases above clinically relevant threshold (≥0.01%, as recommended by NCCN guidelines), and in 28 of 32 (88%) cases below this threshold, with excellent quantitative correlation with conventional analysis on original FC files (linear regression slope = 1.18, intercept = 0.04%, R squared = 0.92, P &lt; 0.0001). Conclusion: We introduce a largely unsupervised AI pipeline that transforms raw BALL-MRD FC data into a markedly smaller, AI-annotated and software-agnostic FC file, including comparison to normal controls. This CCADDAS pipeline simplifies and accelerates detection of BALL-MRD in clinical diagnostics, reducing number of cells analyzed by 87% and manual analysis time by 93%, without impacting test performance. Moreover, CCADDAS can be utilized for any labroatory-developed BALL-MRD assay, and its small-sized export is compatible with any clinical FC software, computing platform and analysis strategy. Adoption of CCADDAS is likely to facilitate the implementation of BALL-MRD FC analysis by more clinical laboratories.

Background: Multiple myeloma (MM) involves the proliferation of monoclonal plasma cells (PCs) in the bone marrow. The plasma cell labeling index (PCLI), indicating the percentage of PCs in the S-phase, is a key prognostic factor in newly diagnosed multiple myeloma (NDMM) patients. Methods: A retrospective study at Mayo Clinic analyzed bone marrow Affymetrix microarray datasets from 136 NDMM patients (2001-2012). Data normalization and quality assessment were followed by differential gene expression analysis using limma-voom, adjusted for age, BMPC percentage, and disease stage (RISS). Selected genes were dichotomized by median and subjected to logistic regression for high/low PCLI. Data from 594 patients with a Proliferation Index at baseline was pooled from the MMRF (CoMMpass) dataset for validation after normalization across assays. Results: The baseline characteristics of the patients are as follows: 62 (46%) were female and 74 (54%) were male, with a mean age at diagnosis of 61.6 ± 11.5 years. RISS staging included 36 patients (26.5%) at stage 1, 32 (23.5%) at stage 2, 29 (21.32%) at stage 3, and 39 patients (28.67%) with unknown stage. PCLI distribution showed 113 patients (83.7%) with &lt;2%, and 23 (16.3%) ≥2%. Differential Analysis: The differential analyses reveal several critical insights. In the primary continuous differential analysis, 21,367 genes were analyzed, identifying 654 as differentially expressed with an FDR-adjusted p-value ≤0.05. Of these, 481 genes were upregulated (top 5: SPC25, TRAT1, PBK, E2F8, ASPM) and 173 were downregulated (top 5: SPARC, CD24, MYOF, CXCL12, VCAM1). Univariate and multivariate stratified analyses showed significant findings: 668 genes in age stratification, 395 genes in BMPC% stratification, and 536 genes in RISS stratification. In multivariate analysis, 408 genes were significant with an FDR-adjusted p-value &lt;0.05 for PCLI and ≥0.05 for BMPC%, age, and RISS. 3-Fold Internal Validation: 240 significant genes were identified in fold 1, 25 in fold 2, and 167 in fold 3, with 21 genes common across all folds (e.g., MCM10, GTSE1, UBE2S, NCAPG, FOXM1, STMN1, NEK2, CENPF, MND1, HJURP, E2F8, TOP2A, SKA1, CCNB2, CCNF, CHAF1A, PBK, EXO1, AURKB, CDCA2, DHFR). 19 genes were common in the sensitivity analysis by excluding RISS 3. 20 genes were common in categorical analysis (Categories &lt;1%, 1-2%, 2-3%, 3-4%, &gt;4%). 3 Gene-Based Proliferation Score (3GPS): On univariate logistic regression analysis of 21 genes for high/low PCLI, 17 significant genes were identified. In the step-wise multivariate analysis, CDCA2 (OR=11.56), DHFR (OR=9.6), and CENPF (OR=6.11) were significant. Log-odds were used as weights. For the high expression(median and above) of gene as 1 and low expression as 0: 3GPS=2.4*CDCA2+ 2.3*DHFR +1.8*CENPF A cut-off of 2.4 was chosen as CDCA2 had the highest positive concordance rate (90.91%) versus DHFR (86.36%) and CENPF (77.27%). External Validation: In the CoMMpass dataset, the sensitivity and specificity for the presence of high expression of any of the 3 genes with ≥2 Proliferation Index (PI) were 92.4% and 73.8%, respectively (AUC=83.1%). The median 3GPS was 0 for &lt;2 PI (n=238) and 6.5 for ≥2 PI (n=356). The sensitivity and specificity for the 3GPS in predicting ≥2 PI were 82.6% and 93.6%, respectively (AUC=88.1%). Pathway Analysis: Significant pathways using Reactome include polo-like kinase mediated events in G2/M transition, condensation of prometaphase chromosomes in metaphase, resolution of sister chromatid cohesion in anaphase, and APC-Cdc20 mediated degradation of Nek2A for regulation of the cell cycle. p53 signaling, cancer pathways, and microtubule regulation were also mapped. Survival Analysis: In our cohort, for CDCA2 quartiles (Q)3 and Q4, for CENPF Q3 and Q4, and for DHFR Q4 were associated with inferior overall survival. 3GPS ≥2.4 was associated with inferior survival of 60.28 months versus 111.5 months for &lt;2.4 (p&lt;0.01). In the validation dataset, median survival was not reached at 100 months (vs. high 79.7, p&lt;0.01). The multivariate Cox model found 3GPS score ≥2.4 as an independent predictor (HR=1.64, 95% CI: 1.23 - 2.19), along with ECOG status, ISS stage, gender, age, and whether the patient underwent Stem Cell Transplant. Conclusion: CDCA2, DHFR, and CENPF are associated with ≥ 2% PCLI. The 3GPS score can serve as a measure of gene-based proliferation index. Prospective validation is needed for clinical practice.

Background: Flow cytometry (FC) is widely utilized for the identification of minimal residual disease (MRD) in chronic lymphocytic leukemia (CLL), providing an essential laboratory indicator for prognostic assessment and therapeutic management. Unfortunately, the high level of expertise required, long manual analysis time, and complexity of FC data files limit the offering of this testing to selected reference laboratories. Methods: We developed an artificial intelligence (AI) pipeline to automatically enhance unaltered FC files, minimizing the expertise and manual analysis time needed to detect CLL MRD on a single-tube 10-color panel (CD5, CD19, CD20, CD22, CD38, CD43, CD45, CD200, kappa and lambda). Raw FC files corresponding to peripheral blood or bone marrow specimens from 166 CLL MRD positive cases (MRD &lt; 5%, median = 0.17%, 122 patients), and 61 MRD-negative cases (45 patients) were processed through our CCADDAS (Clustering and Classification of All events, Dimensionality reduction, Downsampling and Aberrancy Scaling) pipeline on a cloud environment (Google Vertex AI). Automated processing steps included elimination of acquisition errors (FlowCut), state-of-the-art clustering (PARC), dimensionality reduction (UMAP), cluster-based anomaly detection compared to negative controls (15 bone marrow aspirates and 14 peripheral bloods), and cluster-informed downsampling with preservation of low-event subsets. In addition, a deep neural network (DNN) trained on expert-defined subpopulations from negative controls was included for automatic gating of normal subsets (Tensorflow v2). AI-enhanced data files were analyzed using a general purpose flow cytometry analysis software (Kaluza v 3.5, Beckman Coulter), and %MRD estimates compared to reported results based on expert analysis of original FC files. Results: Cluster-informed downsampling reduced the number of cells per case to be manually analyzed from 1.1 million to 165,010 cells on average (85% cellularity reduction); resulting in a smaller FC data file (from 62.1 MB to 15.2 MB; 75% data reduction). Importantly, low-level MRD events were adequately preserved after downsampling (median 100% retention for both MRD&lt;0.01% and MRD 0.01-1%). True number of events were accurately estimated on gated subsets using an “upsampling factor” parameter [observed # events x mean (upsampling factor)]. Gating of normal subsets was completely automated using a DNN classifier parameter. Rapid identification of suspected MRD was aided by an AI-generated “aberrancy scale” parameter that discriminated CLL MRD (mean ± 2SD: 969 ± 144) from background B-lymphoid cells (mean ± 2SD: 395 ± 401) (p&lt;0.0001) (AUC = 0.98), with a performance superior to CD5 (AUC = 0.93), CD20 (AUC = 0.83), CD43 (AUC = 0.8) and CD200 (AUC = 0.7). Immunoglobulin light chain-positive CLL MRD was visually identified on a pre-calculated UMAP plot as a single-colored cluster, after dual coloring for kappa and lambda. The use of AI-enhanced files reduced manual analysis time from 11.6 minutes (SD = 4.5) to 0.8 minutes (SD = 0.4) per case, on average (p &lt; 0.01) (93% reduction). CLL MRD was detected in all positive cases above clinically relevant threshold (≥0.01%, as recommended by NCCN guidelines), and in 17 of 21 (81%) cases below this threshold, with excellent quantitative correlation with conventional analysis on original FC files (linear regression slope = 1.02, intercept = 0.001%, R squared = 0.98, P &lt; 0.0001). Conclusion: We introduce a largely unsupervised AI pipeline that transforms raw CLL-MRD FC data into a markedly smaller, AI-annotated and software-agnostic FC file, including comparison to normal controls. This CCADDAS pipeline simplifies and accelerates detection of CLL-MRD in clinical diagnostics, reducing number of cells analyzed by 85% and manual analysis time by 93%, without impacting test performance. Moreover, CCADDAS can be utilized for any laboratory-developed CLL-MRD assay, and its small-sized export is compatible with any clinical FC software, computing platform and analysis strategy. Adoption of CCADDAS is likely to facilitate the implementation of CLL MRD FC analysis by more clinical laboratories.

Background: Identification of minimal residual disease (MRD) in acute myeloid leukemia (AML) is essential for assessing risk of relapse and guiding therapeutic management. While molecular assays have streamlined AML MRD detection, flow cytometric immunophenotyping (FCI) remains the only available test modality for approximately 50% of patients lacking traceable molecular MRD targets. Unfortunately, due to the high level of expertise required, long manual analysis time and complexity of AML MRD FCI data, only a few reference laboratories offer this as an orderable test. Methods: To facilitate the analysis of AML MRD by FCI, we devised an unsupervised machine learning (ML) pipeline that intakes raw FCI data, eliminates acquisition errors (FlowCut), performs state-of-the-art clustering (PARC) and dimensionality reduction (UMAP), estimates how aberrant or different-from-normal each cluster is compared to a cohort of controls (15 donor and 15 reactive marrow aspirates), downsamples clusters with high event counts without significant loss of low-event clusters, and corrects for downsampling using an upsampling factor. This CCADDAS (Clinically-intended Clustering of All events, Dimensionality reduction, Downsampling and Aberrancy Scaling) pipeline was containerized, deployed in a cloud environment (Google Cloud Vertex AI), and used to process raw FCI files from 50 AML MRD FCI assays (3-tube/10-color panel) with 0.1% to &lt;5% leukemic subsets, and 20 assays with no detectable leukemic subsets by prior conventional analysis. Machine learning-enhanced data files were analyzed using a commercially available FCI analysis software (Kaluza v 2.1, Beckman Coulter). Results: Cluster-informed downsampling reduced the total number of events needed to be analyzed from 1 million cells to 179,531 cells per tube on average (82% data reduction), while preserving low event subsets (clusters smaller than 5000 cells) and allowing for accurate quantitative estimates for any gated population [# events x mean (upsampling factor)]. MRD was detected in all positive cases, and absent on all negative cases when analyzed using the enhanced downsampled data (100% correlation with original data), with accurate quantitative estimates (Spearman r = 0.85). Analysis time was dramatically reduced from 73 minutes to 8 minutes per case on average (89% reduction of manual analysis time). Dimensionality reduction parameters simplified gating of lineage- and maturation-defined subsets. Cluster comparison against control samples facilitated the rapid identification of MRD subsets, based on a calculated aberrancy scale parameter that helped distinguish MRD blasts (mean ± 2SD: 0.85 ± 0.3) from benign blasts (mean ± 2SD: 0.26 ± 0.3) (p&lt;0.0001). Conclusion: We introduce CCADDAS, an unsupervised containerized cloud ML pipeline that enhances raw AML-MRD FCI data with ML annotations, cluster-informed downsampling and comparison to normal controls to create markedly smaller and software-agnostic FCI files. CCADDAS simplifies and accelerates detection of AML-MRD in clinical diagnostics, reducing number of cells analyzed by 82% and manual analysis time by 89%, without impacting test performance. Moreover, CCADDAS can be deployed on any laboratory-developed AML-MRD assay, and its small-sized export is compatible with any clinical FCI software, computing platform and analysis strategy. Adoption of CCADDAS is likely to facilitate the implementation of AML MRD FCI analysis by more clinical laboratories.