Shreve, Jacob

Abstract Introduction: Suboptimal response to therapy, early disease relapse, and limited long-term survival remain critical challenges in a substantial number of patients with multiple myeloma (MM). While cross-sectional imaging studies are commonly obtained in clinical practice, only limited information is currently incorporated into risk stratification efforts and clinical decision making. Here, we investigate the use of novel image-based biomarkers to identify patients at high risk for adverse outcomes. Methods: We obtained the pretreatment 18F-FDG PET/CT scans of patients with newly diagnosed MM with available clinical and laboratory data seen at Mayo Clinic between 2006 and 2022. In this analysis, we included 637 patients with detectable metabolic tumor volume on their pretreatment scan. The native DICOM images were pre-processed and subjected to automated image segmentation using nnU-Net (nnU-Net v1.0, inference using a pre-trained hematologic malignancies model). Limitations of this automated segmentation approach include heterogeneity introduced by varying image quality, technical variability in PET reconstruction, and potential inclusion of non-neoplastic FDG-avid lesions in segmented regions. PyRadiomics was used to extract quantitative radiomics features for clinical prediction modeling. We evaluated the predictive power of tumor size-, shape-, and texture-related radiomics features for identifying high-risk disease (advanced ISS stage at diagnosis, the presence of high-risk FISH abnormalities, failure to achieve a stringent complete response [sCR] to first-line therapy by IWMG criteria, and early mortality). All radiomics features were quantile-normalized for downstream modeling. Forward and backward feature selection as well as testing for multicollinearity were used to remove redundant features. Results: Median age at diagnosis of the 637 myeloma patients was 64 years (range 22–90), and 387 patients (61%) were male. After a median follow-up of 4.4 years (95% CI 0.2–11.7), median overall survival (OS) was 7.3 years (95% CI 6.3–8.3). Increased total tumor volume (upper tertile vs lower tertiles), measured by the total tumor surface area, was associated with advanced ISS stage (III vs I+II; OR 1.73, 95% CI 1.24–2.42, p=0.001, n=637). Increased tumor density (dichotomized at the median), measured by the median gray level intensity (OR 1.93, 95% CI 1.94–3.12, p=0.007, n=303), and decreased tumor heterogeneity (lower tertile vs upper tertiles), measured by the variability of gray-level intensities (OR 1.88, 95% CI 1.49–3.12, p=0.012, n=303), were associated with the presence of high-risk cytogenetics. Both increased total tumor volume (4.9% increase per tertile, 95% CI 2.2–7.6, p<0.001, n=600) and tumor density (8.9% increase per tertile, 95% CI 4.5–13.3, p<0.001, n=600) were independently associated with more extensive bone marrow involvement at diagnosis. Increased tumor heterogeneity (upper tertiles vs lower tertile), as measured by rapid changes of gray level intensities, was associated with failure to achieve a stringent complete response to first-line therapy (OR 0.50, 95% CI 0.30–0.82, p=0.006, n=351). Increased tumor volume was associated with shorter OS (HR 1.35 per tertile, 95% CI 1.11–1.63, p=0.002, n=301), independent of age, sex, ISS stage, and the presence of high-risk cytogenetics. A simple additive score consisting of tertiles of total tumor volume, ISS stage, and plasma cell labeling index ≥2% identified patients at risk of mortality within 3 years of diagnosis (OR 2.00, 95% CI 1.40–2.85, p<0.001, n=126), with moderate discrimination (AUC=0.71). Sensitivity analysis demonstrated total tumor volume and texture as most predictive of overall survival in a random forest survival model (C-index 0.712, 95% CI 0.649–0.775) and tumor texture as most predictive of response to therapy in a random forest classification model (C-index 0.693, 95% CI 0.607–0.779). Conclusions: Automated image segmentation and radiomics feature extraction from 18F-FDG PET/CT scans produces biomarkers that capture disease stage, burden, and risk in newly diagnosed MM. Total tumor volume and heterogeneity are strongly associated with important disease characteristics and can be used to identify patients at high risk for therapeutic resistance, early mortality, and inferior long-term outcomes.

Abstract Background Chimeric-antigen-receptor (CAR) T-cell therapy has transformed outcomes for relapsed or refractory hematologic malignancies, yet early identification of patients at highest risk for death remains elusive. A prognostic model that leverages only the information routinely available at the time of CAR-T evaluationcould guide pre-infusion counseling, intensity of follow-up, and resource allocation. Methods We analyzed a prospectively maintained cohort of CAR-T recipients treated at Mayo Clinic to develop and test our model. For each patient, data (vital signs, lab results, cardiac assessments [ECG and echocardiography], and PET/CT images) from initial CAR-T evaluation (up to 60 days prior to CAR-T infusion) was considered. Vital signs and labs were left un-imputed, while ECG, echocardiographic parameters, and PET/CT features were forward-filled from the preceding 60 days. Each patient-day was labeled as “outcome present” or “no outcome,” enabling supervised training of gradient-boosting models to predict overall survival (OS) at 6 months post CAR-T infusion. Model performance was evaluated using the area under the ROC curve (AUC) with 10,000-fold bootstrapping with data shuffling between each iteration to derive confidence intervals for stability. To enhance interpretability, we also developed a simplified model using the top 19 features (determined by Shapley value-based importance), aiming to preserve predictive accuracy with a smaller feature set. Results A total of 570 CAR-T cell treated patients were included in this study (476 patients in the training cohort and 94 patients in the validation cohort). The median age was 64 years and 61% males. The primary diagnoses were B-cell lymphoma (324 patients) and multiple myeloma (246 patients). A total of 4,851 individual vital-sign readings, 11,781 laboratory measurements were included. PET/CT images were available in 318 patients. The ML model demonstrated high discriminative ability for predicting OS at 6 months post CAR-T infusion. Vital signs alone produced a mean AUC of 0.63; radiomics features alone produced a mean AUC of 0.71, while laboratory tests alone produced a mean AUC of 0.84. The features most contributory to each single-modality model were then combined into a simplified multimodal model with a total of 19 features, which produced a mean AUC of 0.87, with a 95% CI of 0.73 - 0.97. The 19 features included in the simplified model were: LDH, ferritin, uric acid, eosinophils, fibrinogen, D-dimer, heart rate, mean arterial blood pressure, QTc, QT interval, and the radiomics features of sphericity, first-order Kurtosis, first-order mean, first-order energy, first-order total energy, shape maximum2D diameter (sagittal plane), SUV CoV, SUVp90, shape flatness. This multimodal predictive model was then applied to the validation cohort (uninvolved with model training or testing), which produced an AUC of 0.93, with a 95% CI of 0.84 - 0.98. Conclusions Our ML-driven approach successfully predicts overall survival 6 months post CAR-T infusion, highlighting those individuals at most risk of poor outcomes at the time of CAR-T evaluation. The strong performance of the multimodal version of the model suggests that each data domain contributes complementary resolving power, and by including novel data types, such as PET CT radiomics features, the most accurate predictions could be realized. By providing accurate, near-real-time risk stratification for poor outcomes within 6 months of CAR-T infusion, the model could facilitate proactive patient management. Prospective studies of the integration of this ML model into CAR-T treatment workflows will be crucial to confirm its clinical benefits and to facilitate its adoption in practice.

e13622 Background: Chimeric antigen receptor T-cell (CAR-T) therapy has revolutionized treatment of hematologic malignancies; however, severe toxicities like cytokine release syndrome (CRS) remain a limitation to broader utilization. Using a large dataset of patients treated with CAR-T, we developed a machine learning (ML) model leveraging multi-model patient characteristics to predict the risk of CRS onset with clinically meaningful lead times. Methods: A prospectively maintained dataset of CAR-T patients treated at Mayo Clinic was used. Baseline disease/patients characteristics, labs, ECG, echocardiogram, and vital signs (including data from remote patient monitoring program) were collected. A 24-hour rolling window was used to generate daily averages, maximum, minimum, and deltas for each parameter. ECG and echocardiogram data were forward-imputed; vital signs and labs were not imputed. Each day-post-infusion served as a potential event (CRS onset vs. no CRS). A gradient boosting classifier was trained and tested on a subset of patients (n = 351) using 1,000 bootstrap iterations, evaluating 24-, 48-, and 72-hour lead times. Shapley-based feature importance guided simplification to an 11-feature model, externally validated on an independent cohort (n = 51). Models’ performance assessed using the area under the receiver operating characteristic curve (AUC). Results: A total of 402 patients treated with CAR-T were examined. Median age was 64 years and 36% were female. Primary diagnosis was B-cell lymphoma (n = 240) or multiple myeloma (n = 162). A total of 226,332 vital sign measurements and 372,871 lab values, as well as ECG and echocardiogram data, were collected for 30 days post-infusion. Overall, 79.8% of patients experienced CRS within 30 days. The full ML model demonstrated AUCs of 0.90 (95% CI 0.86–0.93), 0.83 (95% CI 0.77–0.88), and 0.82 (95% CI 0.77–0.87) for 24-, 48-, and 72-hour predictions of CRS, respectively. Including “day of CRS onset post CAR-T” as a feature improved the 72-hour AUC to 0.87 (95% CI 0.83–0.90). A model restricted to vital signs alone showed AUCs of 0.88, 0.75, and 0.71 for 24-, 48-, and 72-hour horizons. An 11-feature model—focusing on temperature, blood pressure, fibrinogen, CRP, serum sodium level, QRS interval, and lymphocyte counts—maintained high accuracy at 24 hours (AUC 0.90, 95% CI 0.86–0.93) and was validated externally using an independent dataset with an AUC of 0.92. Conclusions: Our ML model provides a reliable approach for predicting if/when CRS will occur, with up to three days of lead time before clinical manifestation, offering an opportunity for proactive management. The robust performance of a parsimonious 11-feature model supports practical adoption across clinical settings and real-time risk stratification to enhance patient safety. Prospective validation and integration into clinical workflows will be critical to translating these findings into practice.

Introduction: In myeloma, accurate risk stratification remains a challenge due to substantial biological heterogeneity and highly variable survival outcomes. Cross-sectional imaging is commonly obtained at the time of diagnosis, but imaging findings are not currently incorporated into risk stratification efforts in a standardized manner, despite their prognostic significance. Here, we investigate the use of novel image-based prognostic biomarkers obtained by automated image segmentation and radiomics feature extraction using machine learning methods. Methods: We obtained the pre-treatment 18F-FDG PET/CT scans of 443 patients with newly diagnosed myeloma and complete demographic and cytogenetic data (seen at Mayo Clinic between 2005 and 2019). The native DICOM images were pre-processed and subjected to automated image segmentation using nnU-Net (v1.0, inference using a pre-trained hematologic malignancies model). Limitations of this automated segmentation approach include heterogeneity introduced by varying image quality, technical differences related to PET reconstruction, and potential inclusion of non-neoplastic FDG-avid lesions in segmented regions. The quality of the automated image segmentation was evaluated by comparing the radiomics features obtained from ground-truth segmentation (manual lesion segmentation) and model segmentation in a separate cohort of patients with hematologic malignancies (n=72). We selected 42 radiomics features that moderately (r≥0.30, p<0.005, n=23) or strongly (r≥0.50, p<0.001, n=19) correlated between ground-truth (manual) and model (automated) segmentation for further evaluation in myeloma patients. PyRadiomics was used to extract discrete radiomics features for clinical prediction modeling. We evaluated the prognostic significance of tumor size-, shape-, and texture-related radiomics features and compared their predictive power to established prognostic factors (International Staging System [ISS], Revised ISS [R-ISS] stage and high-risk FISH cytogenetics). The outcome of interest was overall survival and the predictive power of competing proportional hazards regression models was assessed using Harrell's C-statistic. For modeling purposes, patients without measurable tumor volume were categorized as having undetectable radiomics features (0 for all features). All radiomics features were median dichotomized for downstream modeling. Forward feature selection and testing for multicollinearity were used to remove redundant features. Select radiomics features were divided into quintiles to generate an additive radiomics risk score and associated risk strata. Results: Median age at diagnosis of the 443 myeloma patients was 63 years (32-91) and 270 patients (62%) were men. After a median follow-up of 2.4 years (95% CI 0.1-7.4), median overall survival was 6.5 years (95% CI 5.5-NR). Most patients had detectable metabolic tumor volume (n=309, 70%) on their pre-treatment 18F-FDG PET/CT scan. Several of the 42 radiomics features were intercorrelated and two features were selected for further investigation based on their independent prognostic significance (forward feature selection). Higher maximum axial tumor diameter (HR 1.90, 95% CI 1.31-2.76, p=0.001) and a lower concentration of low gray-level values in the tumor (HR 1.73, 95% CI 1.20-2.49, p=0.004) were independently associated with overall survival in newly diagnosed myeloma. These associations were independent of age at diagnosis, sex, and R-ISS stage. We devised a simple additive risk score representing the sum of quintiles of the two radiomics features and divided the patients into three risk categories akin to the R-ISS. The predictive power of this radiomics risk score (Harrell's C=0.60) was comparable to the ISS (C=0.59) and the R-ISS (C=0.57). Conclusions: Automated image segmentation and radiomics feature extraction from 18F-FDG PET/CT scans is feasible. This approach yields image-based biomarkers that correlate with results obtained by manual imaging segmentation. Several radiomics features such as tumor size and heterogeneity have independent prognostic significance in newly diagnosed myeloma. These novel biomarkers rival the predictive power of the currently established risk stratification methods and may be used to improve risk stratification in myeloma.

Background: Relapsed DLBCL occurs in up to 40% of patients and has a poor prognosis. Current prognostic tools such as the international prognostic index (IPI) and newer forms (NCCN-IPI) only include baseline clinical features. PET/CT features have not yet been incorporated into prognostic risk models despite being essential for staging and response assessment. Deep learning computer algorithms have consistently demonstrated the ability to analyze imaging data and identify features that are not readily apparent to the human eye. Using a deep learning computer vision model that automatically extracted radiomic features from PET/CTs, we examined the association of changes between baseline PET and interim PET (PET2) radiomic features on survival outcomes and compared the prognostic performance with NCCN-IPI. Methods: Adult patients with newly diagnosed diffuse large B-cell lymphoma (DLBCL) between 2005 and 2015 and enrolled in the prospective MER cohort were included. Native DICOM images from FDG PET/CT scans at baseline and interim PET (PET2, after 2-3 cycles of chemotherapy) were preprocessed for regularity and then segmented using the nnU-Net architecture with an external lymphoma-specific model. Feature extraction and standard uptake value (SUV) characterization were accomplished with PyRadiomics and NiBabel. Our computational method included no manual segmentation correction. Individual radiomics features (n=115) and the % change above (Δ High) and below (Δ low) 50 th quartile from baseline to PET2 were evaluated for association with event free survival (EFS) and overall survival (OS) using a univariate Cox regression model and by Kaplan Meier analysis. EFS was defined as time from PET2 to first systemic relapse or death from any cause. OS was defined as time from PET2 to death from any cause. Additionally, Spearman and Pearson correlation matrices were used to guide feature selection while minimizing collinearity. We then assessed the association of the highest scoring radiomics features with EFS in a multivariate cox regression model. Results: A total of 506 patients had baseline and PET2 scans available for radiomic feature extraction. Median age at diagnosis was 61 years (range 18-90), 43% were females. NCCN-IPI composition and outcomes of our cohort was comparable to the previously published NCCN cohort, with patient distribution and 5-year OS as follows: Low (9%, 5-y OS 83%), Low-intermediate (38%, 5-y OS 49%), High-intermediate (41%, 5-y OS 31%), High (11%, 5-y OS 12%). At a median follow up of 81 months,there were 308 events, and272 patients had died.Several radiomics features were associated with EFS including change in tumor surface area (SA Δ) (HR: 1.75, 95% CI 1.36-2.24, p<0.005, median of 26,927 mm 2, Figure 1A) and change in total metabolic tumor volume (TMTV Δ) (HR: 1.64, 95% CI 1.27-2.10, p<0.005, median of 120,473 mm 3, Figure 1B). As a reference, NCCN-IPI univariate Cox regression analysis of EFS produced a Harrell's C-statistic (c) of 0.624. Single radiomics features of SA Δ and TMTV Δ individually produced unadjusted c-statistics of 0.616 and 0.614 respectively. Combinations of radiomics features (SA Δ, TMTV Δ and NCCN-IPI classification) demonstrated cumulative benefit (c=0.652). Conclusion: Automatic PET-CT radiomic feature extraction using deep learning models is feasible among patients with newly diagnosed DLBCL. The change in tumor SA and TMTV appear to be promising radiomic biomarkers. A low decline in either tumor SA or TMTV by PET2 is associated with higher risk of relapse and poor survival outcomes. As individual characteristics, these features are similar prognostic indicators of EFS as the NCCN-IPI. Further investigation into risk stratification when combining radiomics with clinical and genomic variables is warranted. Figure 1. Event free survival for DLBCL patients with A) a change in tumor surface area by PET2 less than 50% percentile (SA_Δ Low) and greater than 50 th percentile (SA_Δ High) and B) a change in TMTV by PET2 less than 50% percentile (TMTV_Δ Low) and greater than 50 th percentile (TMTV_Δ High).