MBMasha Bondarenko

Masha Bondarenko

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Masha Bondarenko has a total of 48 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Investigating AI’s potential in risk-stratifying lung cancer screenings for extending screening intervals and reducing workload: A retrospective analysis of a tertiary hospital cohort

    by Ali Nowroozi, Masha Bondarenko, Sukhmanjit S Brar, Brian Haas, James Brown, Alison Rustagi, Brian Graham, Mollie Hudson, George Su, Thienkhai H. Vu, et al.

    Objective: We aimed to 1) evaluate whether AI algorithms could identify a very low-risk sub-population Lung-RADS 1/2 cases suitable for a safe switch to biennial screening, and 2) externally validate Sybil, an imaging-based deep learning lung cancer risk prediction tool. Materials and methods: We gathered all lung cancer screenings (LCSs) from a tertiary safety-net hospital (n=3120, up to September 2024), manually extracted relevant clinical data, and analyzed Sybil’s performance across various subgroups, with a specific focus on Lung-RADS 1 and 2 cases. We simulated LCS interval extension using Sybil and the Brock model risk scores calibrated on a subset of the National Lung Screening Trial (NLST). Confidence intervals were obtained through bootstrapping. Results: The median follow-up was 2.3 years (interquartile range 1.5–3.1), with a total 6-year cancer incidence rate of 2.3% (72/3120). Lung-RADS 1 and 2 cases comprised 90.5% of our scans. Sybil’s overall area under the curve (AUC) ranged from 0.819 to 0.912; while its 2-year AUC for Lung-RADS 1 and 2 cases was 0.746. In the screening interval extension simulation, calibrating for a 30% shift to biennial screening translated into 33.8% of negative screens in our data being safely extended using Sybil, with no delayed cancers. The Brock model demonstrated an AUC of 0.513 on our data, with a sensitivity of 18.2% using the 1.5% threshold. Conclusion: Sybil has the potential to safely extend LCS intervals for low-risk individuals, reducing radiology workload and opening spots for life-saving new screenings. Key points: We investigated whether AI risk stratification tools could identify the lowest-risk individuals undergoing lung cancer screening suitable for switching to a biennial screening schedule. In our simulations, Sybil reduced follow-up screening load by at least 30% through switching the lowest-risk Lung-RADS 1 and 2 cases to biennial screening without delayed cancer diagnoses. Very low-risk individuals, as predicted by AI algorithms, could potentially be placed on biennial screening schedules, reducing radiology department workload and freeing time for new patients.

  • Open Access

    Planner-Executor Style Multimodal Agentic System to Answer Patient Questions in Lung Cancer Screening CT

    by Koharu Sakiyama, Adrian Serapio, Meng Ye, Ali Nowroozi, Masha Bondarenko, Yu-Fu Wu, Kang Qi, Jonathan Liu, Maya Vella, Yannan Yu, et al.

    Background: Accurate patient understanding of lung cancer screening (LCS) results is critical for engagement and follow-up adherence, given the currently low screening uptake. While artificial intelligence (AI) systems show promise in facilitating patient communication, its effective use in clinical settings requires appropriate invocation of imaging tools and self-regulation by deferring certain questions to physicians. To this end, we developed and evaluated a planner-executor style multimodal agentic system to answer simulated patient questions about LCS CT results. Methods: This retrospective study utilized 116 LCS CT reports and images collected from a tertiary academic hospital. The system employed a multi-agent architecture (planner and executor) to mimic clinical reasoning. We developed a large language model (LLM)-based question generation framework to prepare a comprehensive set of 699 simulated patient questions, categorized as answerable and defer-to-doctor to assess self-regulation. Performance was evaluated on tool-calling, self-regulation accuracy (ability to correctly defer out-of-scope questions to a human provider) assessed by LLM-as-a-judge, and clinical quality assessed by a reader performance study with four physicians evaluating a subset of 100 responses. Results: The system demonstrated high overall tool-calling accuracy of 93.1% (646/694; 95% CI: 91.2%, 95.0%) and self-regulation accuracy of 92.6% (462/499; 90.2%, 94.8%). For answerable questions, the percentage of responses receiving perfect 5.0 scores across all readers included 78.9% (180/228) for clinical accuracy (inter-rater agreement: Gwet’s AC2=0.92), and 73.2% (167/228) for patient understandability (0.86). For defer-to-doctor questions, the percentage of responses included 77.3% (133/172) for clinical accuracy (0.80) and 70.3% (121/172) for patient understandability (0.78). Conclusions: The planner-executor style multi-modal agentic system reliably answers patient-specific questions about LCS. Its high tool-calling and self-regulation performance demonstrates its potential as a safe and effective digital communication facilitator in LCS programs.

  • Open Access

    Prediction of Subsolid Pulmonary Nodule Evolution from Baseline CT Using Temporal Imaging Models

    by Masha Bondarenko, Kang Qi, Ali Nowroozi, Justin Kim, Brian Kunzang, Aaron Lee, Jonathan Liu, Nicole Tran, Shiny Weng, Maya Vella, et al.

    Background: Prediction of subsolid pulmonary nodule (SSN) progression from baseline CT may improve risk stratification and surveillance planning, but prior approaches have largely relied on fixed follow-up intervals. Methods: This retrospective single-center study evaluated interval-aware temporal imaging models for predicting future SSN growth and morphology across heterogeneous surveillance durations. A total of 24,946 longitudinal scan pairings derived from 2,543 clinician-reviewed SSNs in 426 patients were analyzed. A discriminative deep learning model predicted interval growth from baseline CT, segmentation masks, and interscan interval information, while a temporally conditioned generative model predicted future lesion morphology. Results: The discriminative model achieved an area under the receiver operating characteristic curve of 0.772 (95% confidence interval: 0.704–0.818), with sensitivity of 80.2% and specificity of 58.7% on the test cohort. The generative model predicted future lesion morphology with a Dice similarity coefficient of 0.706 ± 0.186. Prediction performance decreased with increasing follow-up duration, although both models generalized across intervals ranging from months to years. Conclusion: Interval-aware temporal imaging models enable the prediction of future SSN growth and morphology from baseline CT while accounting for variable surveillance intervals. These findings suggest a framework for time-aware, personalized risk assessment that may support individualized surveillance strategies and future AI-assisted management of pulmonary adenocarcinoma spectrum lesions.

  • Open Access

    Pre‐Imaging Clinical Factors Associated With Cardiac MR Image Quality Using Large Language Model‐Enabled Data Extraction

    by Hong Yu, Masha Bondarenko, Ali Nowroozi, Yoo Jin Lee, Adrian Serapio, Punita Kaveti, Jae Ho Sohn

    Background: Poor cardiac MR image quality can prompt repeat examinations and hinder clinical decision‐making. Purpose: To evaluate whether pre‐imaging clinical information, extracted using a large language model (LLM), is independently associated with cardiac MR image quality. Study Type: Retrospective. Population: 1006 adults undergoing clinical cardiac MR examinations. Field Strength/Sequence: 1.5 T and 3 T scanners with cine, black blood, MR angiogram, or late gadolinium enhancement protocols. Assessment: Image quality was categorized per study as excellent, slightly limited, severely limited, or nondiagnostic using institutional reporting conventions finalized by radiologists and cardiologists. A HIPAA‐compliant LLM assigned image quality labels based on radiology reports through an iteratively refined prompt, with reliability confirmed by two radiologists. Labels were binarized as Good (excellent and slightly limited) versus Poor (severely limited and nondiagnostic). A repeat‐imaging‐adjusted image quality label was used in a sensitivity analysis. Pre‐imaging clinical and patient variables were extracted from electronic health records. Associations between variables and image quality labels were investigated. Statistical Tests: Cohen's kappa (κ) for label agreement. Chi‐square and t‐tests for univariate analysis. Variance inflation factor (VIF) and multivariable logistic regression. Significance level: p < 0.05. Results: Binarized image quality labels showed substantial agreement with interpreters' assessments for both the primary dataset (κ = 0.689) and the repeat‐adjusted dataset (κ = 0.879). There was no significant multicollinearity (VIF = 1.01–1.39). Cognitive and communication impairment (OR 1.81, 95% CI [1.30–2.54], p < 0.001) and respiratory issues (1.57 [1.14–2.17], p = 0.006) were independently associated with poor image quality. These associations remained significant in the repeat‐adjusted sensitivity analysis (cognitive and communication impairment (OR 1.75, 95% CI [1.27–2.44], p < 0.001) and respiratory compromise (OR 1.37, 95% CI [1.04–1.82], p = 0.027)). Other clinical variables were not independently associated after adjustment. Data Conclusion: Cognitive/communication impairment and respiratory compromise were independently associated with poor cardiac MR image quality. Level of Evidence: 3. Technical Efficacy: Stage 2.


  • Open Access

    Large language models in radiologic numerical tasks: A thorough evaluation and error analysis

    by Ali Nowroozi, Masha Bondarenko, Adrian Serapio, Tician Schnitzler, Sukhmanjit S Brar, Jae Ho Sohn

    Purpose: To investigate the performance of LLMs in radiology numerical tasks and perform a comprehensive error analysis. Materials and Methods: We defined six tasks: extracting 1-minimum T-score from DEXA report, 2-maximum common bile duct (CBD) diameter from ultrasound report, and 3-maximum lung nodule size from CT report, and judging 1-presence of a highly hypermetabolic region on a PET report, 2-whether a patient is osteoporotic based on a DEXA report, and 3-whether a patient has a dilated CBD based on an ultrasound report. Reports were extracted from the MIMIC III and our institution’s databases, and the ground truths were extracted manually. The models used were Llama 3.1 8b, DeepSeek R1 distilled Llama 8b, OpenAI o1-mini, and OpenAI GPT-5-mini. We manually reviewed all incorrect outputs and performed a comprehensive error analysis. Results: In extraction tasks, while Llama showed relatively variable results (ranging 86%-98.7%) across tasks, other models performed consistently well (accuracies >95%). In judgement tasks, the lowest accuracies of Llama, DeepSeek, o1-mini, and GPT-5-mini were 62.0%, 91.7%, 91.7%, and 99.0%, respectively, while o1-mini and GPT-5-mini did reach 100% performance in detecting osteoporosis. We found no mathematical errors in the outputs of o1-mini and GPT-5-mini. Answer-only output format significantly reduced performance in Llama and DeepSeek but not in o1-mini or GPT-5-mini. Conclusion: True reasoning models perform consistently well in radiology numerical tasks and show no mathematical errors. Simpler non-true reasoning models may also achieve acceptable performance depending on the task.

  • Open Access

    Development and validation of a risk nomogram predicting pneumothorax requiring chest tube placement post-percutaneous CT-guided lung biopsy

    by Masha Bondarenko, Jianxiang Zhang, Ulysis Hugo Baal, Brian Lam, Gunvant Chaudhari, Yoo Jin Lee, Jamie Schroeder, Maya Vella, Brian Haas, Thienkhai Vu, et al.

    Background: Pneumothorax requiring chest tube after CT-guided transthoracic lung biopsy presents added clinical risk and costs to the healthcare system. Identifying high-risk patients can prompt alternative biopsy modes and/or better preparation for more focused post-procedural care. We aimed to develop and externally validate a risk nomogram for pneumothorax requiring chest tube placement following CT-guided lung biopsy, leveraging quantitative emphysema algorithm. Methods: This two-center retrospective study included patients who underwent CT-guided lung biopsy from between 1994 and 2023. Data from one hospital was set aside for validation (n = 613). Emphysema severity was quantified and categorized to 3-point scale using a previously published algorithm based on 3×3×3 kernels and Hounsfield thresholding, and a risk calculator was developed using forward variable selection and logistic regression. The model was validated using bootstrapping and Harrell’s C-index. Results: 2,512 patients (mean age, 64.47 years ± 13.38 [standard deviation]; 1250 men) were evaluated, of whom 157 (6.7%) experienced pneumothorax complications requiring chest tube placement. After forward variable selection to reduce the covariates to maximize clinical usability, the risk score was developed using age over 60 (OR 1.80 [1.15–2.93]), non-prone patient position (OR 2.48 [1.63–3.75]), and severe emphysema (OR 1.99 [1.35–2.94]). The nomogram showed a mean absolute error of 0.5% in calibration and Harrell’s C-index of 0.664 in discrimination in the internal cohort. Conclusion: The developed nomogram predicts age over 60, non-prone position during biopsy, and severe emphysema to be most predictive of pneumothorax requiring chest tube placement following CT-guided lung biopsy.


  • Open Access

    A Nomogram to Predict Pneumothorax Requiring Chest Tube Placement following Percutaneous CT-guided Lung Biopsy

    by Masha Bondarenko, Jianxiang Zhang, Ulysis Hugo Baal, Brian Lam, Gunvant Chaudhari, Yoo Jin Lee, Jamie Schroeder, Maya Vella, Brian Haas, Thienkhai Vu, et al.

    Background Pneumothorax requiring chest tube after CT-guided transthoracic lung biopsy is one of the common complications, and the required hospital stay after chest tube placement represents an added clinical risk to patients and cost to the healthcare system. Identifying high-risk patients can prompt alternative biopsy modes and/or better preparation for more focused post-procedural care. Purpose To develop and externally validate a risk nomogram for pneumothorax requiring chest tube placement following CT-guided lung biopsy, leveraging quantitative emphysema algorithm. Materials & Methods This two-center retrospective study included patients who underwent CT-guided lung biopsy from between 1994 and 2023. Data from one hospital was set aside for validation (n=613). Emphysema severity was quantified and categorized to 3-point scale using a previously published algorithm based on 3×3×3 kernels and Hounsfield thresholding, and a risk calculator was developed using forward variable selection and logistic regression. The model was validated using bootstrapping and Harrell’s C-index. Results 2,512 patients (mean age, 64.47 years +/-13.38 [standard deviation]; 1250 men) were evaluated, of whom 157 (6.7%) experienced pneumothorax complications requiring chest tube placement. After forward variable selection to reduce the covariates to maximize clinical usability, the risk score was developed using age over 60 (OR 1.80 [1.15-2.93]), non-prone patient position (OR 2.48 [1.63-3.75]), and severe emphysema (OR 1.99 [1.35-2.94]). The nomogram showed mean absolute error of 0.5% in calibration and Harrell’s C-index of 0.664 in discrimination in the internal cohort. Conclusion The developed nomogram predicts age over 60, non-prone position during biopsy, and severe emphysema to be most predictive of pneumothorax requiring chest tube placement following CT-guided lung biopsy.