JDJustin Do

Justin Do

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

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Justin Do has a total of 25 co-authors. Here are the top co-authors by number of shared publications:

  • Melissa Chen Xu

    Department of Surgery, Sidney Kimmel Medical College, Thomas Jefferson University, Philadelphia, PA

    3 shared publications
  • Vivaswat Suresh

    Department of Computer Science and Engineering, University of California , San Diego, CA 92093

    3 shared publications
  • Babak Abai

    Division of Vascular and Endovascular Surgery, Department of Surgery, Thomas Jefferson University Hospital, Philadelphia, PA

    2 shared publications
  • Dawn Salvatore

    Division of Vascular and Endovascular Surgery, Department of Surgery, Thomas Jefferson University Hospital, Philadelphia, PA

    2 shared publications
  • Michael Nooromid

    Division of Vascular and Endovascular Surgery, Department of Surgery, Thomas Jefferson University Hospital, Philadelphia, PA

    2 shared publications
  • Nidhir Guggilla

    Department of Electrical Engineering and Computer Sciences, University of California, Berkeley , Berkeley, CA 94720

    2 shared publications

Latest Publications

  • Management Modality of Thoracoabdominal Aortic Aneurysms is Associated with Lower Rates of Follow-Up

    by Melissa Chen Bs Xu, Devi Bs Schwartz, Justin Bs Do, Babak Abai, Dawn Salvatore, Paul Dimuzio, Michael Nooromid

  • Open Access

    Leveraging Multimodal Large Language Models for Fall Risk Reduction in Older Adults in the Home: Proposed Model Design

    by Justin Do, Vivaswat Suresh, Lily Zhang, Bharvi M Chavre, Jeremy Cha, Robert Pugliese

    This research letter proposes a novel model design leveraging natively multimodal large language models to identify fall risks and generate visualizations of recommended home environmental modifications, aiming to improve the accessibility and impact of personalized fall prevention advice for older adults. Through a pilot rating study, this work demonstrates that multimodal large language models can generate safe and actionable advice to reduce fall risk in lived spaces of older adults, and also generate realistic edits based on original images. While this concept needs further testing and clinical comparison, it highlights a promising avenue for further innovation of fall prevention tactics.


  • Open Access

    Integration of intraoperative data in interpretable machine learning models to predict postoperative AKI in noncardiac surgery patients

    by Justin Do, Karan H Shah, Melissa Chen Xu, Andrew Hyunwoo Kim, Vivaswat Suresh, Nidhir Guggilla, Michael Li, Rishi Kothari

    Objectives: We aimed to (1) quantify changes in discrimination when adding intraoperative data to preoperative data and (2) compare tabular machine learning with feature engineering against a time-aware LSTM-based model. Materials and Methods: Retrospective cohort of 46 204 adults undergoing 57 055 eligible noncardiac surgery in the INSPIRE database. We extracted 38 preoperative and 49 intraoperative variables; acute kidney injury (AKI) was defined by KDIGO serum creatinine criteria and modeled as stage 2/3 postoperative AKI. Models were trained on preoperative-only and combined pre- and intraoperative data. Intraoperative series were summarized using eight statistical features for tabular models or integrated directly using an MLP+LSTM architecture. Results: GBT with combined features achieved the highest AUROC (0.896, 95% CI, 0.878-0.914), followed by combined AutoGluon (0.893, 95% CI, 0.877-0.909) and preoperative-only GBT (0.891, 95% CI, 0.871-0.910). ASA-PS ≥3 (AUROC 0.723, 95% CI, 0.700-0.746) and adapted GS-AKI (AUROC 0.719, 95% CI, 0.700-0.739) underperformed machine-learning models. The hybrid MLP+LSTM model did not outperform simpler tabular models (AUROC 0.870, 95% CI, 0.848-0.892). Discussion: The small gain from adding low-frequency intraoperative summaries suggests that most discriminative information for stage 2/3 postoperative AKI was available before surgery. Conclusion: Preoperative tabular ML models provided excellent prediction of postoperative AKI, and added limited incremental discrimination at the available sampling frequency. Future work should evaluate whether higher-frequency intraoperative signals better leverage time-aware architectures.

  • Open Access

    Evaluating Conversational Image Segmentation for Medicine: Performance, Failure Modes, and a Fairness Audit Across Seven Modalities

    by Justin Do, Nidhir Guggilla, Vivaswat Suresh, Rishi Kothari

    Introduction: Medical-image segmentation underpins quantitative diagnostics and research, yet state-of-the-art models remain task-specific and data-hungry. The recent emergence of powerful, multimodal large language models (LLMs) presents a generalizable option; however, their efficacy in the specialized medical domain remains largely unquantified. We aim to benchmark the foundational Gemini 2.5 Flash and Gemini 2.5 Flash-Lite models for zero-shot medical image segmentation and evaluate them for potential bias in performance. Methods: The models were tested on 13,086 medical images spanning seven distinct imaging modalities (endoscopy, fundoscopy, dermoscopy, laparoscopy, ultrasound, radiography, and CT) and 10 clinically relevant segmentation targets (e.g., colorectal polyps, skin lesions, liver tumors). Prompts followed a standard template (“Segment the … in this image.”). Per-image Dice and Intersection-over-Union (IoU) were computed against publicly released expert masks. Bias was assessed on 4,057 dermoscopy images split by Individual-Typology-Angle (ITA) into Light (> 28°, n = 3,499) and Dark (≤ 28°, n = 558) groups. Results: Flash achieved a mean Dice=0.766 (IoU=0.680) for colorectal polyp segmentation, Dice=0.761 (IoU=0.672) for skin lesion segmentation, Dice=0.824 (IoU=0.736) for optic disc segmentation, and Dice=0.718 (IoU=0.616) for surgical tool segmentation, outperforming Flash-Lite by approximately 0.1 Dice points. Accuracy declined on low-contrast radiological tasks (Liver Mass CT Dice=0.071). In the fairness audit, Flash produced a successful mask for 2,839/3,499 light-tone images (81.2%) versus 378/558 dark-tone images (67.7%); χ² = 51.8, p < 0.001. Using all images, mean IoU=0.686 for light tones and IoU=0.591 for dark tones (Kruskal–Wallis H = 62.6, p < 0.001); Cliff’s δ = –0.208 (95% CI – 0.259 to –0.159). Discussion: Gemini 2.5 Flash delivers competitive accuracy on high-contrast photographic datasets at negligible cost. Performance is weaker on radiographic modalities (ultrasound, CT, chest radiography), and in dermoscopy, we observe lower accuracy on darker ITA skin-tone groups. This study informs the field where foundational LLMs are deployment-ready for medical image segmentation and where targeted debiasing or domain adaptation is required.

  • Open Access

    Large language models improve readability of patient education materials on vascular conditions

    by Melissa Chen Xu, Nicola Habash, Justin Do, Jiaying Bi, Dawn Salvatore, Paul Dimuzio, Michael Nooromid, Babak Abai


  • Open Access

    Just as expected? Older adults’ aging expectations are associated with subjective cognition

    by Nikki L. Hill, Sakshi Bhargava, Justin Do, Emily Bratlee-Whitaker, Monique J. Brown, Renata Komalasari, Rachel Wu, Jacqueline Mogle

    This study investigated the relationship between older adults' expectations regarding aging and subjective cognition. Specifically, we examined whether the three domains of aging expectations (physical health, mental health, and cognitive function) were associated with two aspects of subjective cognition: current subjective cognition and subjective cognitive decline (SCD). An online survey was conducted among U.S. adults aged 65-90 (N = 581; Mage=71.4, SD ± 4.81; 51% female). Measures included the 12-item Expectations Regarding Aging scale, the 8-item PROMIS Cognitive Abilities scale (current subjective cognition), and the 12-item Everyday Cognition scale (SCD). We used generalized linear models to examine associations between overall aging expectations and its three domains with current subjective cognition ratings and SCD. We found that more positive expectations regarding physical health, mental health, and cognitive function in aging were associated with higher ratings of current subjective cognition as well as lower SCD. The magnitude of effects across aging expectations domains were similar for both aspects of subjective cognition. Aging expectations are malleable and influence an individual's perceptions of their cognitive functioning. Modifying older adults' aging expectations could support healthier cognitive aging through increased awareness and accurate assumptions about the aging process.