About

Julia Chang is an Associate Professor of Hispanic Studies at Cornell University, affiliated with the Feminist, Gender, and Sexuality Studies Program and the Southeast Asia Program. Her interdisciplinary research and teaching encompass nineteenth-century Spanish literature, with a focus on the realist novel, as well as contemporary Spanish non-fiction and visual art, particularly works by racialized writers and artists. Her scholarly work also engages with Philippine literature in Spanish, feminist and queer theory, histories of racial formation, disability studies, and game studies. Chang's first book, Blood Novels: Gender, Caste, and Race in Spanish Realism, was published by the University of Toronto Press in 2022 and received the inaugural Harriet S. Turner Beca from the Asociación Internacional de Galdosistas. She has been recognized with the 2021 Stephen and Margery Russell Distinguished Teaching Award. Her research includes exploring themes of race, gender, and empire in Spanish literature, as well as issues of disability, sexuality, and racial formation. She has held a faculty position at Brown University prior to her appointment at Cornell and is actively involved in research collectives and editorial projects related to her fields of interest.

Research topics

• Medicine
• Nuclear medicine
• Computer science
• Optics
• Physics

Selected publications

• Development and implementation of a unified curriculum in math, statistics, computing, and informatics with financial considerations for medical physics graduate programs

  Journal of Applied Clinical Medical Physics · 2026-05-01

  articleOpen access1st authorCorresponding

  BACKGROUND: AAPM Report No. 365 recommends that medical physics graduate programs offer courses covering both mathematical and statistical methods (Section 3.1.7) as well as computational methods and medical informatics (Section 3.1.8). While our program seeks to incorporate both of these essential areas into the curriculum, various financial and programmatic constraints have necessitated a more streamlined approach. Accordingly, this work presents a single 2-semester-hour course designed to address these topics in an integrated format. PURPOSE: In this paper, we present our efforts in developing a new teaching approach that addresses both AAPM Report No. 365 recommendations in one course. METHODS: The major challenge of designing this course was the insufficient number of semester hours allocated to teaching both topics. To overcome this obstacle, we implemented a novel approach to homework assignments. Unlike the traditional approach in which students complete homework manually, students in this class were asked to write computer programs to solve most homework questions. These carefully designed assessments not only enhanced students' understanding of the course materials but also required them to utilize appropriate computational methods. Recognizing that students had varying levels of coding experience from their undergraduate studies, the program instructed them, prior to starting the program, to acquire foundational Python skills through self-guided learning to prepare for this course. In addition, basic Python programming guidance was provided with each homework assignment to support students with less coding experience. RESULTS: The final course covered the following key mathematical concepts: signals and systems, Fourier series and transform, probability, statistical inference, image quality, optimization methods, and an introduction to artificial intelligence with an emphasis on machine learning. To complete the homework assignments, students developed coding skills in data visualization, numerical integration, convolution, continuous-time/discrete-time/fast Fourier transforms, random number generator, point estimation, confidence interval, hypothesis testing, linear models, DICOM, the Rose model, conjugate-gradient descent, iterative methods for solving systems of linear equations, and support vector machines. The course was offered in the Spring 2024 and Spring 2025 semesters and received generally positive evaluations, with some noted challenges per the Course and Teacher Rating reports. Overall, students reported that the course was educationally beneficial; however, some indicated that the coding-based assignments were demanding. CONCLUSION: We successfully developed and implemented a course that covers mathematical and statistical methods as well as computational methods and medical informatics as recommended in AAPM Report No. 365.

  PublisherDOI

• Quantifying stent‐induced dose perturbations in intravascular brachytherapy using 3D‐ printed phantoms and film dosimetry

  Journal of Applied Clinical Medical Physics · 2026-01-01

  articleOpen accessSenior author

  BACKGROUND: Coronary artery disease (CAD), the leading cause of death worldwide, is the narrowing of coronary arteries due to atherosclerotic plaque buildup. A common treatment for CAD is percutaneous coronary intervention (PCI), often involving stent placement. However, a common complication or in-stent restenosis (ISR) can occur in 10%-20% of patients which call for the use of therapies like intravascular brachytherapy (IVBT). IVBT delivers targeted beta radiation, typically from Sr-90/Y-90 sources, to inhibit neointimal hyperplasia and reduce restenosis rates. Accurate dose delivery is critical to treatment success, but challenges such as source positioning and dose uniformity persist. Recent advances in 3D printing and radiochromic film dosimetry offer promising tools for more precise dose verification in IVBT, enabling high-resolution assessment of dose distributions and stent-induced perturbations. PURPOSE: IVBT requires precise position to ensure effective treatment. However, stents introduce complexities in dose distribution due to their material and geometry, which can lead to attenuation and impact treatment outcomes. This study aimed to quantify stent-induced dose perturbations using a custom 3D-printed stent phantom and Gafchromic EBT-4 Film, providing insights for dosimetry of IVBT. METHODS: Dose measurements were conducted using a custom designed 3D-printed stent phantom. The film calibration was performed using the RIT film dosimetry package from 0 to 12 Gy. The phantom was designed for a Synergy XD Stent with a diameter of 3 mm with the Sr-90/Y-90 source catheter position designed to be in the center of the stent. Percent depth dose (PDD) distributions were modeled using the third-order exponential polynomial function and compared with Monte Carlo simulations to evaluate agreement. Discrepancies were quantified using root mean square error (RMSE) and mean absolute error (MAE). The stent effect on PDD was analyzed using a paired t-test, and a dose reduction factor (DRF) was calculated to assess attenuation. RESULTS: The third-order exponential polynomial function demonstrated an excellent fit for both configurations, with R-squared values of 0.999 (no stent) and 0.999 (with stent). RMSE and MAE values were slightly higher for the with-stent dataset (0.038 and 0.036, respectively), reflecting increased discrepancies. The paired t-test showed a statistically significant difference between PDD values (t = -6.591, p < 0.0001). The average PDD difference between configurations was 4.26% in the clinically relevant region (2-5 mm). The DRF ranged from 1.18% to 7.92%, with an average attenuation of 4.5%. CONCLUSION: The presence of a stent significantly impacts dose delivery in IVBT, attenuating approximately 4.5% of the dose within clinically relevant depths. These findings highlight the importance of accounting for stent-induced attenuation in treatment planning to ensure accurate dose delivery. The custom stent phantom demonstrates its usefulness in capturing dose perturbations, offering an effective tool for improving IVBT dosimetry.

  PublisherOA PDFDOI

• Characterizing the Variations of Dose Delivery to Coronary Vessels for Sr-90/Y-90 IVBT Using Monte Carlo Simulations

  International Journal of Medical Physics Clinical Engineering and Radiation Oncology · 2026-01-01

  articleOpen accessSenior author
  PublisherDOI

• Bragg-peak FLASH biological optimization enables enhanced normal tissue sparing and dose escalation for ocular stereotactic body radiation therapy

  Physics in Medicine and Biology · 2025-10-02 · 1 citations

  articleOpen access

  Abstract Objective. To evaluate proton Bragg peak FLASH for ocular treatments to enhance normal tissue sparing and enable dose escalation via FLASH biological optimization (FBO). Approach. The FLASH-sparing factors for normal tissues were derived from the literature in modeling the phenomenological FLASH normal tissue sparing effect. Using the single-energy BP-FLASH technique (SEBP-FLASH), an in-house treatment planning system was implemented with the FLASH FBO module. Ten consecutive ocular patients who were treated using conventional dose rate intensity-modulated proton therapy (CONV-IMPT) to 50 Gy in 5 fractions were replanned using the FLASH technique. The dose metrics for the OARs were compared using the two different techniques. The fraction dose was then intentionally escalated from 10 to 12 Gy through FBO to assess whether the plans still met clinical constraints. Main results. In the FLASH regimen without FBO (50 Gy/5 fractions), all ipsilateral OAR dosimetric metrics met clinical objectives with safe margins. While the clinical CONV-IMPT approach demonstrated slightly better dosimetric performance than SEBP-FLASH plans, the incorporation of FBO improved all OAR dose metrics beyond those of CONV- IMPT, except for the mean dose to the cornea (no difference). When the target dose was increased from 50 to 60 Gy using FBO, all OARs remained within clinical limits. The mean and maximum doses to the cornea increased from 11.7 to 15.4 Gy and from 22.8 to 23.6 Gy, respectively, when transitioning from 50 Gy CONV-IMPT to 60 Gy FBO. However, in the 60 Gy FBO plans, the maximum doses were reduced for the eye (102.0%–87.0%), optic nerves (98.7%–74.0%), retina (100.5%–81.8%), lacrimal gland (84.9%–73.2%), and conjunctiva (91%–72.3%). Significance. SEBP-FLASH achieves plan quality comparable to CONV-IMPT using 50 Gy/5 fractions and enables dose escalation via FLASH FBO while meeting clinical standards, potentially improving tumor control with acceptable toxicity.

  PublisherDOI

• L4 Automation of IMRT QA with the AutoFrame Robotic Platform

  International Journal of Medical Physics Clinical Engineering and Radiation Oncology · 2025-01-01

  articleOpen accessSenior author
  PublisherDOI

• A comprehensive comparative study of the two predominant techniques for delivering conformal proton pencil beam scanning FLASH radiotherapy

  International Journal of Particle Therapy · 2025-03-01

  articleOpen access
  PublisherDOI

• Contour Guided-Deep Radon Prior: an Unsupervised Framework for CT Reconstruction

  2025-11-01

  article

  Background: The inverse problem in computed tomography (CT) reconstruction has long been a critical research topic in medical imaging. Although deep learning-based approaches have achieved remarkable progress in solving CT reconstruction problems, most methods heavily depend on large-scale high-quality training datasets and lack interpretability. Method: This paper proposes a novel Contour-Guided Deep Radon Prior (CG-DRP), a fully unsupervised framework for CT reconstruction. Built upon the Deep Radon Prior (DRP), CG-DRP incorporates geometric contour information to guide the reconstruction process. Specifically, the proposed method learns projection priors in the Radon domain while integrating edge-aware constraints to effectively compensate for missing projection data. By embedding a neural network as an implicit prior into the iterative reconstruction process, CG-DRP enables synergistic optimization through gradient feedback in both the image and Radon domains. Results: Experimental results demonstrate the effectiveness of CG-DRP in addressing inverse problems such as limited-angle and sparse-view CT reconstruction. The contour-guided mechanism significantly enhances edge preservation and improves the clarity of structural boundaries in reconstructed images. Compared with existing supervised learning approaches, CG-DRP achieves comparable or superior reconstruction performance without requiring any training data. Conclusions: CG-DRP introduces a new paradigm for CT image reconstruction by leveraging deep self-correlation in the Radon domain and contour priors to facilitate effective collaboration between neural network manifolds and image reconstruction. This unsupervised learning framework not only enhances reconstruction quality but also improves algorithm interpretability, providing a scalable solution for medical image reconstruction.

  PublisherDOI

• Bragg-peak FLASH biological optimization with dose escalation in ocular SBRT

  International Journal of Particle Therapy · 2025-03-01

  articleOpen access
  PublisherDOI

• The Directive Application: Enhancing Clinical Workflow and Transforming Oncology Care with Centralized Data Management

  International Journal of Radiation Oncology*Biology*Physics · 2024-09-27

  article
  PublisherDOI

• 730 Convolutional neural networks: What dermatologists should know about vision AI

  Journal of Investigative Dermatology · 2024-07-19

  articleOpen access
  PublisherOA PDFDOI

Frequent coauthors

• Dattatreyudu Nori

  Cornell University

  159 shared

• A Sabbas

  NewYork–Presbyterian Hospital

  120 shared

• Samuel Trichter

  104 shared

• Fridon Kulidzhanov

  NewYork–Presbyterian Hospital

  100 shared

• K Chao

  98 shared

• A. Gabriella Wernicke

  Hofstra University

  80 shared

• Bhupesh Parashar

  73 shared

• Lili Zhou

  Zhejiang Normal University

  68 shared

Education

• Ph.D., Electrical Engineering

  New York University Polytechnic Institute

  1995

Awards & honors

• Harriet S. Turner Beca by the Asociación Internacional de Ga…
• 2021 Stephen and Margery Russell Distinguished Teaching Awar…
• Fellow in the CIVIC Research Collaborative in Media Studies…