News

The Bagci Lab received a blue ribbon for one of the best posters at the 2024 Children’s Brain Tumor Summit in Arlington, a premier multidisciplinary meeting in Pediatric Neuro-oncology. Max Bengtsson, an undergraduate student supervised by Dr. Elif Keles, Dr.Angela Waanders and  Dr. Ulas Bagci, was the first author. The project focused on segmenting pediatric brain tumors in MRIs, specifically targeting all four subregions, using deep learning algorithms on real-world data.

Our group will present three abstract in the annual meeting of APA. Papers are the following: 1) Deep learning based automated pancreas segmentation and volumetry measurement in patients with acute pancreatitis, 2) Patients with acute pancreatitis have significant decline in pancreas volume over time , and 3) The association of volumetric changes and disease severity with endocrine dysfunction following acute pancreatitis.

20th edition of the IEEE International Conference on Advanced Video and Signal-Based Surveillance (AVSS), taking place in the beautiful city of Niagara Falls, ON, Canada. Bagci lab has a paper to present (oral). The paper can be found in ArXiv link!

Dr. Bagci gave an invited talk at Wake Forest University, hosted by Prof. Metin Gurcan. In this talk, titled “Trustworthy AI for Imaging-based Diagnosis”, he focused on the failures of deep learning / AI algorithms and proposed several approaches to increase robustness of AI powered medical imaging systems.

Dr. Bagci gave an invited talk in IEEE ISBI 2024, MRI Beyond the Norm Special Session.
Organized by Prof. Rad and Dr. Fujimoto, Dr. Bagci was one of the keynotes in this highly regarded special session. Other speakers were Dr. Wald from Harvard Medical School, Dr de Lara from Harvard Medical School, and Dr Sadleir from Arizona State University. Dr. Bagci talked about recent progress and future aspects of “Generative AI for pre- and post-processing MRI”.

The Segment Anything Model (SAM) originally designed for general-purpose segmentation tasks has been used recently for polyp segmentation. Nonetheless fine-tuning SAM with data from new imaging centers or clinics poses significant challenges. To this end we utilize variable perturbed bounding box prompts (BBP) to enrich the learning context and enhance the model’s robustness to BBP perturbations during inference. Rigorous experiments on polyp segmentation benchmarks reveal that our variable BBP perturbation significantly improves model resilience. Notably on Kvasir 1-shot fine-tuning boosts the DICE score by 20% and 37% with 50 and 100-pixel BBP perturbations during inference respectively. Our results motivate the broader applicability of our PP-SAM for other medical imaging tasks with limited samples.

Although artificial intelligence (AI) based computer-aided diagnosis systems have been shown to be useful in medical image analysis, current deep learning methods still suffer (1) challenging localization of lesions and (2) low-efficient clinical practice (3) lack of expert knowledge. Eye tracking research is important in computer vision because it can help us understand how humans interact with the visual world. Specifically for high-risk applications, such as medical imaging, eye tracking can help us comprehend how radiologists and other medical professionals search, analyze, and interpret images for diagnostic and clinical purposes. In this study, we investigate how to apply eye tracking techniques to real clinical practice, which builds a time-efficient, robust, and radiologist-centered computer-aided diagnosis system.

In this paper, we review recently developed machine learning and deep learning-based solutions for neonatology applications. We systematically evaluate the roles of both classical machine learning and deep learning in neonatology applications, define the methodologies, including algorithmic developments, and describe the remaining challenges in the assessment of neonatal diseases by using PRISMA 2020 guidelines. We also discuss possible directions for new AI models and the future of neonatology with the rising power of AI, suggesting roadmaps for the integration of AI into neonatal intensive care units

Healthy-to-Patients Domain-Adaptive Deep Learning for Time-Resolved Segmentation of Left Atrium in Short-Axis Cine MRI Images

1. A review of prognostic prediction of coronary artery disease patients with myocardial perfusion scintigraphy and artificial intelligence.
2. Predicting Short Term Mortality In Patients With Acute Pulmonary Embolism With Deep Learning.

Our joint partnership between the Machine and Hybrid Intelligence at Northwestern University and the Mayo Clinic Pancreas Cancer Early Detection team of Michael B Wallace continues to produce major advances toward our goal of detecting the earliest and treatable forms of pancreatic cancer. Recent breakthroughs include highly reliable systems for pancreatic segmentation on CT and MRI, which should improve the detection of very early malignant change and solid and cystic pancreatic tumors.

This article outlines the fundamental statistical issues in FL, tackles device-related problems, addresses security challenges, and navigates the complexity of privacy concerns, all while highlighting its transformative potential in the medical field. Our study primarily focuses on medical applications of FL, particularly in the context of global cancer diagnosis. We highlight the potential of FL to enable computer-aided diagnosis tools that address this challenge with greater effectiveness than traditional data-driven methods. Recent literature has shown that FL models are robust and generalize well to new data, which is essential for medical applications.

paper link: https://ieeexplore.ieee.org/document/10304218

This study presents an innovative approach for automatic road detection with deep learning, employing fusion strategies to utilize both lower-resolution satellite imagery and GPS trajectory data, a concept never explored before.

paper link: https://link.springer.com/article/10.1007/s12145-023-01201-6

Our research advances domain generalization by integrating soft-thresholding with Fourier-transform techniques, reducing background interference in retinal fundus image segmentation. This method, proven on public datasets, markedly enhances model generalization and accuracy, surpassing current standards.

paper link: https://arxiv.org/html/2309.09866v3

The COVID-19 pandemic, alongside the widespread availability of graphics processing units, has accelerated advancements in AI and medical imaging, offering new avenues to enhance patient care despite initial challenges in model performance on novel data. These developments not only promise to improve future medical AI applications but also underscore the potential of AI in infectious disease imaging research, addressing evidence gaps and fostering innovative solutions in the field.

paper link: https://academic.oup.com/jid/article/228

The aim of this study was to investigate the prognostic relevance of PIV in predicting in-hospital mortality in acute pulmonary embolism (PE) patients and to compare it with the well-known risk scoring system, PE severity index (PESI), which is commonly used for a short-term mortality prediction in such patients.

We recently released a multi-center gastrointestinal tract dataset with 8K images from 27 classes. We hope that it opens new possibilities in GI endoscopy & colonoscopy research.

Dataset link: https://osf.io/84e7f/
paper link: https://arxiv.org/pdf/2307.08140.pdf

This study introduces eye gaze as a novel interactive prompt for image segmentation, different than previous model-based applications. Specifically, leveraging the real-time interactive prompting feature of the recently proposed Segment Anything Model (SAM), we present the GazeSAM system to enable users to collect target segmentation masks by simply looking at the region of interest.

Github Page: https://ukaukaaaa.github.io/projects/gazesam/gazesam.html
paper link: https://openreview.net/pdf?id=hJ5DREWdjs

The annual Research Day at Northwestern is a campuswide event to promote faculty and trainee development by sharing exciting research and conversations with colleagues. This year we have 11 posters to present:

1. GazeSAM: What You See is What You Segment
2. Deep Learning Algorithm for Accurate and Real-time Liver Segmentation
3. Visual Explanations for Radiology AI Applications with Information Bottleneck
4. Ensemble learning with Residual Transformer for Brian Tumor Segmentation
5. Monkeypox Diagnosis With Interpretable Deep Learning
6. A Fully Automatic AI System for Pancreas Segmentation from
   Multicenter MRI Scans
7. AI-Based Pancreatic Cancer Risk Classification in IPMN Patients
8. Fast Forward Forward: Analytic Training for Deep Neural Networks
9. Computer-Aided Automatic Measurement of Pancreatic Cysts
10. Cognitive Perspectives on Complexity of Diagnostic Search and Its Analogy in Artificial Intelligence
11. Pancreatic Cancer: Decoding Risks with Duct Diameter, Cyst Size, and Abdominal Circumference

At the International Symposium on Current Advances in Medical Imaging Techniques, Prof. Bagci highlighted our recent findings from several high-risk, high-gain projects, funded by NIH.

Two papers got accepted by Winter Conference on Applications of Computer Vision (WACV2024).

Our lab MHIL was recently featured in Northwestern Medicine’s newsletter, highlighting progress in AI for Pancreatic Diseases.  Read more details at:

GastroVision: A Multi-class Endoscopy Image Dataset for Computer Aided Gastrointestinal Disease Detection.

by Debesh J. et al.

We’re proud to announce that Dr. Bagci represented our lab as a panelist in the ML4MHD (Machine Learning for Multimodal Healthcare Data) session at ICML 2023. Chaired by Dr. Pallavi Tiwari, the panel also featured esteemed experts like Dr. Daniel Rueckert and Dr. Xiaoxiao Li.

We were honored to host Dr. Bo Zhou, a distinguished final-year Ph.D. Candidate from the Department of Biomedical Engineering at Yale University, for a captivating seminar on AI applications in Multi-modal Nuclear Medicine (NM) Imaging. Dr. Zhou’s extensive research addresses the critical challenges of high radiation doses, image quality degradation, and prolonged acquisition times in imaging modalities like PET-CT and PET-MRI. His innovative deep-learning strategies aim to enhance the safety, efficiency, and quality of these imaging techniques, potentially elevating their clinical utility. Dr. Zhou, with a commendable academic and research background marked by awards and over 50 publications, shared insights that promise to redefine the boundaries of medical imaging. Explore more about Dr. Zhou’s work on his homepage.

Congratulations to our visiting student, Harmony, on her stellar presentation at Northwestern University! Her insightful study on “Pancreatic Cancer: Decoding Risks with Duct Diameter, Cyst Size, and Abdominal Circumference” showcased an in-depth analysis and a fresh perspective, further contributing to the evolving discourse on pancreatic cancer research. We’re proud of her achievements and commend her dedication to this vital field of study.

On July 21st, our lab resonated with inspiration and forward-thinking, thanks to our remarkable women scientists at the forefront of medical imaging research. Their tales of dedication, challenges, and groundbreaking discoveries were the highlight of this special gathering.

A highly-anticipated keynote speech is set to be given at the upcoming IEEE ICECCME 2023 conference. It will focus on the failures of deep learning / AI algorithms and propose several approaches to increase robustness of AI powered medical imaging systems.

We are excited to announce the acceptance of our MIDL 2023 paper for publication! Stay tuned for more details on the study.

We are honored to welcome Dr. Mubarak Shah, a renowned professor from the University of Central Florida (UCF), to our lab. During his visit, we will share our latest research findings and engage in fruitful discussions to explore potential collaborations and future directions in our respective fields.

The new book co-edited by Prof. Bagci, “AI in Clinical Medicine: A Practical Guide for Healthcare Professionals,” is now available for purchase in both physical and electronic formats from Wiley. Get your copy at :

On March 16, 2023, an insightful talk was delivered at the University of Pennsylvania’s Radiology department, delving into the diagnosis and risk stratification of pancreatic cysts using explainable AI techniques.

On February 15, 2023, an invited talk was given at the University of Wisconsin Madison’s BME department, highlighting the significance of trustworthy AI in imaging-based diagnosis.

An improved method of performing object segmentation and classification that reduces the memory required to perform these tasks, while increasing predictive accuracy.

Prof. Bagci will be sharing expertise at the MICCAI 2023 workshop focusing on medical image learning with limited and noisy data.

Bagci lab received a grant funded by RSNA Emerging Issues – Long-Term COVID Effects EILTC2208, entitled “PASC Pulmonary Fibrosis Prediction with Deep Learning and Multimodal Data.”

Our team attended RSNA 2022 at Chicago. Ugur and Debesh are presenting our recent progress.