About

Frank J. Brooks is a Research Assistant Professor in Bioengineering at the University of Illinois Urbana-Champaign, with a background in physics and mathematics. He earned his Ph.D. in Physics from Washington University in St. Louis in 2008, along with a Master of Science in Physics from the University of Missouri and a Bachelor of Arts in Mathematics from the University of St. Louis. His research focuses on medical imaging, computational tools, and the application of AI in medicine, including the development of new computational methods for medical image analysis and the investigation of hallucinations in tomographic image reconstruction. Brooks has contributed to understanding the robustness of learning-based methods for quantitative phase retrieval, assessing generative adversarial networks for medical image statistics, and evaluating high-order spatial arrangements in diagnostic images. He has also been involved in interdisciplinary collaborations related to cancer imaging, optoacoustic tomography, and lung imaging, among others. His work aims to improve diagnostic accuracy and develop innovative imaging techniques through advanced computational and AI methodologies.

Research topics

• Machine Learning
• Artificial Intelligence
• Computer Science
• Computer vision
• Statistics
• Mathematics

Selected publications

• A Taxonomy of Machine Hallucination in Radiology

  Radiology Artificial Intelligence · 2026-03-01

  articleOpen access1st authorCorresponding

  A taxonomy of machine hallucination is proposed to codify intended interpretations of hallucination when evaluating the utility and trustworthiness of generative artificial intelligence deployed in radiology.

  PublisherOA PDFDOI

• On the Utility of Virtual Staining for Downstream Applications as it relates to Task Network Capacity

  ArXiv.org · 2025-07-31

  preprintOpen access

  Virtual staining, or in-silico-labeling, has been proposed to computationally generate synthetic fluorescence images from label-free images by use of deep learning-based image-to-image translation networks. In most reported studies, virtually stained images have been assessed only using traditional image quality measures such as structural similarity or signal-to-noise ratio. However, in biomedical imaging, images are typically acquired to facilitate an image-based inference, which we refer to as a downstream biological or clinical task. This study systematically investigates the utility of virtual staining for facilitating clinically relevant downstream tasks (like segmentation or classification) with consideration of the capacity of the deep neural networks employed to perform the tasks. Comprehensive empirical evaluations were conducted using biological datasets, assessing task performance by use of label-free, virtually stained, and ground truth fluorescence images. The results demonstrated that the utility of virtual staining is largely dependent on the ability of the segmentation or classification task network to extract meaningful task-relevant information, which is related to the concept of network capacity. Examples are provided in which virtual staining does not improve, or even degrades, segmentation or classification performance when the capacity of the associated task network is sufficiently large. The results demonstrate that task network capacity should be considered when deciding whether to perform virtual staining.

  PublisherOA PDFDOI

• <i>In vivo</i> volumetric visualization and quantification of cumulus expansion in mice with intravital optical coherence tomography

  Biomedical Optics Express · 2025-11-25

  articleOpen access

  Ovulation is preceded by a critical physiological process known as cumulus expansion, during which the cumulus cell layer surrounding the oocyte undergoes structural remodeling. Despite the recognized importance of this process for reproductive success, live quantitative imaging of cumulus expansion has not been previously achieved due to limitations of current imaging technologies for deeply located ovaries. In this study, we employed intravital optical coherence tomography for three-dimensional visualization of mouse follicles containing cumulus-oocyte complexes (COC) within the physiological context of the ovary, both ex vivo and in vivo . This method enabled time-lapse measurement of cumulus layer thickness and COC volume. Longitudinal imaging in live mice revealed the physiological spatiotemporal dynamics of cumulus matrix expansion preceding ovulation. These findings establish a novel in vivo platform for dynamic investigation of previously inaccessible preovulatory processes within a physiological context.

  PublisherDOI

• On the Utility of Virtual Staining for Downstream Applications as it relates to Task Network Capacity

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-08-06 · 1 citations

  preprintOpen access

  Virtual staining, or in-silico-labeling, has been proposed to computationally generate synthetic fluorescence images from label-free images by use of deep learning-based image-to-image translation networks. In most reported studies, virtually stained images have been assessed only using traditional image quality measures such as structural similarity or signal-to-noise ratio. However, in biomedical imaging, images are typically acquired to facilitate an image-based inference, which we refer to as a downstream biological or clinical task. This study systematically investigates the utility of virtual staining for facilitating clinically relevant downstream tasks (like segmentation or classification) with consideration of the capacity of the deep neural networks employed to perform the tasks. Comprehensive empirical evaluations were conducted using biological datasets, assessing task performance by use of label-free, virtually stained, and ground truth fluorescence images. The results demonstrated that the utility of virtual staining is largely dependent on the ability of the segmentation or classification task network to extract meaningful task-relevant information, which is related to the concept of network capacity. Examples are provided in which virtual staining does not improve, or even degrades, segmentation or classification performance when the capacity of the associated task network is sufficiently large. The results demonstrate that task network capacity should be considered when deciding whether to perform virtual staining.

  PublisherOA PDFDOI

• On the utility of virtual staining for downstream applications as it relates to task network capacity

  Biomedical Optics Express · 2025-09-22

  articleOpen access

  Virtual staining, or in-silico labeling, employs deep learning-based image-to-image translation networks to computationally generate synthetic stained images, either from label-free inputs or by digitally transforming one stain type into another. In most reported studies, virtually stained images have been assessed only using traditional image quality measures such as structural similarity or signal-to-noise ratio. However, in biomedical imaging, images are typically acquired to facilitate an image-based inference, which we refer to as a downstream biological or clinical task. This study systematically investigates the utility of virtual staining for facilitating clinically relevant downstream tasks (like segmentation or classification) with consideration of the capacity of the deep neural networks employed to perform the tasks. Comprehensive empirical evaluations were conducted using biological datasets, assessing task performance by use of label-free, virtually stained, and ground truth fluorescence images. The results demonstrated that the utility of virtual staining is largely dependent on the ability of the segmentation or classification task network to extract meaningful task-relevant information, which is related to the concept of network capacity. Examples are provided in which virtual staining does not improve, or even degrades, segmentation or classification performance when the capacity of the associated task network is sufficiently large. The results demonstrate that task network capacity should be considered when deciding whether to perform virtual staining.

  PublisherDOI

• A method for evaluating deep generative models of images for hallucinations in high-order spatial context

  Pattern Recognition Letters · 2024-09-03 · 3 citations

  articleSenior authorCorresponding
  PublisherDOI

• Evaluating the capacity of a diffusion generative model to reproduce spatial context relevant to diagnostic imaging

  2024-03-29

  articleSenior author

  The rapid evolution of deep generative models (DGMs) has highlighted their great potential in medical imaging research. Recently, it has been claimed that a diffusion generative model: denoising diffusion probabilistic model (DDPM), performs better at image synthesis than the previously popular DGMs: generative adversarial networks (GANs). However, this claim is based on evaluations employing measures intended for natural images, and thus, does not resolve questions about their relevance to medical imaging tasks. To partially address this problem, we performed a series of assessments to evaluate the ability of a DDPM to reproduce diagnostically relevant spatial context. Our findings show that in all our studies, although context was generally well replicated in DDPM-generated ensembles, it was never perfectly reproduced in the entire ensemble.

  PublisherDOI

• Exploring a method to evaluate image-conditioned deep generative models for their capacity to reproduce domain-relevant spatial context

  2024-02-16

  articleSenior author

  In domains such as biomedical imaging, the evaluation of deep generative models (DGMs) for image-to-image translation tasks is additionally challenged by the need for substantial domain expertise, even for visual evaluation. To partially circumvent this problem, we propose a data-driven, human interpretable method to evaluate image-conditioned DGMs for the reproducibility of domain-relevant spatial context before the DGMs are considered for diagnostic tasks and real-world deployment.

  PublisherDOI

• Report on the AAPM grand challenge on deep generative modeling for learning medical image statistics

  Medical Physics · 2024-10-24 · 9 citations

  articleOpen access

  BACKGROUND: The findings of the 2023 AAPM Grand Challenge on Deep Generative Modeling for Learning Medical Image Statistics are reported in this Special Report. PURPOSE: The goal of this challenge was to promote the development of deep generative models for medical imaging and to emphasize the need for their domain-relevant assessments via the analysis of relevant image statistics. METHODS: 512. For the evaluation of submissions to the Challenge, an ensemble of 10 000 DGM-generated images from each submission was employed. The evaluation procedure consisted of two stages. In the first stage, a preliminary check for memorization and image quality (via the Fréchet Inception Distance [FID]) was performed. Submissions that passed the first stage were then evaluated for the reproducibility of image statistics corresponding to several feature families including texture, morphology, image moments, fractal statistics, and skeleton statistics. A summary measure in this feature space was employed to rank the submissions. Additional analyses of submissions was performed to assess DGM performance specific to individual feature families, the four classes in the training data, and also to identify various artifacts. RESULTS: Fifty-eight submissions from 12 unique users were received for this Challenge. Out of these 12 submissions, 9 submissions passed the first stage of evaluation and were eligible for ranking. The top-ranked submission employed a conditional latent diffusion model, whereas the joint runners-up employed a generative adversarial network, followed by another network for image superresolution. In general, we observed that the overall ranking of the top 9 submissions according to our evaluation method (i) did not match the FID-based ranking, and (ii) differed with respect to individual feature families. Another important finding from our additional analyses was that different DGMs demonstrated similar kinds of artifacts. CONCLUSIONS: This Grand Challenge highlighted the need for domain-specific evaluation to further DGM design as well as deployment. It also demonstrated that the specification of a DGM may differ depending on its intended use.

  PublisherOA PDFDOI

• Assessing the Capacity of a Denoising Diffusion Probabilistic Model to Reproduce Spatial Context

  IEEE Transactions on Medical Imaging · 2024-06-14 · 13 citations

  articleOpen accessSenior author

  Diffusion models have emerged as a popular family of deep generative models (DGMs). In the literature, it has been claimed that one class of diffusion models-denoising diffusion probabilistic models (DDPMs)-demonstrate superior image synthesis performance as compared to generative adversarial networks (GANs). To date, these claims have been evaluated using either ensemble-based methods designed for natural images, or conventional measures of image quality such as structural similarity. However, there remains an important need to understand the extent to which DDPMs can reliably learn medical imaging domain-relevant information, which is referred to as 'spatial context' in this work. To address this, a systematic assessment of the ability of DDPMs to learn spatial context relevant to medical imaging applications is reported for the first time. A key aspect of the studies is the use of stochastic context models (SCMs) to produce training data. In this way, the ability of the DDPMs to reliably reproduce spatial context can be quantitatively assessed by use of post-hoc image analyses. Error-rates in DDPM-generated ensembles are reported, and compared to those corresponding to other modern DGMs. The studies reveal new and important insights regarding the capacity of DDPMs to learn spatial context. Notably, the results demonstrate that DDPMs hold significant capacity for generating contextually correct images that are 'interpolated' between training samples, which may benefit data-augmentation tasks in ways that GANs cannot.

  PublisherDOI

Frequent coauthors

• Mark A. Anastasio

  University of Illinois Urbana-Champaign

  39 shared

• Sayantan Bhadra

  National Institutes of Health Clinical Center

  12 shared

• Perry W. Grigsby

  Washington University in St. Louis

  11 shared

• Weimin Zhou

  10 shared

• Varun A. Kelkar

  University of Illinois Urbana-Champaign

  9 shared

• R. Deshpande

  8 shared

• Hua Li

  Washington University in St. Louis

  8 shared

• Dimitrios S. Gotsis

  University of Illinois Urbana-Champaign

  5 shared

Labs

• Frank J Brooks LabPI