David Fuentes

Adrian Celaya, Beatrice Riviere, David Fuentes

Within medical imaging segmentation, the Dice coefficient and Hausdorff-based metrics are standard measures of success for deep learning models. However, modern loss functions for medical image segmentation often only consider the Dice coefficient or similar region-based metrics during training. As a result, segmentation architectures trained over such loss functions run the risk of achieving high accuracy for the Dice coefficient but low accuracy for Hausdorff-based metrics. Low accuracy on Hausdorff-based metrics can be problematic for applications such as tumor segmentation, where such benchmarks are crucial. For example, high Dice scores accompanied by significant Hausdorff errors could indicate that the predictions fail to detect small tumors. We propose the Generalized Surface Loss function, a novel loss function to minimize Hausdorff-based metrics with more desirable numerical properties than current methods and with weighting terms for class imbalance. Our loss function outperforms other losses when tested on the LiTS and BraTS datasets using the state-of-the-art nnUNet architecture. These results suggest we can improve medical imaging segmentation accuracy with our novel loss function.

Adrian Celaya, Beatrice Riviere, David Fuentes

Accurate medical imaging segmentation is critical for precise and effective medical interventions. However, despite the success of convolutional neural networks (CNNs) in medical image segmentation, they still face challenges in handling fine-scale features and variations in image scales. These challenges are particularly evident in complex and challenging segmentation tasks, such as the BraTS multi-label brain tumor segmentation challenge. In this task, accurately segmenting the various tumor sub-components, which vary significantly in size and shape, remains a significant challenge, with even state-of-the-art methods producing substantial errors. Therefore, we propose two architectures, FMG-Net and W-Net, that incorporate the principles of geometric multigrid methods for solving linear systems of equations into CNNs to address these challenges. Our experiments on the BraTS 2020 dataset demonstrate that both FMG-Net and W-Net outperform the widely used U-Net architecture regarding tumor subcomponent segmentation accuracy and training efficiency. These findings highlight the potential of incorporating the principles of multigrid methods into CNNs to improve the accuracy and efficiency of medical imaging segmentation.

Adrian Celaya, Evan Lim, Rachel Glenn et al.

Medical imaging segmentation is a highly active area of research, with deep learning-based methods achieving state-of-the-art results in several benchmarks. However, the lack of standardized tools for training, testing, and evaluating new methods makes the comparison of methods difficult. To address this, we introduce the Medical Imaging Segmentation Toolkit (MIST), a simple, modular, and end-to-end medical imaging segmentation framework designed to facilitate consistent training, testing, and evaluation of deep learning-based medical imaging segmentation methods. MIST standardizes data analysis, preprocessing, and evaluation pipelines, accommodating multiple architectures and loss functions. This standardization ensures reproducible and fair comparisons across different methods. We detail MIST's data format requirements, pipelines, and auxiliary features and demonstrate its efficacy using the BraTS Adult Glioma Post-Treatment Challenge dataset. Our results highlight MIST's ability to produce accurate segmentation masks and its scalability across multiple GPUs, showcasing its potential as a powerful tool for future medical imaging research and development.

Wolfgang Stefan, David Fuentes, Erol Yeniaras et al.

MR images have a magnitude and a phase, but in almost all clinical applications only the magnitude images are used, because the phase images have a smooth but strong background signal that masks useful information. The phase contains information such as the temperature during thermal ablation, and the iron content of brain tissue. We describe a novel method to suppress the background that is based on higher order edge detection and sparse image representation.

David Fuentes, Diana McSpadden, Sodiq Adewole

In this paper, we propose a distributed Generative Adversarial Networks (discGANs) to generate synthetic tabular data specific to the healthcare domain. While using GANs to generate images has been well studied, little to no attention has been given to generation of tabular data. Modeling distributions of discrete and continuous tabular data is a non-trivial task with high utility. We applied discGAN to model non-Gaussian multi-modal healthcare data. We generated 249,000 synthetic records from original 2,027 eICU dataset. We evaluated the performance of the model using machine learning efficacy, the Kolmogorov-Smirnov (KS) test for continuous variables and chi-squared test for discrete variables. Our results show that discGAN was able to generate data with distributions similar to the real data.

Adrian Celaya, Keegan Kirk, David Fuentes et al.

In recent years, there has been a growing interest in leveraging deep learning and neural networks to address scientific problems, particularly in solving partial differential equations (PDEs). However, many neural network-based methods like PINNs rely on auto differentiation and sampling collocation points, leading to a lack of interpretability and lower accuracy than traditional numerical methods. As a result, we propose a fully unsupervised approach, requiring no training data, to estimate finite difference solutions for PDEs directly via small linear convolutional neural networks. Our proposed approach uses substantially fewer parameters than similar finite difference-based approaches while also demonstrating comparable accuracy to the true solution for several selected elliptic and parabolic problems compared to the finite difference method.

Awj Twam, Adrian E. Celaya, Megan C. Jacobsen et al.

Cervical cancer remains the fourth most common malignancy amongst women worldwide.1 Concurrent chemoradiotherapy (CRT) serves as the mainstay definitive treatment regimen for locally advanced cervical cancers and includes external beam radiation followed by brachytherapy.2 Integral to radiotherapy treatment planning is the routine contouring of both the target tumor at the level of the cervix, associated gynecologic anatomy and the adjacent organs at risk (OARs). However, manual contouring of these structures is both time and labor intensive and associated with known interobserver variability that can impact treatment outcomes. While multiple tools have been developed to automatically segment OARs and the high-risk clinical tumor volume (HR-CTV) using computed tomography (CT) images,3,4,5,6 the development of deep learning-based tumor segmentation tools using routine T2-weighted (T2w) magnetic resonance imaging (MRI) addresses an unmet clinical need to improve the routine contouring of both anatomical structures and cervical cancers, thereby increasing quality and consistency of radiotherapy planning. This work applied a novel deep-learning model (PocketNet) to segment the cervix, vagina, uterus, and tumor(s) on T2w MRI. The performance of the PocketNet architecture was evaluated, when trained on data via five-fold cross validation. PocketNet achieved a mean Dice-Sorensen similarity coefficient (DSC) exceeding 70% for tumor segmentation and 80% for organ segmentation. Validation on a publicly available dataset from The Cancer Imaging Archive (TCIA) demonstrated the models robustness, achieving DSC scores of 67.3% for tumor segmentation and 80.8% for organ segmentation. These results suggest that PocketNet is robust to variations in contrast protocols, providing reliable segmentation of the regions of interest.

Cito Balsells, Beatrice Riviere, David Fuentes

We formulate truncated singular value decompositions of entire convolutional neural networks. We demonstrate the computed left and right singular vectors are useful in identifying which images the convolutional neural network is likely to perform poorly on. To create this diagnostic tool, we define two metrics: the Right Projection Ratio and the Left Projection Ratio. The Right (Left) Projection Ratio evaluates the fidelity of the projection of an image (label) onto the computed right (left) singular vectors. We observe that both ratios are able to identify the presence of class imbalance for an image classification problem. Additionally, the Right Projection Ratio, which only requires unlabeled data, is found to be correlated to the model's performance when applied to image segmentation. This suggests the Right Projection Ratio could be a useful metric to estimate how likely the model is to perform well on a sample.

Adrian Celaya, Yimo Wang, David Fuentes et al.

In recent years, there has been an increasing interest in using deep learning and neural networks to tackle scientific problems, particularly in solving partial differential equations (PDEs). However, many neural network-based methods, such as physics-informed neural networks, depend on automatic differentiation and the sampling of collocation points, which can result in a lack of interpretability and lower accuracy compared to traditional numerical methods. To address this issue, we propose two approaches for learning discontinuous Galerkin solutions to PDEs using small linear convolutional neural networks. Our first approach is supervised and depends on labeled data, while our second approach is unsupervised and does not rely on any training data. In both cases, our methods use substantially fewer parameters than similar numerics-based neural networks while also demonstrating comparable accuracy to the true and DG solutions for elliptic problems.

Adrian Celaya, David Fuentes, Beatrice Riviere

Physics-informed neural networks (PINNs) have gained significant attention for solving forward and inverse problems related to partial differential equations (PDEs). While advancements in loss functions and network architectures have improved PINN accuracy, the impact of collocation point sampling on their performance remains underexplored. Fixed sampling methods, such as uniform random sampling and equispaced grids, can fail to capture critical regions with high solution gradients, limiting their effectiveness for complex PDEs. Adaptive methods, inspired by adaptive mesh refinement from traditional numerical methods, address this by dynamically updating collocation points during training but may overlook residual dynamics between updates, potentially losing valuable information. To overcome this limitation, we propose two adaptive collocation point selection strategies utilizing the QR Discrete Empirical Interpolation Method (QR-DEIM), a reduced-order modeling technique for efficiently approximating nonlinear functions. Our results on benchmark PDEs demonstrate that our QR-DEIM-based approaches improve PINN accuracy compared to existing methods, offering a promising direction for adaptive collocation point strategies.

Alec Reinhardt, Newsha Nikzad, Raven J. Hollis et al.

Diagnostic imaging has gained prominence as potential biomarkers for early detection and diagnosis in a diverse array of disorders including cancer. However, existing methods routinely face challenges arising from various factors such as image heterogeneity. We develop a novel imaging-based distributional data analysis (DDA) approach that incorporates the probability (quantile) distribution of the pixel-level features as covariates. The proposed approach uses a smoothed quantile distribution (via a suitable basis representation) as functional predictors in a scalar-on-functional quantile regression model. Some distinctive features of the proposed approach include the ability to: (i) account for heterogeneity within the image; (ii) incorporate granular information spanning the entire distribution; and (iii) tackle variability in image sizes for unregistered images in cancer applications. Our primary goal is risk prediction in Hepatocellular carcinoma that is achieved via predicting the change in tumor grades at post-diagnostic visits using pre-diagnostic enhancement pattern mapping (EPM) images of the liver. Along the way, the proposed DDA approach is also used for case versus control diagnosis and risk stratification objectives. Our analysis reveals that when coupled with global structural radiomics features derived from the corresponding T1-MRI scans, the proposed smoothed quantile distributions derived from EPM images showed considerable improvements in sensitivity and comparable specificity in contrast to classification based on routinely used summary measures that do not account for image heterogeneity. Given that there are limited predictive modeling approaches based on heterogeneous images in cancer, the proposed method is expected to provide considerable advantages in image-based early detection and risk prediction.

Rajarajeswari Muthusivarajan, Adrian Celaya, Joshua P. Yung et al.

Deep neural networks with multilevel connections process input data in complex ways to learn the information.A networks learning efficiency depends not only on the complex neural network architecture but also on the input training images.Medical image segmentation with deep neural networks for skull stripping or tumor segmentation from magnetic resonance images enables learning both global and local features of the images.Though medical images are collected in a controlled environment,there may be artifacts or equipment based variance that cause inherent bias in the input set.In this study, we investigated the correlation between the image quality metrics of MR images with the neural network segmentation accuracy.For that we have used the 3D DenseNet architecture and let the network trained on the same input but applying different methodologies to select the training data set based on the IQM values.The difference in the segmentation accuracy between models based on the random training inputs with IQM based training inputs shed light on the role of image quality metrics on segmentation accuracy.By running the image quality metrics to choose the training inputs,further we may tune the learning efficiency of the network and the segmentation accuracy.

Adrian Celaya, Jonas A. Actor, Rajarajeswari Muthusivarajan et al.

Medical imaging deep learning models are often large and complex, requiring specialized hardware to train and evaluate these models. To address such issues, we propose the PocketNet paradigm to reduce the size of deep learning models by throttling the growth of the number of channels in convolutional neural networks. We demonstrate that, for a range of segmentation and classification tasks, PocketNet architectures produce results comparable to that of conventional neural networks while reducing the number of parameters by multiple orders of magnitude, using up to 90% less GPU memory, and speeding up training times by up to 40%, thereby allowing such models to be trained and deployed in resource-constrained settings.

Reza Madankan, Wolfgang Stefan, Samuel Fahrenholtz et al.

An accelerated model-based information theoretic approach is presented to perform the task of Magnetic Resonance (MR) thermal image reconstruction from a limited number of observed samples on k-space. The key idea of the proposed approach is to utilize information theoretic techniques to optimally detect samples of k-space that are information rich with respect to a model of the thermal data acquisition. These highly informative k-space samples are then used to refine the mathematical model and reconstruct the image. The information theoretic reconstruction is demonstrated retrospectively in data acquired during MR guided Laser Induced Thermal Therapy (MRgLITT) procedures. The approach demonstrates that locations of high-information content with respect to a model based reconstruction of MR thermometry may be quantitatively identified. The predicted locations of high-information content are sorted and retrospectively extracted from the fully sampled k-space measurements data set. The effect of interactively increasing the predicted number of data points used in the subsampled reconstruction is quantified using the L2-norm of the distance between the subsampled and fully sampled reconstruction. Performance of the proposed approach is also compared with clinically available subsampling techniques (rectilinear subsampling and variable-density Poisson disk undersampling). It is shown that the proposed subsampling scheme results in accurate reconstructions using small fraction of k-space points and suggest that the reconstruction technique may be useful in improving the efficiency of the thermometry data temporal resolution.