Search

Machine learning (ML) applications in medical artificial intelligence (AI) systems have shifted from traditional and statistical methods to increasing application of deep learning models. This survey navigates the current landscape of multimodal ML, focusing on its profound impact on medical image analysis and clinical decision support systems. Emphasizing challenges and innovations in addressing multimodal representation, fusion, translation, alignment, and co-learning, the paper explores the transformative potential of multimodal models for clinical predictions. It also highlights the need for principled assessments and practical implementation of such models, bringing attention to the dynamics between decision support systems and healthcare providers and personnel. Despite advancements, challenges such as data biases and the scarcity of "big data" in many biomedical domains persist. We conclude with a discussion on principled innovation and collaborative efforts to further the mission of seamless integration of multimodal ML models into biomedical practice. [ABSTRACT FROM AUTHOR]

Glioblastoma is the most common malignant brain tumor with less than 15 months median survival. To aid prognosis, there is a need for decision tools that leverage diagnostic modalities such as MRI to inform survival. In this study, we examine higher-order spatial proximity characteristics from habitats and propose two graph-based methods (minimum spanning tree and graph run-length matrix) to characterize spatial heterogeneity over tumor MRI-derived intensity habitats and assess their relationships with overall survival as well as the immune signature status of patients with glioblastoma. A data set of 74 patients was studied based on the availability of post-contrast T1-weighted and T2-weighted fluid attenuated inversion recovery (FLAIR) image data in The Cancer Image Archive (TCIA). We assessed the predictive value of MST- and GRLM-derived features from 2D images for prediction of 12-month survival status and immune signature status of patients with glioblastoma via a receiver operating characteristic curve analysis. For 12-month survival prediction using MST-based method, sensitivity and specificity were 0.82 and 0.79 respectively. For GRLM-based method, sensitivity and specificity were 0.73 and 0.77 respectively. For immune status, sensitivity and specificity were 0.91 and 0.69, respectively, for the GRLM-based method with an immune effector. Our results show that the proposed MST- and GRLM-derived features are predictive of 12-month survival status as well as the immune signature status of patients with glioblastoma. To our knowledge, this is the first application of MST- and GRLM-based proximity analyses for the study of radiologically-defined tumor habitats in glioblastoma. [ABSTRACT FROM AUTHOR]

Glioblastoma is the most common malignant brain tumor with less than 15 months median survival. To aid prognosis, there is a need for decision tools that leverage diagnostic modalities such as MRI to inform survival. In this study, we examine higher-order spatial proximity characteristics from habitats and propose two graph-based methods (minimum spanning tree and graph run-length matrix) to characterize spatial heterogeneity over tumor MRI-derived intensity habitats and assess their relationships with overall survival as well as the immune signature status of patients with glioblastoma. A data set of 74 patients was studied based on the availability of post-contrast T1-weighted and T2-weighted fluid attenuated inversion recovery (FLAIR) image data in The Cancer Image Archive (TCIA). We assessed the predictive value of MST- and GRLM-derived features from 2D images for prediction of 12-month survival status and immune signature status of patients with glioblastoma via a receiver operating characteristic curve analysis. For 12-month survival prediction using MST-based method, sensitivity and specificity were 0.82 and 0.79 respectively. For GRLM-based method, sensitivity and specificity were 0.73 and 0.77 respectively. For immune status, sensitivity and specificity were 0.91 and 0.69, respectively, for the GRLM-based method with an immune effector. Our results show that the proposed MST- and GRLM-derived features are predictive of 12-month survival status as well as the immune signature status of patients with glioblastoma. To our knowledge, this is the first application of MST- and GRLM-based proximity analyses for the study of radiologically-defined tumor habitats in glioblastoma. [ABSTRACT FROM AUTHOR]

Machine learning applied to digital pathology has been increasingly used to assess kidney function and diagnose the underlying cause of chronic kidney disease (CKD). We developed a novel computational framework, clustering-based spatial analysis (CluSA), that leverages unsupervised learning to learn spatial relationships between local visual patterns in kidney tissue. This framework minimizes the need for time-consuming and impractical expert annotations. 107,471 histopathology images obtained from 172 biopsy cores were used in the clustering and in the deep learning model. To incorporate spatial information over the clustered image patterns on the biopsy sample, we spatially encoded clustered patterns with colors and performed spatial analysis through graph neural network. A random forest classifier with various groups of features were used to predict CKD. For predicting eGFR at the biopsy, we achieved a sensitivity of 0.97, specificity of 0.90, and accuracy of 0.95. AUC was 0.96. For predicting eGFR changes in one-year, we achieved a sensitivity of 0.83, specificity of 0.85, and accuracy of 0.84. AUC was 0.85. This study presents the first spatial analysis based on unsupervised machine learning algorithms. Without expert annotation, CluSA framework can not only accurately classify and predict the degree of kidney function at the biopsy and in one year, but also identify novel predictors of kidney function and renal prognosis. [ABSTRACT FROM AUTHOR]

Pathologists use visual classification to assess patient kidney biopsy samples when diagnosing the underlying cause of kidney disease. However, the assessment is qualitative, or semi-quantitative at best, and reproducibility is challenging. To discover previously unknown features which predict patient outcomes and overcome substantial interobserver variability, we developed an unsupervised bag-of-words model. Our study applied to the C-PROBE cohort of patients with chronic kidney disease (CKD). 107,471 histopathology images were obtained from 161 biopsy cores and identified important morphological features in biopsy tissue that are highly predictive of the presence of CKD both at the time of biopsy and in one year. To evaluate the performance of our model, we estimated the AUC and its 95% confidence interval. We show that this method is reliable and reproducible and can achieve 0.93 AUC at predicting glomerular filtration rate at the time of biopsy as well as predicting a loss of function at one year. Additionally, with this method, we ranked the identified morphological features according to their importance as diagnostic markers for chronic kidney disease. In this study, we have demonstrated the feasibility of using an unsupervised machine learning method without human input in order to predict the level of kidney function in CKD. The results from our study indicate that the visual dictionary, or visual image pattern, obtained from unsupervised machine learning can predict outcomes using machine-derived values that correspond to both known and unknown clinically relevant features. [ABSTRACT FROM AUTHOR]

Radiomics involves high-throughput extraction of large numbers of quantitative features from medical images and analysis of these features to predict patients' outcome and support clinical decision-making. However, radiomics features are sensitive to several factors, including scanning protocols. The purpose of this study was to investigate the robustness of magnetic resonance imaging (MRI) radiomics features with various MRI scanning protocol parameters and scanners using an MRI radiomics phantom. The variability of the radiomics features with different scanning parameters and repeatability measured using a test–retest scheme were evaluated using the coefficient of variation and intraclass correlation coefficient (ICC) for both T1- and T2-weighted images. For variability measures, the features were categorized into three groups: large, intermediate, and small variation. For repeatability measures, the average T1- and T2-weighted image ICCs for the phantom (0.963 and 0.959, respectively) were higher than those for a healthy volunteer (0.856 and 0.849, respectively). Our results demonstrated that various radiomics features are dependent on different scanning parameters and scanners. The radiomics features with a low coefficient of variation and high ICC for both the phantom and volunteer can be considered good candidates for MRI radiomics studies. The results of this study will assist current and future MRI radiomics studies. [ABSTRACT FROM AUTHOR]

One of the most common and aggressive malignant brain tumors is Glioblastoma multiforme. Despite the multimodality treatment such as radiation therapy and chemotherapy (temozolomide: TMZ), the median survival rate of glioblastoma patient is less than 15 months. In this study, we investigated the association between measures of spatial diversity derived from spatial point pattern analysis of multiparametric magnetic resonance imaging (MRI) data with molecular status as well as 12-month survival in glioblastoma. We obtained 27 measures of spatial proximity (diversity) via spatial point pattern analysis of multiparametric T1 post-contrast and T2 fluid-attenuated inversion recovery MRI data. These measures were used to predict 12-month survival status (≤12 or >12 months) in 74 glioblastoma patients. Kaplan-Meier with receiver operating characteristic analyses was used to assess the relationship between derived spatial features and 12-month survival status as well as molecular subtype status in patients with glioblastoma. Kaplan-Meier survival analysis revealed that 14 spatial features were capable of stratifying overall survival in a statistically significant manner. For prediction of 12-month survival status based on these diversity indices, sensitivity and specificity were 0.86 and 0.64, respectively. The area under the receiver operating characteristic curve and the accuracy were 0.76 and 0.75, respectively. For prediction of molecular subtype status, proneural subtype shows highest accuracy of 0.93 among all molecular subtypes based on receiver operating characteristic analysis. We find that measures of spatial diversity from point pattern analysis of intensity habitats from T1 post-contrast and T2 fluid-attenuated inversion recovery images are associated with both tumor subtype status and 12-month survival status and may therefore be useful indicators of patient prognosis, in addition to providing potential guidance for molecularly-targeted therapies in Glioblastoma multiforme. [ABSTRACT FROM AUTHOR]

This study compared three methods for analyzing DCE-MRI data with a reference region (RR) model: a linear least-square fitting with numerical analysis (LLSQ-N), a nonlinear least-square fitting with numerical analysis (NLSQ-N), and an analytical analysis (NLSQ-A). The accuracy and precision of estimating the pharmacokinetic parameter ratios KR and VR, where KR is defined as a ratio between the two volume transfer constants,K trans,TOI and K trans,RR, and VR is the ratio between the two extracellular extravascular volumes,ve,TOI and ve,RR, were assessed using simulations under various signal-to-noise ratios (SNRs) and temporal resolutions (4, 6, 30, and 60s). When no noise was added, the simulations showed that the mean percent error (MPE) for the estimated KR and VR using the LLSQ-N and NLSQ-N methods ranged from 1.2% to 31.6% with various temporal resolutions while the NLSQ-A method maintained a very high accuracy (<1.0×104%) regardless of the temporal resolution. The simulation also indicated that the LLSQ-N and NLSQ-N methods appear to underestimate the parameter ratios more than the NLSQ-A method. In addition, seven in vivo DCE-MRI datasets from spontaneously occurring canine brain tumors were analyzed with each method. Results for the in vivo study showed that KR(ranging from 0.63 to 3.11) and VR(ranging from 2.82 to 19.16) for the NLSQ-A method were both higher than results for the other two methods (KR ranging from 0.01 to 1.29 and VR ranging from 1.48 to 19.59). A temporal downsampling experiment showed that the averaged percent error for the NLSQ-A method (8.45%) was lower than the other two methods (22.97% for LLSQ-N and 65.02% for NLSQ-N) for KR, and the averaged percent error for the NLSQ-A method (6.33%) was lower than the other two methods (6.57% for LLSQ-N and 13.66% for NLSQ-N) for VR. Using simulations, we showed that the NLSQ-A method can estimate the ratios of pharmacokinetic parameters more accurately and precisely than the NLSQ-N and LLSQ-N methods over various SNRs and temporal resolutions. All simulations were validated with in vivo DCE MRI data. [ABSTRACT FROM AUTHOR]

Abstract: Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is performed by obtaining sequential MRI images, before, during and after the injection of a contrast agent. It is usually used to observe the exchange of contrast agent between the vascular space and extravascular extracellular space (EES), and provide information about blood volume and microvascular permeability. To estimate the kinetic parameters derived from the pharmacokinetic model, accurate knowledge of the arterial input function (AIF) is very important. However, the AIF is usually unknown, and it remains very difficult to obtain such information noninvasively. In this article, without knowledge of the AIF, we applied a reference region (RR) model to analyze the kinetic parameters. The RR model usually depends on kinetic parameters found in previous studies of a reference region. However, both the assignment of reference region parameters (intersubject variation) and the selection of the reference region itself (intrasubject variation) may confound the results obtained by RR methods. Instead of using literature values for those pharmacokinetic parameters of the reference region, we proposed to use two pharmacokinetic parameter ratios between the tissue of interest (TOI) and the reference region. Specifically, one parameter K R is calculated as the ratio between the volume transfer constant K trans of the TOI and RR. Similarly, another parameter V R is calculated as the ratio between the extravascular extracellular volume fraction v e of the TOI and RR. To investigate the consistency of the two ratios, the K trans of the RR was varied ranging from 0.1 to 1.0 min−1, covering the cited literature values. A simulated dataset with different levels of Gaussian noises and an in vivo dataset acquired from five canine brains with spontaneous occurring brain tumors were used to study the proposed ratios. It is shown from both datasets that these ratios are independent of K trans of the RR, implying that there is potentially no need to obtain information about literature values from the reference region for future pharmacokinetic modeling and analysis. [Copyright &y& Elsevier]

A large eddy simulation has been performed to investigate the effect of swirl on the heat and momentum transfer in an annular pipe flow with a rotating inner wall. The compressible filtered Navier-Stokes equations were solved using a second-order-accurate finite-volume method. Low-Mach-number preconditioning was used to enable the compressible code to work efficiently at low Mach numbers. A dynamic subgrid-scale stress model accounted for the subgrid-scale turbulence. A nonuniform grid in the radial direction yielded very accurate solutions using a reasonable number of grid points. The numerical results are summarized and compared with the experimental results of previous studies. The simulations indicated that the Nusselt number and the wall friction coefficient increased with increasing rotation speed of the wall. It was also observed that the axial velocity profile became flattened and turbulent intensities were enhanced due to swirl. This modification of the turbulent structures was closely related to the increase of the Nusselt number and the friction coefficient. [ABSTRACT FROM AUTHOR]

Differentiating pseudoprogression from true tumor progression has become a significant challenge in follow-up of diffuse infiltrating gliomas, particularly high grade, which leads to a potential treatment delay for patients with early glioma recurrence. In this study, we proposed to use a multiparametric MRI data as a sequence input for the convolutional neural network with the recurrent neural network based deep learning structure to discriminate between pseudoprogression and true tumor progression. In this study, 43 biopsy-proven patient data identified as diffuse infiltrating glioma patients whose disease progressed/recurred were used. The dataset consists of five original MRI sequences; pre-contrast T1-weighted, post-contrast T1-weighted, T2-weighted, FLAIR, and ADC images as well as two engineered sequences; T1post–T1pre and T2–FLAIR. Next, we used three CNN-LSTM models with a different set of sequences as input sequences to pass through CNN-LSTM layers. We performed threefold cross-validation in the training dataset and generated the boxplot, accuracy, and ROC curve, AUC from each trained model with the test dataset to evaluate models. The mean accuracy for VGG16 models ranged from 0.44 to 0.60 and the mean AUC ranged from 0.47 to 0.59. For CNN-LSTM model, the mean accuracy ranged from 0.62 to 0.75 and the mean AUC ranged from 0.64 to 0.81. The performance of the proposed CNN-LSTM with multiparametric sequence data was found to outperform the popular convolutional CNN with a single MRI sequence. In conclusion, incorporating all available MRI sequences into a sequence input for a CNN-LSTM model improved diagnostic performance for discriminating between pseudoprogression and true tumor progression. [ABSTRACT FROM AUTHOR]

Background: Machine learning models provide significant opportunities for improvement in health care, but their "black-box" nature poses many risks. Methods: We built a custom Python module as part of a framework for generating artifacts that are meant to be tunable and describable to allow for future testing needs. We conducted an analysis of a previously published digital pathology classification model and an internally developed kidney tissue segmentation model, utilizing a variety of generated artifacts including testing their effects. The artifacts simulated were bubbles, tissue folds, uneven illumination, marker lines, uneven sectioning, altered staining, and tissue tears. Results: We found that there is some performance degradation on the tiles with artifacts, particularly with altered stains but also with marker lines, tissue folds, and uneven sectioning. We also found that the response of deep learning models to artifacts could be nonlinear. Conclusions: Generated artifacts can provide a useful tool for testing and building trust in machine learning models by understanding where these models might fail. [ABSTRACT FROM AUTHOR]

First-order radiomic features, such as metabolic tumor volume (MTV) and total lesion glycolysis (TLG), are associated with disease progression in early-stage classical Hodgkin lymphoma (HL). We hypothesized that a model incorporating first- and second-order radiomic features would more accurately predict outcome than MTV or TLG alone. We assessed whether radiomic features extracted from baseline PET scans predicted relapsed or refractory disease status in a cohort of 251 patients with stage I-II HL who were managed at a tertiary cancer center. Models were developed and tested using a machine-learning algorithm. Features extracted from mediastinal sites were highly predictive of primary refractory disease. A model incorporating 5 of the most predictive features had an area under the curve (AUC) of 95.2% and total error rate of 1.8%. By comparison, the AUC was 78% for both MTV and TLG and was 65% for maximum standardize uptake value (SUVmax). Furthermore, among the patients with refractory mediastinal disease, our model distinguished those who were successfully salvaged from those who ultimately died of HL. We conclude that our PET radiomic model may improve upfront stratification of early-stage HL patients with mediastinal disease and thus contribute to risk-adapted, individualized management. [ABSTRACT FROM AUTHOR]

Recently, the development of durable hydrophobic surfaces has received much attention, with anti-icing applications in harsh environments such as aircrafts, wind turbines, power lines, and marine vessels. In this paper we describe a design methodology employing a lattice Boltzmann method to determine the optimal geometry of microstructures to achieve superhydrophobicity. We describe a top-down fabrication method to form superhydrophobic micro-hierarchical metal surface using photolithography, nanoimprinting, and continuous metal-to-metal replication using pulse-reverse-current electrochemical deposition. The surfaces were formed of nickel, which has a large hardness and is resistant to corrosion, making it suitable for use in harsh external conditions. We compared the measured wettability of fabricated micro-hierarchical metal surface with that from numerical simulations. The contact angle and contact angle hysteresis of four metal surfaces were measured (i.e., a bare surface, a random nanostructured surface, an engineered nanostructured surface, and an engineered hierarchical structured surface), and the anti-icing properties of these four metal surfaces were investigated. [ABSTRACT FROM AUTHOR]