Noninvasive Molecular Subtyping of Pediatric Low-Grade Glioma with Self-Supervised Transfer Learning.

Purpose To develop and externally test a scan-to-prediction deep learning pipeline for noninvasive, MRI-based BRAF mutational status classification for pediatric low-grade glioma. Materials and Methods This retrospective study included two pediatric low-grade glioma datasets with linked genomic and diagnostic T2-weighted MRI data of patients: Dana-Farber/Boston Children's Hospital (development dataset, n = 214 [113 (52.8%) male; 104 (48.6%) BRAF wild type, 60 (28.0%) BRAF fusion, and 50 (23.4%) BRAF V600E]) and the Children's Brain Tumor Network (external testing, n = 112 [55 (49.1%) male; 35 (31.2%) BRAF wild type, 60 (53.6%) BRAF fusion, and 17 (15.2%) BRAF V600E]). A deep learning pipeline was developed to classify BRAF mutational status (BRAF wild type vs BRAF fusion vs BRAF V600E) via a two-stage process: (a) three-dimensional tumor segmentation and extraction of axial tumor images and (b) section-wise, deep learning-based classification of mutational status. Knowledge-transfer and self-supervised approaches were investigated to prevent model overfitting, with a primary end point of the area under the receiver operating characteristic curve (AUC). To enhance model interpretability, a novel metric, center of mass distance, was developed to quantify the model attention around the tumor. Results A combination of transfer learning from a pretrained medical imaging-specific network and self-supervised label cross-training (TransferX) coupled with consensus logic yielded the highest classification performance with an AUC of 0.82 (95% CI: 0.72, 0.91), 0.87 (95% CI: 0.61, 0.97), and 0.85 (95% CI: 0.66, 0.95) for BRAF wild type, BRAF fusion, and BRAF V600E, respectively, on internal testing. On external testing, the pipeline yielded an AUC of 0.72 (95% CI: 0.64, 0.86), 0.78 (95% CI: 0.61, 0.89), and 0.72 (95% CI: 0.64, 0.88) for BRAF wild type, BRAF fusion, and BRAF V600E, respectively. Conclusion Transfer learning and self-supervised cross-training improved classification performance and generalizability for noninvasive pediatric low-grade glioma mutational status prediction in a limited data scenario. Keywords: Pediatrics, MRI, CNS, Brain/Brain Stem, Oncology, Feature Detection, Diagnosis, Supervised Learning, Transfer Learning, Convolutional Neural Network (CNN) Supplemental material is available for this article. © RSNA, 2024.

Automated temporalis muscle quantification and growth charts for children through adulthood.

Lean muscle mass (LMM) is an important aspect of human health. Temporalis muscle thickness is a promising LMM marker but has had limited utility due to its unknown normal growth trajectory and reference ranges and lack of standardized measurement. Here, we develop an automated deep learning pipeline to accurately measure temporalis muscle thickness (iTMT) from routine brain magnetic resonance imaging (MRI). We apply iTMT to 23,876 MRIs of healthy subjects, ages 4 through 35, and generate sex-specific iTMT normal growth charts with percentiles. We find that iTMT was associated with specific physiologic traits, including caloric intake, physical activity, sex hormone levels, and presence of malignancy. We validate iTMT across multiple demographic groups and in children with brain tumors and demonstrate feasibility for individualized longitudinal monitoring. The iTMT pipeline provides unprecedented insights into temporalis muscle growth during human development and enables the use of LMM tracking to inform clinical decision-making.

Noninvasive molecular subtyping of pediatric low-grade glioma with self-supervised transfer learning.

Purpose: To develop and externally validate a scan-to-prediction deep-learning pipeline for noninvasive, MRI-based BRAF mutational status classification for pLGG. Materials and Methods: We conducted a retrospective study of two pLGG datasets with linked genomic and diagnostic T2-weighted MRI of patients: BCH (development dataset, n=214 [60 (28%) BRAF fusion, 50 (23%) BRAF V600E, 104 (49%) wild-type), and Child Brain Tumor Network (CBTN) (external validation, n=112 [60 (53%) BRAF-Fusion, 17 (15%) BRAF-V600E, 35 (32%) wild-type]). We developed a deep learning pipeline to classify BRAF mutational status (V600E vs. fusion vs. wildtype) via a two-stage process: 1) 3D tumor segmentation and extraction of axial tumor images, and 2) slice-wise, deep learning-based classification of mutational status. We investigated knowledge-transfer and self-supervised approaches to prevent model overfitting with a primary endpoint of the area under the receiver operating characteristic curve (AUC). To enhance model interpretability, we developed a novel metric, COMDist, that quantifies the accuracy of model attention around the tumor. Results: A combination of transfer learning from a pretrained medical imaging-specific network and self-supervised label cross-training (TransferX) coupled with consensus logic yielded the highest macro-average AUC (0.82 [95% CI: 0.70-0.90]) and accuracy (77%) on internal validation, with an AUC improvement of +17.7% and a COMDist improvement of +6.4% versus training from scratch. On external validation, the TransferX model yielded AUC (0.73 [95% CI 0.68-0.88]) and accuracy (75%). Conclusion: Transfer learning and self-supervised cross-training improved classification performance and generalizability for noninvasive pLGG mutational status prediction in a limited data scenario.

Expert-level pediatric brain tumor segmentation in a limited data scenario with stepwise transfer learning.

Purpose: Artificial intelligence (AI)-automated tumor delineation for pediatric gliomas would enable real-time volumetric evaluation to support diagnosis, treatment response assessment, and clinical decision-making. Auto-segmentation algorithms for pediatric tumors are rare, due to limited data availability, and algorithms have yet to demonstrate clinical translation. Methods: We leveraged two datasets from a national brain tumor consortium (n=184) and a pediatric cancer center (n=100) to develop, externally validate, and clinically benchmark deep learning neural networks for pediatric low-grade glioma (pLGG) segmentation using a novel in-domain, stepwise transfer learning approach. The best model [via Dice similarity coefficient (DSC)] was externally validated and subject to randomized, blinded evaluation by three expert clinicians wherein clinicians assessed clinical acceptability of expert- and AI-generated segmentations via 10-point Likert scales and Turing tests. Results: The best AI model utilized in-domain, stepwise transfer learning (median DSC: 0.877 [IQR 0.715-0.914]) versus baseline model (median DSC 0.812 [IQR 0.559-0.888]; p<0.05). On external testing (n=60), the AI model yielded accuracy comparable to inter-expert agreement (median DSC: 0.834 [IQR 0.726-0.901] vs. 0.861 [IQR 0.795-0.905], p=0.13). On clinical benchmarking (n=100 scans, 300 segmentations from 3 experts), the experts rated the AI model higher on average compared to other experts (median Likert rating: 9 [IQR 7-9]) vs. 7 [IQR 7-9], p<0.05 for each). Additionally, the AI segmentations had significantly higher (p<0.05) overall acceptability compared to experts on average (80.2% vs. 65.4%). Experts correctly predicted the origins of AI segmentations in an average of 26.0% of cases. Conclusions: Stepwise transfer learning enabled expert-level, automated pediatric brain tumor auto-segmentation and volumetric measurement with a high level of clinical acceptability. This approach may enable development and translation of AI imaging segmentation algorithms in limited data scenarios.

Deep learning-enabled quantification of simultaneous PET/MRI for cell transplantation monitoring.

Current methods of in vivo imaging islet cell transplants for diabetes using magnetic resonance imaging (MRI) are limited by their low sensitivity. Simultaneous positron emission tomography (PET)/MRI has greater sensitivity and ability to visualize cell metabolism. However, this dual-modality tool currently faces two major challenges for monitoring cells. Primarily, the dynamic conditions of PET such as signal decay and spatiotemporal change in radioactivity prevent accurate quantification of the transplanted cell number. In addition, selection bias from different radiologists renders human error in segmentation. This calls for the development of artificial intelligence algorithms for the automated analysis of PET/MRI of cell transplantations. Here, we combined K-means++ for segmentation with a convolutional neural network to predict radioactivity in cell-transplanted mouse models. This study provides a tool combining machine learning with a deep learning algorithm for monitoring islet cell transplantation through PET/MRI. It also unlocks a dynamic approach to automated segmentation and quantification of radioactivity in PET/MRI.

Magnetic Particle Imaging of Transplanted Human Islets Using a Machine Learning Algorithm.

Human islet transplantation is a promising therapy to restore normoglycemia for type 1 diabetes (T1D). Despite recent advances, human islet transplantation remains suboptimal due to significant islet graft loss after transplantation. Various immunological and nonimmunological factors contribute to this loss therefore signifying a need for strategies and approaches for visualizing and monitoring transplanted human islet grafts. One such imaging approach is magnetic particle imaging (MPI), an emerging imaging modality that detects the magnetization of iron oxide nanoparticles. MPI is known for its specificity due to its high image contrast and sensitivity. MPI through its noninvasive nature provides the means for monitoring transplanted human islets in real time. Here we summarize an approach to track transplanted human islets using MPI. We label human islet from donors with dextran-coated ferucarbotran iron oxide nanoparticles, transplant the labeled human islet into under the left kidney capsule, and image graft cells using an MPI scanner. We engineer a K-means++, clustering-based unsupervised machine learning algorithm for standardized image segmentation and iron quantification of the MPI, which solves problems with selection bias and indiscriminate signal boundary that accompanies this newer imaging modality. In this chapter, we summarize the methods of this emerging imaging modality of MPI in conjunction with unsupervised machine learning to monitor and visualize islets after transplantation.

Nanotechnology in Immunotherapy for Type 1 Diabetes: Promising Innovations and Future Advances.

Diabetes is a chronic condition which affects the glucose metabolism in the body. In lieu of any clinical "cure," the condition is managed through the administration of pharmacological aids, insulin supplements, diet restrictions, exercise, and the like. The conventional clinical prescriptions are limited by their life-long dependency and diminished potency, which in turn hinder the patient's recovery. This necessitated an alteration in approach and has instigated several investigations into other strategies. As Type 1 diabetes (T1D) is known to be an autoimmune disorder, targeting the immune system in activation and/or suppression has shown promise in reducing beta cell loss and improving insulin levels in response to hyperglycemia. Another strategy currently being explored is the use of nanoparticles in the delivery of immunomodulators, insulin, or engineered vaccines to endogenous immune cells. Nanoparticle-assisted targeting of immune cells holds substantial potential for enhanced patient care within T1D clinical settings. Herein, we summarize the knowledge of etiology, clinical scenarios, and the current state of nanoparticle-based immunotherapeutic approaches for Type 1 diabetes. We also discuss the feasibility of translating this approach to clinical practice.

3D in vivo Magnetic Particle Imaging of Human Stem Cell-Derived Islet Organoid Transplantation Using a Machine Learning Algorithm.

Stem cell-derived islet organoids constitute a promising treatment of type 1 diabetes. A major hurdle in the field is the lack of appropriate in vivo method to determine graft outcome. Here, we investigate the feasibility of in vivo tracking of transplanted stem cell-derived islet organoids using magnetic particle imaging (MPI) in a mouse model. Human induced pluripotent stem cells-L1 were differentiated to islet organoids and labeled with superparamagnetic iron oxide nanoparticles. The phantoms comprising of different numbers of labeled islet organoids were imaged using an MPI system. Labeled islet organoids were transplanted into NOD/scid mice under the left kidney capsule and were then scanned using 3D MPI at 1, 7, and 28 days post transplantation. Quantitative assessment of the islet organoids was performed using the K-means++ algorithm analysis of 3D MPI. The left kidney was collected and processed for immunofluorescence staining of C-peptide and dextran. Islet organoids expressed islet cell markers including insulin and glucagon. Image analysis of labeled islet organoids phantoms revealed a direct linear correlation between the iron content and the number of islet organoids. The K-means++ algorithm showed that during the course of the study the signal from labeled islet organoids under the left kidney capsule decreased. Immunofluorescence staining of the kidney sections showed the presence of islet organoid grafts as confirmed by double staining for dextran and C-peptide. This study demonstrates that MPI with machine learning algorithm analysis can monitor islet organoids grafts labeled with super-paramagnetic iron oxide nanoparticles and provide quantitative information of their presence in vivo.

A concise review: the synergy between artificial intelligence and biomedical nanomaterials that empowers nanomedicine.

Nanomedicine has recently experienced unprecedented growth and development. However, the complexity of operations at the nanoscale introduces a layer of difficulty in the clinical translation of nanodrugs and biomedical nanotechnology. This problem is further exacerbated when engineering and optimizing nanomaterials for biomedical purposes. To navigate this issue, artificial intelligence (AI) algorithms have been applied for data analysis and inference, allowing for a more applicable understanding of the complex interaction amongst the abundant variables in a system involving the synthesis or use of nanomedicine. Here, we report on the current relationship and implications of nanomedicine and AI. Particularly, we explore AI as a tool for enabling nanomedicine in the context of nanodrug screening and development, brain-machine interfaces and nanotoxicology. We also report on the current state and future direction of nanomedicine and AI in cancer, diabetes, and neurological disorder therapy.

A Concise Review: The Role of Stem Cells in Cancer Progression and Therapy.

The properties of cancer stem cells (CSCs) have recently gained attention as an avenue of intervention for cancer therapy. In this review, we highlight some of the key roles of CSCs in altering the cellular microenvironment in favor of cancer progression. We also report on various studies in this field which focus on transformative properties of CSCs and their influence on surrounding cells or targets through the release of cellular cargo in the form of extracellular vesicles. The findings from these studies encourage the development of novel interventional therapies that can target and prevent cancer through efficient, more effective methods. These methods include targeting immunosuppressive proteins and biomarkers, promoting immunization against tumors, exosome-mediated CSC conversion, and a focus on the quiescent properties of CSCs and their role in cancer progression. The resulting therapeutic benefit and transformative potential of these novel approaches to stem cell-based cancer therapy provide a new direction in cancer treatment, which can focus on nanoscale, molecular properties of the cellular microenvironment and establish a more precision medicine-oriented paradigm of treatment.

Artificial Intelligence Analysis of Magnetic Particle Imaging for Islet Transplantation in a Mouse Model.

PURPOSE: Current approaches to quantification of magnetic particle imaging (MPI) for cell-based therapy are thwarted by the lack of reliable, standardized methods of segmenting the signal from background in images. This calls for the development of artificial intelligence (AI) systems for MPI analysis. PROCEDURES: We utilize a canonical algorithm in the domain of unsupervised machine learning, known as K-means++, to segment the regions of interest (ROI) of images and perform iron quantification analysis using a standard curve model. We generated in vitro, in vivo, and ex vivo data using islets and mouse models and applied the AI algorithm to gain insight into segmentation and iron prediction on these MPI data. In vitro models included imaging the VivoTrax-labeled islets in varying numbers. In vivo mouse models were generated through transplantation of increasing numbers of the labeled islets under the kidney capsule of mice. Ex vivo data were obtained from the MPI images of excised kidney grafts. RESULTS: The K-means++ algorithms segmented the ROI of in vitro phantoms with minimal noise. A linear correlation between the islet numbers and the increasing prediction of total iron value (TIV) in the islets was observed. Segmentation results of the ROI of the in vivo MPI scans showed that with increasing number of transplanted islets, the signal intensity increased with linear trend. Upon segmenting the ROI of ex vivo data, a linear trend was observed in which increasing intensity of the ROI yielded increasing TIV of the islets. Through statistical evaluation of the algorithm performance via intraclass correlation coefficient validation, we observed excellent performance of K-means++-based model on segmentation and quantification analysis of MPI data. CONCLUSIONS: We have demonstrated the ability of the K-means++-based model to provide a standardized method of segmentation and quantification of MPI scans in an islet transplantation mouse model.

Current Progress and Perspective: Clinical Imaging of Islet Transplantation.

Islet transplantation has great potential as a cure for type 1 diabetes. At present; the lack of a clinically validated non-invasive imaging method to track islet grafts limits the success of this treatment. Some major clinical imaging modalities and various molecular probes, which have been studied for non-invasive monitoring of transplanted islets, could potentially fulfill the goal of understanding pathophysiology of the functional status and viability of the islet grafts. In this current review, we summarize the recent clinical studies of a variety of imaging modalities and molecular probes for non-invasive imaging of transplanted beta cell mass. This review also includes discussions on in vivo detection of endogenous beta cell mass using clinical imaging modalities and various molecular probes, which will be useful for longitudinally detecting the status of islet transplantation in Type 1 diabetic patients. For the conclusion and perspectives, we highlight the applications of multimodality and novel imaging methods in islet transplantation.

Molecular imaging and deep learning analysis of uMUC1 expression in response to chemotherapy in an orthotopic model of ovarian cancer.

Artificial Intelligence (AI) algorithms including deep learning have recently demonstrated remarkable progress in image-recognition tasks. Here, we utilized AI for monitoring the expression of underglycosylated mucin 1 (uMUC1) tumor antigen, a biomarker for ovarian cancer progression and response to therapy, using contrast-enhanced in vivo imaging. This was done using a dual-modal (magnetic resonance and near infrared optical imaging) uMUC1-specific probe (termed MN-EPPT) consisted of iron-oxide magnetic nanoparticles (MN) conjugated to a uMUC1-specific peptide (EPPT) and labeled with a near-infrared fluorescent dye, Cy5.5. In vitro studies performed in uMUC1-expressing human ovarian cancer cell line SKOV3/Luc and control uMUC1low ES-2 cells showed preferential uptake on the probe by the high expressor (n = 3, p < .05). A decrease in MN-EPPT uptake by SKOV3/Luc cells in vitro due to uMUC1 downregulation after docetaxel therapy was paralleled by in vivo imaging studies that showed a reduction in probe accumulation in the docetaxel treated group (n = 5, p < .05). The imaging data were analyzed using deep learning-enabled segmentation and quantification of the tumor region of interest (ROI) from raw input MRI sequences by applying AI algorithms including a blend of Convolutional Neural Networks (CNN) and Fully Connected Neural Networks. We believe that the algorithms used in this study have the potential to improve studying and monitoring cancer progression, amongst other diseases.