Detection of glaucoma progression on longitudinal series of en-face macular optical coherence tomography angiography images with a deep learning model.

BACKGROUND/AIMS: To design a deep learning (DL) model for the detection of glaucoma progression with a longitudinal series of macular optical coherence tomography angiography (OCTA) images. METHODS: 202 eyes of 134 patients with open-angle glaucoma with ≥4 OCTA visits were followed for an average of 3.5 years. Glaucoma progression was defined as having a statistically significant negative 24-2 visual field (VF) mean deviation (MD) rate. The baseline and final macular OCTA images were aligned according to centre of fovea avascular zone automatically, by checking the highest value of correlation between the two images. A customised convolutional neural network (CNN) was designed for classification. A comparison of the CNN to logistic regression model for whole image vessel density (wiVD) loss on detection of glaucoma progression was performed. The performance of the model was defined based on the confusion matrix of the validation dataset and the area under receiver operating characteristics (AUC). RESULTS: The average (95% CI) baseline VF MD was -3.4 (-4.1 to -2.7) dB. 28 (14%) eyes demonstrated glaucoma progression. The AUC (95% CI) of the DL model for the detection of glaucoma progression was 0.81 (0.59 to 0.93). The sensitivity, specificity and accuracy (95% CI) of DL model were 67% (34% to 78%), 83% (42% to 97%) and 80% (52% to 95%), respectively. The AUC (95% CI) for the detection of glaucoma progression based on the logistic regression model was lower than the DL model (0.69 (0.50 to 0.88)). CONCLUSION: The optimised DL model detected glaucoma progression based on longitudinal macular OCTA images showed good performance. With external validation, it could enhance detection of glaucoma progression. TRIAL REGISTRATION NUMBER: NCT00221897.

Diagnosing Solid Lesions in the Pancreas With Multimodal Artificial Intelligence: A Randomized Crossover Trial.

Importance: Diagnosing solid lesions in the pancreas via endoscopic ultrasonographic (EUS) images is challenging. Artificial intelligence (AI) has the potential to help with such diagnosis, but existing AI models focus solely on a single modality. Objective: To advance the clinical diagnosis of solid lesions in the pancreas through developing a multimodal AI model integrating both clinical information and EUS images. Design, Setting, and Participants: In this randomized crossover trial conducted from January 1 to June 30, 2023, from 4 centers across China, 12 endoscopists of varying levels of expertise were randomly assigned to diagnose solid lesions in the pancreas with or without AI assistance. Endoscopic ultrasonographic images and clinical information of 439 patients from 1 institution who had solid lesions in the pancreas between January 1, 2014, and December 31, 2022, were collected to train and validate the joint-AI model, while 189 patients from 3 external institutions were used to evaluate the robustness and generalizability of the model. Intervention: Conventional or AI-assisted diagnosis of solid lesions in the pancreas. Main Outcomes and Measures: In the retrospective dataset, the performance of the joint-AI model was evaluated internally and externally. In the prospective dataset, diagnostic performance of the endoscopists with or without the AI assistance was compared. Results: The retrospective dataset included 628 patients (400 men [63.7%]; mean [SD] age, 57.7 [27.4] years) who underwent EUS procedures. A total of 130 patients (81 men [62.3%]; mean [SD] age, 58.4 [11.7] years) were prospectively recruited for the crossover trial. The area under the curve of the joint-AI model ranged from 0.996 (95% CI, 0.993-0.998) in the internal test dataset to 0.955 (95% CI, 0.940-0.968), 0.924 (95% CI, 0.888-0.955), and 0.976 (95% CI, 0.942-0.995) in the 3 external test datasets, respectively. The diagnostic accuracy of novice endoscopists was significantly enhanced with AI assistance (0.69 [95% CI, 0.61-0.76] vs 0.90 [95% CI, 0.83-0.94]; P < .001), and the supplementary interpretability information alleviated the skepticism of the experienced endoscopists. Conclusions and Relevance: In this randomized crossover trial of diagnosing solid lesions in the pancreas with or without AI assistance, the joint-AI model demonstrated positive human-AI interaction, which suggested its potential to facilitate a clinical diagnosis. Nevertheless, future randomized clinical trials are warranted. Trial Registration: ClinicalTrials.gov Identifier: NCT05476978.

Improving image classification of gastrointestinal endoscopy using curriculum self-supervised learning.

Endoscopy, a widely used medical procedure for examining the gastrointestinal (GI) tract to detect potential disorders, poses challenges in manual diagnosis due to non-specific symptoms and difficulties in accessing affected areas. While supervised machine learning models have proven effective in assisting clinical diagnosis of GI disorders, the scarcity of image-label pairs created by medical experts limits their availability. To address these limitations, we propose a curriculum self-supervised learning framework inspired by human curriculum learning. Our approach leverages the HyperKvasir dataset, which comprises 100k unlabeled GI images for pre-training and 10k labeled GI images for fine-tuning. By adopting our proposed method, we achieved an impressive top-1 accuracy of 88.92% and an F1 score of 73.39%. This represents a 2.1% increase over vanilla SimSiam for the top-1 accuracy and a 1.9% increase for the F1 score. The combination of self-supervised learning and a curriculum-based approach demonstrates the efficacy of our framework in advancing the diagnosis of GI disorders. Our study highlights the potential of curriculum self-supervised learning in utilizing unlabeled GI tract images to improve the diagnosis of GI disorders, paving the way for more accurate and efficient diagnosis in GI endoscopy.

Improving long COVID-related text classification: a novel end-to-end domain-adaptive paraphrasing framework.

The emergence of long COVID during the ongoing COVID-19 pandemic has presented considerable challenges for healthcare professionals and researchers. The task of identifying relevant literature is particularly daunting due to the rapidly evolving scientific landscape, inconsistent definitions, and a lack of standardized nomenclature. This paper proposes a novel solution to this challenge by employing machine learning techniques to classify long COVID literature. However, the scarcity of annotated data for machine learning poses a significant obstacle. To overcome this, we introduce a strategy called medical paraphrasing, which diversifies the training data while maintaining the original content. Additionally, we propose a Data-Reweighting-Based Multi-Level Optimization Framework for Domain Adaptive Paraphrasing, supported by a Meta-Weight-Network (MWN). This innovative approach incorporates feedback from the downstream text classification model to influence the training of the paraphrasing model. During the training process, the framework assigns higher weights to the training examples that contribute more effectively to the downstream task of long COVID text classification. Our findings demonstrate that this method substantially improves the accuracy and efficiency of long COVID literature classification, offering a valuable tool for physicians and researchers navigating this complex and ever-evolving field.

Interpretable unsupervised learning enables accurate clustering with high-throughput imaging flow cytometry.

A primary challenge of high-throughput imaging flow cytometry (IFC) is to analyze the vast amount of imaging data, especially in applications where ground truth labels are unavailable or hard to obtain. We present an unsupervised deep embedding algorithm, the Deep Convolutional Autoencoder-based Clustering (DCAEC) model, to cluster label-free IFC images without any prior knowledge of input labels. The DCAEC model first encodes the input images into the latent representations and then clusters based on the latent representations. Using the DCAEC model, we achieve a balanced accuracy of 91.9% for human white blood cell (WBC) clustering and 97.9% for WBC/leukemia clustering using the 3D IFC images and 3D DCAEC model. Above all, although no human recognizable features can separate the clusters of cells with protein localization, we demonstrate the fused DCAEC model can achieve a cluster balanced accuracy of 85.3% from the label-free 2D transmission and 3D side scattering images. To reveal how the neural network recognizes features beyond human ability, we use the gradient-weighted class activation mapping method to discover the cluster-specific visual patterns automatically. Evaluation results show that the automatically identified salient image regions have strong cluster-specific visual patterns for different clusters, which we believe is a stride for the interpretable neural network for cell analysis with high-throughput IFCs.

Improve the performance of CT-based pneumonia classification via source data reweighting.

Pneumonia is a life-threatening disease. Computer tomography (CT) imaging is broadly used for diagnosing pneumonia. To assist radiologists in accurately and efficiently detecting pneumonia from CT scans, many deep learning methods have been developed. These methods require large amounts of annotated CT scans, which are difficult to obtain due to privacy concerns and high annotation costs. To address this problem, we develop a three-level optimization based method which leverages CT data from a source domain to mitigate the lack of labeled CT scans in a target domain. Our method automatically identifies and downweights low-quality source CT data examples which are noisy or have large domain discrepancy with target data, by minimizing the validation loss of a target model trained on reweighted source data. On a target dataset with 2218 CT scans and a source dataset with 349 CT images, our method achieves an F1 score of 91.8% in detecting pneumonia and an F1 score of 92.4% in detecting other types of pneumonia, which are significantly better than those achieved by state-of-the-art baseline methods.

Brain tumor classification based on neural architecture search.

Brain tumor is a life-threatening disease and causes about 0.25 million deaths worldwide in 2020. Magnetic Resonance Imaging (MRI) is frequently used for diagnosing brain tumors. In medically underdeveloped regions, physicians who can accurately diagnose and assess the severity of brain tumors from MRI are highly lacking. Deep learning methods have been developed to assist physicians in detecting brain tumors from MRI and determining their subtypes. In existing methods, neural architectures are manually designed by human experts, which is time-consuming and labor-intensive. To address this problem, we propose to automatically search for high-performance neural architectures for classifying brain tumors from MRIs, by leveraging a Learning-by-Self-Explanation (LeaSE) architecture search method. LeaSE consists of an explainer model and an audience model. The explainer aims at searching for a highly performant architecture by encouraging the architecture to generate high-fidelity explanations of prediction outcomes, where explanations' fidelity is evaluated by the audience model. LeaSE is formulated as a four-level optimization problem involving a sequence of four learning stages which are conducted end-to-end. We apply LeaSE for MRI-based brain tumor classification, including four classes: glioma, meningioma, pituitary tumor, and healthy, on a dataset containing 3264 MRI images. Results show that our method can search for neural architectures that achieve better classification accuracy than manually designed deep neural networks while having fewer model parameters. For example, our method achieves a test accuracy of 90.6% and an AUC of 95.6% with 3.75M parameters while the accuracy and AUC of a human-designed network-ResNet101-is 84.5% and 90.1% respectively with 42.56M parameters. In addition, our method outperforms state-of-the-art neural architecture search methods.

Neural architecture search for pneumonia diagnosis from chest X-rays.

Pneumonia is one of the diseases that causes the most fatalities worldwide, especially in children. Recently, pneumonia-caused deaths have increased dramatically due to the novel Coronavirus global pandemic. Chest X-ray (CXR) images are one of the most readily available and common imaging modality for the detection and identification of pneumonia. However, the detection of pneumonia from chest radiography is a difficult task even for experienced radiologists. Artificial Intelligence (AI) based systems have great potential in assisting in quick and accurate diagnosis of pneumonia from chest X-rays. The aim of this study is to develop a Neural Architecture Search (NAS) method to find the best convolutional architecture capable of detecting pneumonia from chest X-rays. We propose a Learning by Teaching framework inspired by the teaching-driven learning methodology from humans, and conduct experiments on a pneumonia chest X-ray dataset with over 5000 images. Our proposed method yields an area under ROC curve (AUC) of 97.6% for pneumonia detection, which improves upon previous NAS methods by 5.1% (absolute).

Assessment of Motor Dysfunction with Virtual Reality in Patients Undergoing [123I]FP-CIT SPECT/CT Brain Imaging.

[123I]FP-CIT SPECT has been valuable for distinguishing Parkinson disease (PD) from essential tremor. However, its performance for quantitative assessment of motor dysfunction has not been established. A virtual reality (VR) application was developed and compared with [123I]FP-CIT SPECT/CT for detection of severity of motor dysfunction. Forty-four patients (21 males, 23 females, age 64.5 ± 12.4) with abnormal [123I]FP-CIT SPECT/CT underwent assessment of bradykinesia, activities of daily living, and tremor with VR. Support vector machines (SVM) machine learning models were applied to VR and SPECT data. Receiver operating characteristic (ROC) analysis demonstrated greater area under the curve (AUC) for VR (0.8418, 95% CI 0.6071-0.9617) compared with brain SPECT (0.5357, 95% CI 0.3373-0.7357, p = 0.029) for detection of motor dysfunction. Logistic regression identified VR as an independent predictor of motor dysfunction (Odds Ratio 326.4, SE 2.17, p = 0.008). SVM for prediction of the Unified Parkinson's Disease Rating Scale Part III (UPDRS-III) demonstrated greater R-squared of 0.713 (p = 0.008) for VR, compared with 0.0764 (p = 0.361) for brain SPECT. This study demonstrates that VR can be safely used in patients prior to [123I]FP-CIT SPECT imaging and may improve prediction of motor dysfunction. This test has the potential to provide a simple, objective, quantitative analysis of motor symptoms in PD patients.

Molecular dynamics simulation on the allosteric analysis of the c-di-GMP class I riboswitch induced by ligand binding.

Riboswitches are RNA molecules that regulate gene expression using conformation change, affected by binding of small molecule ligands. Although a number of ligand-bound aptamer complex structures have been solved, it is important to know ligand-free conformations of the aptamers in order to understand the mechanism of specific binding by ligands. In this paper, we use dynamics simulations on a series of models to characterize the ligand-free and ligand-bound aptamer domain of the c-di-GMP class I (GEMM-I) riboswitch. The results revealed that the ligand-free aptamer has a stable state with a folded P2 and P3 helix, an unfolded P1 helix and open binding pocket. The first Mg ions binding to the aptamer is structurally favorable for the successive c-di-GMP binding. The P1 helix forms when c-di-GMP is successive bound. Three key junctions J1/2, J2/3 and J1/3 in the GEMM-I riboswitch contributing to the formation of P1 helix have been found. The binding of the c-di-GMP ligand to the GEMM-I riboswitch induces the riboswitch's regulation through the direct allosteric communication network in GEMM-I riboswitch from the c-di-GMP binding sites in the J1/2 and J1/3 junctions to the P1 helix, the indirect ones from those in the J2/3 and P2 communicating to P1 helix via the J1/2 and J1/3 media.

Structural Studies of the 3',3'-cGAMP Riboswitch Induced by Cognate and Noncognate Ligands Using Molecular Dynamics Simulation.

Riboswtich RNAs can control gene expression through the structural change induced by the corresponding small-molecule ligands. Molecular dynamics simulations and free energy calculations on the aptamer domain of the 3',3'-cGAMP riboswitch in the ligand-free, cognate-bound and noncognate-bound states were performed to investigate the structural features of the 3',3'-cGAMP riboswitch induced by the 3',3'-cGAMP ligand and the specificity of ligand recognition. The results revealed that the aptamer of the 3',3'-cGAMP riboswitch in the ligand-free state has a smaller binding pocket and a relatively compact structure versus that in the 3',3'-cGAMP-bound state. The binding of the 3',3'-cGAMP molecule to the 3',3'-cGAMP riboswitch induces the rotation of P1 helix through the allosteric communication from the binding sites pocket containing the J1/2, J1/3 and J2/3 junction to the P1 helix. Simultaneously, these simulations also revealed that the preferential binding of the 3',3'-cGAMP riboswitch to its cognate ligand, 3',3'-cGAMP, over its noncognate ligand, c-di-GMP and c-di-AMP. The J1/2 junction in the 3',3'-cGAMP riboswitch contributing to the specificity of ligand recognition have also been found.