Multivariate time series (MTS) data in electronic health records (EHR) pose unique challenges due to their sparsity and irregular time intervals. Existing methods often tend towards imputation or isolated encoding, resulting in suboptimal representation learning. Moreover, most existing research solely focuses on single-shot diagnosis, neglecting the importance of continuous diagnosis, particularly for critically ill patients. Continuous diagnosis provides significant opportunities for timely intervention and rational resource allocation. To address these challenges, we propose an innovative multivariate time-series representation learning for continuous medical diagnosis. Specifically, we first address sparsity issues by combining feature names and record values encoded by time. Then, we utilize a transformer variant with gated units to extract contextual features. Additionally, we introduce Time Update Block, a component that combines the strengths of long short-term memory and attention mechanisms, aimed at improving the model's ability for continuous diagnosis. Based on extensive experimental evaluations on real-world medical datasets, we demonstrate the superior performance of the proposed method.

Despite significant advances in image-text medical visual language modeling, the high cost of fine-grained annotation of images to align radiology reports has led current approaches to focus primarily on semantic alignment between the image and the full report, neglecting the critical diagnostic information contained in the text. This is insufficient in medical scenarios demanding high explainability. To address this problem, in this paper, we introduce radiology reports as images in prompt learning. Specifically, we extract key clinical concepts, lesion locations, and positive labels from easily accessible radiology reports and combine them with an external medical knowledge base to form fine-grained self-supervised signals. Moreover, we propose a novel Report-Concept Textual-Prompt Learning ( RC-TPL ), which aligns radiology reports at multiple levels. In the inference phase, the report-level and concept-level prompts provide rich global and local semantic understanding for X-ray images. Extensive experiments on X-ray image datasets demonstrate the superior performance of our approach with respect to various baselines, especially in the presence of scarce imaging data. Our study not only significantly improves the accuracy of data-constrained medical X-ray diagnosis, but also demonstrates how the integration of domain-specific conceptual knowledge can enhance the explainability of medical image analysis.

Acute Coronary Syndrome (ACS), an emergent medical condition, is intricately linked to environmental factors like air pollution and meteorological conditions. Harnessing regional environmental data, such as weather metrics, can promptly forecast ACS incidence rates, enabling optimised medical resource allocation and increased patient recovery rates. However, the prediction task is rendered complex due to disparities in data collection capabilities across institutions, yielding datasets with analogous features but significant label variations, impeding the application of universal models. Challenges abound due to the heterogeneity of multi-factor data, temporal alignment disparities, and the intricacies of sparse data. To address these challenges, this paper introduces the Domain Adaptation with Multi-factor Associative Structures (DAMAS), a time-series domain adaptation approach based on multi-factor sparse associative frameworks. Augmented by an isomorphic attention-driven variable label mapping scheme and combined with multi-layer perceptrons, our approach skilfully negotiates label imbalances. This results in refined prediction precision connecting environmental factors to regional ACS incidences.

The 12-lead Electrocardiography (ECG) is one of the most commonly used diagnostic tools for cardiovascular disease. Widely available ECG databases and deep learning algorithms present an opportunity to substantially improve the accuracy and scalability of automated ECG abnormal identification. However, existing methods mainly model leads individually and then aggregate them for prediction, ignoring the relationship between leads, which is an important diagnostic reference for clinicians. In this paper, we propose a novel model, called ECGNN, which main consists a feature extractor backbone and a graph neural network module. The feature extractor backbone is a neural network used to extract features of ECG signal for subsequent prediction and initialization of the graph nodes. Specifically, the proposed graph neural network module combines graph convolution and graph pooling into a unified module to generate hierarchical representations of graphs and can be integrated into various feature extractor backbones. Experimental results on two largescale 12-lead ECG databases demonstrate the effectiveness of our proposed model.

Recently, deep learning techniques based on electronic health record (EHR) data have achieved success in medical prediction. However, due to the complexity, heterogeneity nature of EHR data, most previous studies build models based on single-modal data (e.g. the structured data or the unstructured free-text data). Although some studies have trained the models based on multimodal EHR data and achieved more advanced performance, they still suffer from the clinical practicability problems, as they require separate modeling for each medical prediction task. Moreover, they ignore the potential correlation between clinical prediction tasks. In this work, we propose UniMed, a Unified model handles multiple Medical prediction tasks simultaneously by learning from multimodal EHR data. Our UniMed model encodes each input modality separately and uses a transformer decoder followed by task-specific prediction heads to predict each medical task. Experimental results conducted on publicly available EHR dataset demonstrate that there is a time-progressive correlation between medical prediction tasks and show the effectiveness of our method.

Symptom diagnosis in dialogue remains a challenging task because the symptom entities and their status need to be extracted correctly at the same time. Most previous studies treat symptom diagnosis as a classification or sequence labeling task and focus on using single-sentence dialogue as input. Unique from past studies, in this paper, we propose a new framework for dialogue symptom diagnosis, which formulate it as a machine reading comprehension (MRC) task. We first use window-level multi-turn of dialogue as input and extract the symptom entities. Then, we generate a question for each entity to infer the symptom status in the form of question answering (QA). Benefit from the MRC formalization, our proposed framework can encode more informative prior knowledge, which can effectively improve the performance of symptom status inference. Experiments on the Chinese medical dialogue dataset show that the proposed framework outperforms the previous best model and several competitive baselines, which indicates that our framework provides a useful direction for dialogue symptom diagnosis. The code and data are publicly available at https://github.com/zhaoxiongjun/DSD.

Chest X-ray is still the most common and important method for diagnosing pneumonia. However, the analysis of chest radiographs requires professional radiologists, and overreliance on radiologists may lead to erroneous diagnosis or missed diagnosis. Using convolutional neural networks(CNNs) for diagnosis chest diseases on chest X-ray has achieved better results, but most of the previous models are only trained by frontal-view X-ray images. Unique from past studies, in this paper, we proposed a model that can learn multi-view semantic information from chest X-rays to detect pneumonia. Our model includes two stages of feature extraction and feature fusion, and is trained on MIMIC-CXR-JPG dataset, currently the largest publicly available chest x-ray dataset, containing 377,110 JPG format images. We demonstrate that such multi-view weighted feature fusion model outperforms the models that use features only from one view. Our results are better than previous models for pneumonia detection.