Han Yuan

Mingxuan Liu, Siqi Li, Han Yuan et al.

Objective: The proper handling of missing values is critical to delivering reliable estimates and decisions, especially in high-stakes fields such as clinical research. The increasing diversity and complexity of data have led many researchers to develop deep learning (DL)-based imputation techniques. We conducted a systematic review to evaluate the use of these techniques, with a particular focus on data types, aiming to assist healthcare researchers from various disciplines in dealing with missing values. Methods: We searched five databases (MEDLINE, Web of Science, Embase, CINAHL, and Scopus) for articles published prior to August 2021 that applied DL-based models to imputation. We assessed selected publications from four perspectives: health data types, model backbone (i.e., main architecture), imputation strategies, and comparison with non-DL-based methods. Based on data types, we created an evidence map to illustrate the adoption of DL models. Results: We included 64 articles, of which tabular static (26.6%, 17/64) and temporal data (37.5%, 24/64) were the most frequently investigated. We found that model backbone(s) differed among data types as well as the imputation strategy. The "integrated" strategy, that is, the imputation task being solved concurrently with downstream tasks, was popular for tabular temporal (50%, 12/24) and multi-modal data (71.4%, 5/7), but limited for other data types. Moreover, DL-based imputation methods yielded better imputation accuracy in most studies, compared with non-DL-based methods. Conclusion: DL-based imputation models can be customized based on data type, addressing the corresponding missing patterns, and its associated "integrated" strategy can enhance the efficacy of imputation, especially in scenarios where data is complex. Future research may focus on the portability and fairness of DL-based models for healthcare data imputation.

Han Yuan, Mingxuan Liu, Lican Kang et al.

Nowadays, the interpretation of why a machine learning (ML) model makes certain inferences is as crucial as the accuracy of such inferences. Some ML models like the decision tree possess inherent interpretability that can be directly comprehended by humans. Others like artificial neural networks (ANN), however, rely on external methods to uncover the deduction mechanism. SHapley Additive exPlanations (SHAP) is one of such external methods, which requires a background dataset when interpreting ANNs. Generally, a background dataset consists of instances randomly sampled from the training dataset. However, the sampling size and its effect on SHAP remain to be unexplored. In our empirical study on the MIMIC-III dataset, we show that the two core explanations - SHAP values and variable rankings fluctuate when using different background datasets acquired from random sampling, indicating that users cannot unquestioningly trust the one-shot interpretation from SHAP. Luckily, such fluctuation decreases with the increase of the background dataset size. Also, we notice an U-shape in the stability assessment of SHAP variable rankings, demonstrating that SHAP is more reliable in ranking the most and least important variables compared to moderately important ones. Overall, our results suggest that users should take into account how background data affects SHAP results, with improved SHAP stability as the background sample size increases.

Siqi Li, Yilin Ning, Marcus Eng Hock Ong et al.

We propose FedScore, a privacy-preserving federated learning framework for scoring system generation across multiple sites to facilitate cross-institutional collaborations. The FedScore framework includes five modules: federated variable ranking, federated variable transformation, federated score derivation, federated model selection and federated model evaluation. To illustrate usage and assess FedScore's performance, we built a hypothetical global scoring system for mortality prediction within 30 days after a visit to an emergency department using 10 simulated sites divided from a tertiary hospital in Singapore. We employed a pre-existing score generator to construct 10 local scoring systems independently at each site and we also developed a scoring system using centralized data for comparison. We compared the acquired FedScore model's performance with that of other scoring models using the receiver operating characteristic (ROC) analysis. The FedScore model achieved an average area under the curve (AUC) value of 0.763 across all sites, with a standard deviation (SD) of 0.020. We also calculated the average AUC values and SDs for each local model, and the FedScore model showed promising accuracy and stability with a high average AUC value which was closest to the one of the pooled model and SD which was lower than that of most local models. This study demonstrates that FedScore is a privacy-preserving scoring system generator with potentially good generalizability.

Mingxuan Liu, Yilin Ning, Han Yuan et al.

Objective: Shapley additive explanations (SHAP) is a popular post-hoc technique for explaining black box models. While the impact of data imbalance on predictive models has been extensively studied, it remains largely unknown with respect to SHAP-based model explanations. This study sought to investigate the effects of data imbalance on SHAP explanations for deep learning models, and to propose a strategy to mitigate these effects. Materials and Methods: We propose to adjust class distributions in the background and explanation data in SHAP when explaining black box models. Our data balancing strategy is to compose background data and explanation data with an equal distribution of classes. To evaluate the effects of data adjustment on model explanation, we propose to use the beeswarm plot as a qualitative tool to identify "abnormal" explanation artifacts, and quantitatively test the consistency between variable importance and prediction power. We demonstrated our proposed approach in an empirical study that predicted inpatient mortality using the Medical Information Mart for Intensive Care (MIMIC-III) data and a multilayer perceptron. Results: Using the data balancing strategy would allow us to reduce the number of the artifacts in the beeswarm plot, thus mitigating the negative effects of data imbalance. Additionally, with the balancing strategy, the top-ranked variables from the corresponding importance ranking demonstrated improved discrimination power. Discussion and Conclusion: Our findings suggest that balanced background and explanation data could help reduce the noise in explanation results induced by skewed data distribution and improve the reliability of variable importance ranking. Furthermore, these balancing procedures improve the potential of SHAP in identifying patients with abnormal characteristics in clinical applications.

LeoWu TomyEnrique, Xiangcheng Du, Kangliang Liu et al.

Scene text image super-resolution has significantly improved the accuracy of scene text recognition. However, many existing methods emphasize performance over efficiency and ignore the practical need for lightweight solutions in deployment scenarios. Faced with the issues, our work proposes an efficient framework called SGENet to facilitate deployment on resource-limited platforms. SGENet contains two branches: super-resolution branch and semantic guidance branch. We apply a lightweight pre-trained recognizer as a semantic extractor to enhance the understanding of text information. Meanwhile, we design the visual-semantic alignment module to achieve bidirectional alignment between image features and semantics, resulting in the generation of highquality prior guidance. We conduct extensive experiments on benchmark dataset, and the proposed SGENet achieves excellent performance with fewer computational costs. Code is available at https://github.com/SijieLiu518/SGENet

Han Yuan, Chuan Hong, Nguyen Tuan Anh Tran et al.

Pneumothorax is a medical emergency caused by abnormal accumulation of air in the pleural space - the potential space between the lungs and chest wall. On 2D chest radiographs, pneumothorax occurs within the thoracic cavity and outside of the mediastinum and we refer to this area as "lung+ space". While deep learning (DL) has increasingly been utilized to segment pneumothorax lesions in chest radiographs, many existing DL models employ an end-to-end approach. These models directly map chest radiographs to clinician-annotated lesion areas, often neglecting the vital domain knowledge that pneumothorax is inherently location-sensitive. We propose a novel approach that incorporates the lung+ space as a constraint during DL model training for pneumothorax segmentation on 2D chest radiographs. To circumvent the need for additional annotations and to prevent potential label leakage on the target task, our method utilizes external datasets and an auxiliary task of lung segmentation. This approach generates a specific constraint of lung+ space for each chest radiograph. Furthermore, we have incorporated a discriminator to eliminate unreliable constraints caused by the domain shift between the auxiliary and target datasets. Our results demonstrated significant improvements, with average performance gains of 4.6%, 3.6%, and 3.3% regarding Intersection over Union (IoU), Dice Similarity Coefficient (DSC), and Hausdorff Distance (HD). Our research underscores the significance of incorporating medical domain knowledge about the location-specific nature of pneumothorax to enhance DL-based lesion segmentation.

Han Yuan, Chuan Hong, Pengtao Jiang et al.

Background: Pneumothorax is an acute thoracic disease caused by abnormal air collection between the lungs and chest wall. To address the opaqueness often associated with deep learning (DL) models, explainable artificial intelligence (XAI) methods have been introduced to outline regions related to pneumothorax diagnoses made by DL models. However, these explanations sometimes diverge from actual lesion areas, highlighting the need for further improvement. Method: We propose a template-guided approach to incorporate the clinical knowledge of pneumothorax into model explanations generated by XAI methods, thereby enhancing the quality of these explanations. Utilizing one lesion delineation created by radiologists, our approach first generates a template that represents potential areas of pneumothorax occurrence. This template is then superimposed on model explanations to filter out extraneous explanations that fall outside the template's boundaries. To validate its efficacy, we carried out a comparative analysis of three XAI methods with and without our template guidance when explaining two DL models in two real-world datasets. Results: The proposed approach consistently improved baseline XAI methods across twelve benchmark scenarios built on three XAI methods, two DL models, and two datasets. The average incremental percentages, calculated by the performance improvements over the baseline performance, were 97.8% in Intersection over Union (IoU) and 94.1% in Dice Similarity Coefficient (DSC) when comparing model explanations and ground-truth lesion areas. Conclusions: In the context of pneumothorax diagnoses, we proposed a template-guided approach for improving AI explanations. We anticipate that our template guidance will forge a fresh approach to elucidating AI models by integrating clinical domain expertise.

Yingjian Chen, Haoran Liu, Yinhong Liu et al.

Large language models (LLMs) are widely used, but they often generate subtle factual errors, especially in long-form text. These errors are fatal in some specialized domains such as medicine. Existing fact-checking with grounding documents methods face two main challenges: (1) they struggle to understand complex multihop relations in long documents, often overlooking subtle factual errors; (2) most specialized methods rely on pairwise comparisons, requiring multiple model calls, leading to high resource and computational costs. To address these challenges, we propose GraphCheck, a fact-checking framework that uses extracted knowledge graphs to enhance text representation. Graph Neural Networks further process these graphs as a soft prompt, enabling LLMs to incorporate structured knowledge more effectively. Enhanced with graph-based reasoning, GraphCheck captures multihop reasoning chains that are often overlooked by existing methods, enabling precise and efficient fact-checking in a single inference call. Experimental results on seven benchmarks spanning both general and medical domains demonstrate up to a 7.1% overall improvement over baseline models. Notably, GraphCheck outperforms existing specialized fact-checkers and achieves comparable performance with state-of-the-art LLMs, such as DeepSeek-V3 and OpenAI-o1, with significantly fewer parameters.

Han Yuan, Lican Kang, Yong Li

While deep learning has exhibited remarkable predictive capabilities in various medical image tasks, its inherent black-box nature has hindered its widespread implementation in real-world healthcare settings. Our objective is to unveil the decision-making processes of deep learning models in the context of glaucoma classification by employing several Class Activation Map (CAM) techniques to generate model focus regions and comparing them with clinical domain knowledge of the anatomical area (optic cup, optic disk, and blood vessels). Four deep neural networks, including VGG-11, ResNet-18, DeiT-Tiny, and Swin Transformer-Tiny, were developed using binary diagnostic labels of glaucoma and five CAM methods (Grad-CAM, XGrad-CAM, Score-CAM, Eigen-CAM, and Layer-CAM) were employed to highlight the model focus area. We applied the paired-sample t-test to compare the percentage of anatomies in the model focus area to the proportion of anatomies in the entire image. After that, Pearson's and Spearman's correlation tests were implemented to examine the relationship between model predictive ability and the percentage of anatomical structures in the model focus area. On five public glaucoma datasets, all deep learning models consistently displayed statistically significantly higher percentages of anatomical structures in the focus area than the proportions of anatomical structures in the entire image. Also, we validated the positive relationship between the percentage of anatomical structures in the focus area and model predictive performance. Our study provides evidence of the convergence of decision logic between deep neural networks and human clinicians through rigorous statistical tests. We anticipate that it can help alleviate clinicians' concerns regarding the trustworthiness of deep learning in healthcare. For reproducibility, the code and dataset have been released at GitHub.

Feiyang Li, Yingjian Chen, Haoran Liu et al.

Large Language Models (LLMs) have shown remarkable progress in medical question answering (QA), yet their effectiveness remains predominantly limited to English due to imbalanced multilingual training data and scarce medical resources for low-resource languages. To address this critical language gap in medical QA, we propose Multilingual Knowledge Graph-based Retrieval Ranking (MKG-Rank), a knowledge graph-enhanced framework that enables English-centric LLMs to perform multilingual medical QA. Through a word-level translation mechanism, our framework efficiently integrates comprehensive English-centric medical knowledge graphs into LLM reasoning at a low cost, mitigating cross-lingual semantic distortion and achieving precise medical QA across language barriers. To enhance efficiency, we introduce caching and multi-angle ranking strategies to optimize the retrieval process, significantly reducing response times and prioritizing relevant medical knowledge. Extensive evaluations on multilingual medical QA benchmarks across Chinese, Japanese, Korean, and Swahili demonstrate that MKG-Rank consistently outperforms zero-shot LLMs, achieving maximum 35.03% increase in accuracy, while maintaining an average retrieval time of only 0.0009 seconds.

Bo Hu, Han Yuan, Vlad Pandelea et al.

The rapid advancement of large language models (LLMs) has sparked widespread adoption across diverse applications, making robust evaluation frameworks crucial for assessing their performance. While conventional evaluation metrics remain applicable for shorter texts, their efficacy diminishes when evaluating the quality of long-form answers. This limitation is particularly critical in real-world scenarios involving extended questions, extensive context, and long-form answers, such as financial analysis or regulatory compliance. In this paper, we use a practical financial use case to illustrate applications that handle "long question-context-answer triplets". We construct a real-world financial dataset comprising long triplets and demonstrate the inadequacies of traditional metrics. To address this, we propose an effective Extract, Match, and Score (EMS) evaluation approach tailored to the complexities of long-form LLMs' outputs, providing practitioners with a reliable methodology for assessing LLMs' performance in complex real-world scenarios.

Han Yuan, Yilin Wu, Li Zhang et al.

Small language models (SLMs) are increasingly used for financial classification due to their fast inference and local deployability. However, compared with large language models, SLMs are more prone to factual hallucinations in reasoning and exhibit weaker classification performance. This raises a natural question: Can mitigating factual hallucinations improve SLMs' financial classification? To address this, we propose a three-step pipeline named AAAI (Association Identification, Automated Detection, and Adaptive Inference). Experiments on three representative SLMs reveal that: (1) factual hallucinations are positively correlated with misclassifications; (2) encoder-based verifiers effectively detect factual hallucinations; and (3) incorporating feedback on factual errors enables SLMs' adaptive inference that enhances classification performance. We hope this pipeline contributes to trustworthy and effective applications of SLMs in finance.

Han Yuan, Yue Zhao, Li Zhang et al.

Structured output from large language models (LLMs) has enhanced efficiency in processing generated information and is increasingly adopted in industrial applications. Prior studies have investigated the impact of structured output on LLMs' generation quality, often presenting one-way findings. Some suggest that structured format enhances completeness and factual accuracy, while others argue that it restricts the reasoning capacity of LLMs and leads to reductions in standard evaluation metrics. Potential limitations of these assessments include restricted testing scenarios, weakly controlled comparative settings, and reliance on coarse metrics. In this work, we present a refined analysis using causal inference. Based on one assumed and two guaranteed constraints, we derive five potential causal structures characterizing the influence of structured output on LLMs' generation: (1) collider without m-bias, (2) collider with m-bias, (3) single cause from instruction, (4) single cause from output format, and (5) independence. Across seven public and one developed reasoning tasks, we find that coarse metrics report positive, negative, or neutral effects of structured output on GPT-4o's generation. However, causal inference reveals no causal impact in 43 out of 48 scenarios. In the remaining 5, 3 involve multifaceted causal structures influenced by concrete instructions.

Han Yuan, Li Zhang, Zheng Ma

Language models (LMs) have exhibited exceptional versatility in reasoning and in-depth financial analysis through their proprietary information processing capabilities. Previous research focused on evaluating classification performance while often overlooking explainability or pre-conceived that refined explanation corresponds to higher classification accuracy. Using a public dataset in finance domain, we quantitatively evaluated self-explanations by LMs, focusing on their factuality and causality. We identified the statistically significant relationship between the accuracy of classifications and the factuality or causality of self-explanations. Our study built an empirical foundation for approximating classification confidence through self-explanations and for optimizing classification via proprietary reasoning.

Han Yuan, Chuan Hong

Active learning selects the most informative samples from the unlabelled dataset to annotate in the context of a limited annotation budget. While numerous methods have been proposed for subsequent sample selection based on an initialized model, scant attention has been paid to the indispensable phase of active learning: selecting samples for model cold-start initialization. Most of the previous studies resort to random sampling or naive clustering. However, random sampling is prone to fluctuation, and naive clustering suffers from convergence speed, particularly when dealing with high-dimensional data such as imaging data. In this work, we propose to integrate foundation models with clustering methods to select samples for cold-start active learning initialization. Foundation models refer to those trained on massive datasets by the self-supervised paradigm and capable of generating informative and compacted embeddings for various downstream tasks. Leveraging these embeddings to replace raw features such as pixel values, clustering quickly converges and identifies better initial samples. For a comprehensive comparison, we included a classic ImageNet-supervised model to acquire embeddings. Experiments on two clinical tasks of image classification and segmentation demonstrated that foundation model-based clustering efficiently pinpointed informative initial samples, leading to models showcasing enhanced performance than the baseline methods. We envisage that this study provides an effective paradigm for future cold-start active learning.

Feng Xie, Han Yuan, Yilin Ning et al.

Objective: Temporal electronic health records (EHRs) can be a wealth of information for secondary uses, such as clinical events prediction or chronic disease management. However, challenges exist for temporal data representation. We therefore sought to identify these challenges and evaluate novel methodologies for addressing them through a systematic examination of deep learning solutions. Methods: We searched five databases (PubMed, EMBASE, the Institute of Electrical and Electronics Engineers [IEEE] Xplore Digital Library, the Association for Computing Machinery [ACM] digital library, and Web of Science) complemented with hand-searching in several prestigious computer science conference proceedings. We sought articles that reported deep learning methodologies on temporal data representation in structured EHR data from January 1, 2010, to August 30, 2020. We summarized and analyzed the selected articles from three perspectives: nature of time series, methodology, and model implementation. Results: We included 98 articles related to temporal data representation using deep learning. Four major challenges were identified, including data irregularity, data heterogeneity, data sparsity, and model opacity. We then studied how deep learning techniques were applied to address these challenges. Finally, we discuss some open challenges arising from deep learning. Conclusion: Temporal EHR data present several major challenges for clinical prediction modeling and data utilization. To some extent, current deep learning solutions can address these challenges. Future studies can consider designing comprehensive and integrated solutions. Moreover, researchers should incorporate additional clinical domain knowledge into study designs and enhance the interpretability of the model to facilitate its implementation in clinical practice.

Han Yuan, Feng Xie, Marcus Eng Hock Ong et al.

Background: Medical decision-making impacts both individual and public health. Clinical scores are commonly used among a wide variety of decision-making models for determining the degree of disease deterioration at the bedside. AutoScore was proposed as a useful clinical score generator based on machine learning and a generalized linear model. Its current framework, however, still leaves room for improvement when addressing unbalanced data of rare events. Methods: Using machine intelligence approaches, we developed AutoScore-Imbalance, which comprises three components: training dataset optimization, sample weight optimization, and adjusted AutoScore. All scoring models were evaluated on the basis of their area under the curve (AUC) in the receiver operating characteristic analysis and balanced accuracy (i.e., mean value of sensitivity and specificity). By utilizing a publicly accessible dataset from Beth Israel Deaconess Medical Center, we assessed the proposed model and baseline approaches in the prediction of inpatient mortality. Results: AutoScore-Imbalance outperformed baselines in terms of AUC and balanced accuracy. The nine-variable AutoScore-Imbalance sub-model achieved the highest AUC of 0.786 (0.732-0.839) while the eleven-variable original AutoScore obtained an AUC of 0.723 (0.663-0.783), and the logistic regression with 21 variables obtained an AUC of 0.743 (0.685-0.800). The AutoScore-Imbalance sub-model (using down-sampling algorithm) yielded an AUC of 0. 0.771 (0.718-0.823) with only five variables, demonstrating a good balance between performance and variable sparsity. Conclusions: The AutoScore-Imbalance tool has the potential to be applied to highly unbalanced datasets to gain further insight into rare medical events and to facilitate real-world clinical decision-making.

Feng Xie, Yilin Ning, Han Yuan et al.

Scoring systems are highly interpretable and widely used to evaluate time-to-event outcomes in healthcare research. However, existing time-to-event scores are predominantly created ad-hoc using a few manually selected variables based on clinician's knowledge, suggesting an unmet need for a robust and efficient generic score-generating method. AutoScore was previously developed as an interpretable machine learning score generator, integrated both machine learning and point-based scores in the strong discriminability and accessibility. We have further extended it to time-to-event data and developed AutoScore-Survival, for automatically generating time-to-event scores with right-censored survival data. Random survival forest provides an efficient solution for selecting variables, and Cox regression was used for score weighting. We illustrated our method in a real-life study of 90-day mortality of patients in intensive care units and compared its performance with survival models (i.e., Cox) and the random survival forest. The AutoScore-Survival-derived scoring model was more parsimonious than survival models built using traditional variable selection methods (e.g., penalized likelihood approach and stepwise variable selection), and its performance was comparable to survival models using the same set of variables. Although AutoScore-Survival achieved a comparable integrated area under the curve of 0.782 (95% CI: 0.767-0.794), the integer-valued time-to-event scores generated are favorable in clinical applications because they are easier to compute and interpret. Our proposed AutoScore-Survival provides an automated, robust and easy-to-use machine learning-based clinical score generator to studies of time-to-event outcomes. It provides a systematic guideline to facilitate the future development of time-to-event scores for clinical applications.