Medical Datasets Research Articles

An Improved Deep Learning Framework for Multimodal Medical Data Analysis

Lung disease is one of the leading causes of death worldwide. This emphasizes the need for early diagnosis in order to provide appropriate treatment and save lives. Physicians typically require information about patients’ clinical symptoms, various laboratory and pathology tests, along with chest X-rays to confirm the diagnosis of lung disease. In this study, we present a transformer-based multimodal deep learning approach that incorporates imaging and clinical data for effective lung disease diagnosis on a new multimodal medical dataset. The proposed method employs a cross-attention transformer module to merge features from the heterogeneous modalities. Then unified fused features are used for disease classification. The experiments were performed and evaluated on several classification metrics to illustrate the performance of the proposed approach. The study’s results revealed that the proposed method achieved an accuracy of 95% in terms of accurate classification of tuberculosis and outperformed other traditional fusion methods on multimodal tuberculosis data used in this study.

Medical Federated Model with Mixture of Personalized and Shared Components.

Although data-driven methods usually have noticeable performance on disease diagnosis and treatment, they are suspected of leakage of privacy due to collecting data for model training. Recently, federated learning provides a secure and trustable alternative to collaboratively train model without any exchange of medical data among multiple institutes. Therefore, it has draw much attention due to its natural merit on privacy protection. However, when heterogenous medical data exists between different hospitals, federated learning usually has to face with degradation of performance. In the paper, we propose a new personalized framework of federated learning to handle the problem. It successfully yields personalized models based on awareness of similarity between local data, and achieves better tradeoff between generalization and personalization than existing methods. After that, we further design a differentially sparse regularizer to improve communication efficiency during procedure of model training. Additionally, we propose an effective method to reduce the computational cost, which improves computation efficiency significantly. Furthermore, we collect five real medical datasets, including two public medical image datasets and three private multi-center clinical diagnosis datasets, and evaluate its performance by conducting nodule classification, tumor segmentation, and clinical risk prediction tasks. Comparing with 14 existing related methods, the proposed method successfully achieves the best model performance, and meanwhile up to 60% improvement of communication efficiency. Source code is public, and can be accessed at: https://github.com/ ApplicationTechnologyOfMedicalBigData/pFedNet-code.

HI-Net: A novel histopathologic image segmentation model for metastatic breast cancer via lightweight dataset construction

Since 2020, breast cancer has remained the most prevalent cancer worldwide and the World Health Organisation projects significant increases by 2040, with new cases expected to exceed 3 million annually (a 40% increase) and deaths to surpass 1 million (a 50% increase), highlighting the urgent need for advancements in detection and treatment. Current detection of metastasis is highly dependent on labour-intensive and error-prone pathological examination of large-scale biotissue. Given the high-resolution (100,000 × 100,000 gigapixels) but limited quantity of open-source pathological slide datasets, existing deep learning models face preprocessing challenges. This paper introduces HI-Net, a high-speed panoramic feature-extraction pyramid network for rapid and accurate detection of metastatic breast cancer, balancing panoramic segmentation and local attention. Additionally, a lightweight pathological slide dataset optimised for 512 x 512-pixel resolution, derived from downsampled and reassembled competitive datasets, accelerates training and reduces computational costs. HI-Net demonstrates superior performance on existing medical imaging competition datasets and our lightweight dataset, evidencing its effectiveness across datasets and potential for contributing to the generalisation of intelligent diagnostics.

Implementasi K-Means Clustering pada Pengelompokan Pasien Penyakit Jantung

Heart disease is a prominent global health concern, necessitating early identification and patient grouping for effective management. This study employs the K-Means clustering algorithm with a medical dataset of 303 patients, encompassing various attributes. These include Age, Gender, Chest Pain Type, Blood Pressure, Serum Cholesterol Level, Fasting Blood Sugar, Resting Electrocardiographic Results, Maximum Heart Rate, Angina, ST Depression, and Slope of the ST Segment. The goal is to categorize patients into four clusters based on chest pain types, a crucial symptom indicating disease severity. The computation concludes after the sixth iteration, revealing Cluster 1 (27 patients), Cluster 2 (135 patients), Cluster 3 (15 patients), and Cluster 4 (126 patients). Collaborative analysis with medical experts highlights that Cluster 1, mainly comprising older males, exhibits high-risk indicators. While this grouping aids in personalized treatment strategy development, further clinical validation involving more experts and datasets is imperative for enhanced reliability.

Boosting medical diagnostics with a novel gradient-based sample selection method

In the rapidly expanding landscape of medical data, the need for innovative approaches to maximize classification performance has become increasingly critical. As data volumes grow, ensuring that diagnostic systems work with accurate and relevant data is paramount for effective and generalizable classification. This study introduces a novel gradient-based sample selection method, the first of its kind in the literature, specifically designed to enhance classification accuracy by removing redundant and non-informative data. Unlike traditional methods that focus solely on feature selection, this approach integrates an advanced sample selection technique to optimize the input data, leading to more accurate and efficient diagnostics. The method is validated on multiple disease datasets, including the Wisconsin Diagnostic Breast Cancer (WDBC) dataset and the Cleveland Coronary Artery Disease (CAD) dataset, demonstrating its broad applicability and effectiveness. To address dataset imbalance, the Adaptive Synthetic Sampling (ADASYN) method is employed, followed by Particle Swarm Optimization (PSO) for feature selection. The refined datasets are then classified using a Support Vector Machine (SVM), showing that even traditional classifiers can achieve substantial improvements when enhanced with advanced sample selection. The results underscore the critical importance of precise sample selection in boosting classification performance, setting a new standard for computer-aided diagnostics and paving the way for future innovations in handling large and complex medical datasets.

Computational modeling of tumor invasion from limited and diverse data in Glioblastoma

For diseases with high morbidity rates such as Glioblastoma Multiforme, the prognostic and treatment planning pipeline requires a comprehensive analysis of imaging, clinical, and molecular data. Many mutations have been shown to correlate strongly with the median survival rate and response to therapy of patients. Studies have demonstrated that these mutations manifest as specific visual biomarkers in tumor imaging modalities such as MRI. To minimize the number of invasive procedures on a patient and for the overall resource optimization for the prognostic and treatment planning process, the correlation of imaging and molecular features has garnered much interest. While the tumor mass is the most significant feature, the impacted tissue surrounding the tumor is also a significant biomarker contributing to the visual manifestation of mutations — which has not been studied as extensively. The pattern of tumor growth impacts the surrounding tissue accordingly, which is a reflection of tumor properties as well. Modeling how the tumor growth impacts the surrounding tissue can reveal important information about the patterns of tumor enhancement, which in turn has significant diagnostic and prognostic value. This paper presents the first work to automate the computational modeling of the impacted tissue surrounding the tumor using generative deep learning. The paper isolates and quantifies the impact of the Tumor Invasion (TI) on surrounding tissue based on change in mutation status, subsequently assessing its prognostic value. Furthermore, a TI Generative Adversarial Network (TI-GAN) is proposed to model the tumor invasion properties. Extensive qualitative and quantitative analyses, cross-dataset testing, and radiologist blind tests are carried out to demonstrate that TI-GAN can realistically model the tumor invasion under practical challenges of medical datasets such as limited data and high intra-class heterogeneity.

Potential of multimodal large language models for data mining of medical images and free-text reports

Medical images and radiology reports are essential for physicians to diagnose medical conditions. However, the vast diversity and cross-source heterogeneity inherent in these data have posed significant challenges to the generalizability of current data-mining methods for clinical decision-making. Recently, multimodal large language models (MLLMs), especially Gemini-Vision-series (Gemini) and GPT-4-series (GPT-4) models, have revolutionized numerous domains, significantly impacting the medical field. In this study, we conducted a detailed evaluation of the performance of the Gemini series models (including Gemini-1.0-Pro-Vision, Gemini-1.5-Pro, and Gemini-1.5-Flash) and GPT series models (including GPT-4o, GPT-4-Turbo, and GPT-3.5-Turbo) across 14 medical datasets, covering 5 medical imaging categories (dermatology, radiology, dentistry, ophthalmology, and endoscopy) and 3 radiology report datasets. The investigated tasks encompass disease classification, lesion segmentation, anatomical localization, disease diagnosis, report generation, and lesion detection. Moreover, we also validated the performance of the Claude-3-Opus, Yi-Large, Yi-Large-Turbo, and LLaMA 3 models to gain a comprehensive understanding of the MLLM models in the medical field. Our experimental results demonstrated that Gemini-series models excelled in report generation and lesion detection but faces challenges in disease classification and anatomical localization. Conversely, GPT-series models exhibited proficiency in lesion segmentation and anatomical localization but encountered difficulties in disease diagnosis and lesion detection. Additionally, both the Gemini series and GPT series contain models that have demonstrated commendable generation efficiency. While both models hold promise in reducing physician workload, alleviating pressure on limited healthcare resources, and fostering collaboration between clinical practitioners and artificial intelligence technologies, substantial enhancements and comprehensive validations remain imperative before clinical deployment.

Impact of feature selection techniques on machine learning and deep learning techniques for cardiovascular disease prediction-an analysis

Cardio vascular disease is one of the life-threatening diseases which affects individuals worldwide. Early diagnosis may allow for the prevention or mitigation of cardiovascular diseases, which may minor mortality rates. A feasible Deep Learning and Machine Learning algorithms are used to find risk variables. Machine Learning and Deep Learning system anticipates heart diseases early on and reduce death rates from clinical data. To detect heart diseases or determine the patient's severity level, numerous research studies recently used various machine learning techniques. The volume of internationally recognised medical data sets is growing in terms of both qualities and records. This paper delivers brief outline of various feature extraction methods such as LASSO, RELIEF, RFE, MR-MR and RELIEF on deep learning and machine learning techniques for diagnosing cardiac disease. The performance metrics taken into consideration are Accuracy, Precision, Recall, F1score and the error measures are least Mean Squared Error and Mean Absolute Error. The feature selection methods with more features selected outpaced other approaches. Finally, crucial findings from the evaluated studies are outlined.

Double kernel and minimum variance embedded broad learning system based autoencoder for one-class classification

One-class classification methods are often used for anomaly detection in healthcare, quality control in manufacturing, and fraud detection in financial services. Particularly in medical datasets, instances of rare cancers are considerably fewer in comparison to those of healthy individuals. One-class classification methods, which concentrate on learning models using just samples from a specified class, have drawn increasing attention as a solution to this problem. To enhance the model’s ability to capture essential feature information, we propose a novel strategy in this paper. This strategy involves the initial application of double kernel mapping, translating the original features into a new feature space. The autoencoder based on the Broad Learning System is then modified to incorporate the minimum variance so that the model can learn the potential feature information. We validated the effectiveness of our proposed strategy using UCI dataset and concordia digits dataset. To compare the performance with other one-class methods, the average F1 scores and the average G−mean are employed as comparative measures. Based on experimental results, it is evident that DKVBLS-AE not only outperforms but also demonstrates greater stability compared to the existing methods.

Lazy Resampling: Fast and information preserving preprocessing for deep learning

Background and Objective:Preprocessing of data is a vital step for almost all deep learning workflows. In computer vision, manipulation of data intensity and spatial properties can improve network stability and can provide an important source of generalisation for deep neural networks. Models are frequently trained with preprocessing pipelines composed of many stages, but these pipelines come with a drawback; each stage that resamples the data costs time, degrades image quality, and adds bias to the output. Long pipelines can also be complex to design, especially in medical imaging, where cropping data early can cause significant artifacts. Methods:We present Lazy Resampling, a software that rephrases spatial preprocessing operations as a graphics pipeline. Rather than each transform individually modifying the data, the transforms generate transform descriptions that are composited together into a single resample operation wherever possible. This reduces pipeline execution time and, most importantly, limits signal degradation. It enables simpler pipeline design as crops and other operations become non-destructive. Lazy Resampling is designed in such a way that it provides the maximum benefit to users without requiring them to understand the underlying concepts or change the way that they build pipelines. Results:We evaluate Lazy Resampling by comparing traditional pipelines and the corresponding lazy resampling pipeline for the following tasks on Medical Segmentation Decathlon datasets. We demonstrate lower information loss in lazy pipelines vs. traditional pipelines. We demonstrate that Lazy Resampling can avoid catastrophic loss of semantic segmentation label accuracy occurring in traditional pipelines when passing labels through a pipeline and then back through the inverted pipeline. Finally, we demonstrate statistically significant improvements when training UNets for semantic segmentation. Conclusion:Lazy Resampling reduces the loss of information that occurs when running processing pipelines that traditionally have multiple resampling steps and enables researchers to build simpler pipelines by making operations such as rotation and cropping effectively non-destructive. It makes it possible to invert labels back through a pipeline without catastrophic loss of accuracy.A reference implementation for Lazy Resampling can be found at https://github.com/KCL-BMEIS/LazyResampling. Lazy Resampling is being implemented as a core feature in MONAI, an open source python-based deep learning library for medical imaging, with a roadmap for a full integration.

Modeling of joint extraction of entity relationships in clinical electronic medical records

The advancement of medical informatization necessitates extracting entities and their relationships from electronic medical records. Presently, research on electronic medical records predominantly concentrates on single-entity relationship extraction. However, clinical electronic medical records frequently exhibit overlapping complex entity relationships, thereby heightening the challenge of information extraction. To rectify the absence of a clinical medical relationship extraction dataset, this study utilizes electronic medical records from 584 patients in a hospital to create a compact clinical medical relationship extraction dataset. To address the pipelined relationship extraction model’s limitation in overlooking the one-to-many correlation problem between entities and relationships, this paper introduces a cascading relationship extraction model. This model integrates the MacBERT pre-training model, gated recurrent network, and multi-head self-attention mechanism to enhance the extraction of text features. Simultaneously, adversarial learning is incorporated to bolster the model’s robustness. In scenarios involving one-to-many relationships between entities, a two-phase task is employed. Initially, the main entity is predicted, followed by predicting the associated object and their correspondences. Employing this cascade-structured approach enables the model to flexibly manage intricate entity relationships, thereby enhancing extraction accuracy. Experimental results demonstrate the model’s efficiency, yielding F1-scores of 82.8%, 76.8%, and 88.2% for fulfilling relational extraction requirements and tasks on DuIE, CHIP-CDEE, and private datasets, respectively. These scores represent improvements over the benchmark model. The findings indicate the model’s applicability in practical domains, particularly in tasks such as biomedical information extraction.

Intelligent cardiovascular disease diagnosis using deep learning enhanced neural network with ant colony optimization

To identify patterns in big medical datasets and use Deep Learning and Machine Learning (ML) to reliably diagnose Cardio Vascular Disease (CVD), researchers are currently delving deeply into these fields. Training on large datasets and producing highly accurate validation results is exceedingly difficult. Furthermore, early and precise diagnosis is necessary due to the increased global prevalence of cardiovascular disease (CVD). However, the increasing complexity of healthcare datasets makes it challenging to detect feature connections and produce precise predictions. To address these issues, the Intelligent Cardiovascular Disease Diagnosis based on Ant Colony Optimisation with Enhanced Deep Learning (ICVD-ACOEDL) model was developed. This model employs feature selection (FS) and hyperparameter optimization to diagnose CVD. Applying a min–max scaler, medical data is first consistently prepared. The key feature that sets ICVD-ACOEDL apart is the use of Ant Colony Optimisation (ACO) to select an optimal feature subset, which in turn helps to upgrade the performance of the ensuring deep learning enhanced neural network (DLENN) classifier. The model reforms the hyperparameters of DLENN for CVD classification using Bayesian optimization. Comprehensive evaluations on benchmark medical datasets show that ICVD-ACOEDL exceeds existing techniques, indicating that it could have a significant impact on CVD diagnosis. The model furnishes a workable way to increase CVD classification efficiency and accuracy in real-world medical situations by incorporating ACO for feature selection, min–max scaling for data pre-processing, and Bayesian optimization for hyperparameter tweaking.

Hybrid multiple instance learning network for weakly supervised medical image classification and localization

Weakly supervised medical image analysis is of great significance for computer-aided diagnosis due to the difficulty in obtaining accurately labeled medical data. In this paper, we proposed a new Multi-instance Learning (MIL) framework called HybridMIL integrating CNN Convolutional Neural Networks (CNN) and Broad Learning Systems (BLS). Our HybridMIL can overcome several challenging issues over existing MIL methods based on either CNN or BLS alone: (i) Multiple levels (i.e., different resolutions) of feature information can be simultaneously extracted through a newly proposed instance-level feature enhancement (IFE) module; (ii) Global-level semantic information contained in the deep layers can be better represented under the global-level semantic enhancement (GSE) module; (iii) Hybrid feature fusion (HFF) module is newly designed to effectively fuse and align the multi-level outputs of IFE and global-level semantic information of GSE for subsequent classification and localization tasks. The proposed HybridMIL is evaluated on various public medical and MIL benchmark datasets. The results indicate that HybridMIL surpasses other recent MIL models in terms of classification and localization performance by up to 8.5% and 9.0%, respectively. Lastly, we demonstrate the highly competitive performance of HybridMIL in general MIL problems, going beyond weakly supervised medical image analysis.

Leukemia detection and classification using computer-aided diagnosis system with falcon optimization algorithm and deep learning

Leukemia is a type of blood tumour that occurs because of abnormal enhancement in WBCs (white blood cells) in the bone marrow of the human body. Blood-forming tissue cancer influences the lymphatic and bone marrow system. The early diagnosis and detection of leukaemia, i.e., the accurate difference of malignant leukocytes with little expense at the beginning of the disease, is a primary challenge in the disease analysis field. Despite the higher occurrence of leukemia, there is a lack of flow cytometry tools, and the procedures accessible at medical diagnostics centres are time-consuming. Distinct researchers have implemented computer-aided diagnostic (CAD) and machine learning (ML) methods for laboratory image analysis, aiming to manage the restrictions of late leukemia analysis. This study proposes a new Falcon optimization algorithm with deep convolutional neural network for Leukemia detection and classification (FOADCNN-LDC) technique. The main objective of the FOADCNN-LDC technique is to classify and recognize leukemia. The FOADCNN-LDC technique utilizes a median filtering (MF) based noise removal process to eradicate the image noise. Besides, the FOADCNN-LDC technique employs the ShuffleNetv2 model for the feature extraction process. Moreover, the detection and classification of the leukemia process are performed by utilizing the convolutional denoising autoencoder (CDAE) model. The FOA is implemented to select the hyperparameter of the CDAE model. The simulation process of the FOADCNN-LDC approach is performed on a benchmark medical dataset. The investigational analysis of the FOADCNN-LDC approach highlighted a superior accuracy value of 99.62% over existing techniques.

Dual Channel‐Spatial Self‐Attention Transformer and CNN synergy network for 3D medical image segmentation

Even though the Vision Transformer leverages the self-attention mechanism to capture long-range dependencies, showing significant potential in medical image segmentation, the limited annotations in the image dataset make it difficult for the Transformer model to extract different global features, resulting in attention collapse and generating similar or identical attention maps. Previous studies have attempted to solve the problem by integrating convolutional neural layers into Transformer-based architectures. However, improper integration may lead to the inability of the model to effectively capture local and global information in both spatial and channel dimensions. To address the above issue, we propose a hybrid architecture using the Dual Channel-Spatial Self-Attention Transformer and CNN Synergy Network (DTC-SUNETR) for medical image segmentation. Specifically, we redesigned the self-attention mechanism. A novel Channel-Spatial Self-Attention (CSSA) block is introduced to integrate the enhanced channel and spatial self-attention mechanism to capture the global relationship and local structure among image features. This helps the model to more comprehensively understand the inter-dependencies between different channels and capture the relationships between different pixels, thus enhancing the feature representation of the corresponding dimensions. Simultaneously, it also improves the overall computational efficiency of the network. Extensive experiments on four different medical image segmentation datasets, including Synapse, ACDC, Brain Tumor, and Lung Tumor, demonstrate the superiority of the proposed DTC-SUNETR over state-of-the-art methods.

Enhancing the Interpretability of Malaria and Typhoid Diagnosis with Explainable AI and Large Language Models.

Malaria and Typhoid fever are prevalent diseases in tropical regions, and both are exacerbated by unclear protocols, drug resistance, and environmental factors. Prompt and accurate diagnosis is crucial to improve accessibility and reduce mortality rates. Traditional diagnosis methods cannot effectively capture the complexities of these diseases due to the presence of similar symptoms. Although machine learning (ML) models offer accurate predictions, they operate as "black boxes" with non-interpretable decision-making processes, making it challenging for healthcare providers to comprehend how the conclusions are reached. This study employs explainable AI (XAI) models such as Local Interpretable Model-agnostic Explanations (LIME), and Large Language Models (LLMs) like GPT to clarify diagnostic results for healthcare workers, building trust and transparency in medical diagnostics by describing which symptoms had the greatest impact on the model's decisions and providing clear, understandable explanations. The models were implemented on Google Colab and Visual Studio Code because of their rich libraries and extensions. Results showed that the Random Forest model outperformed the other tested models; in addition, important features were identified with the LIME plots while ChatGPT 3.5 had a comparative advantage over other LLMs. The study integrates RF, LIME, and GPT in building a mobile app to enhance the interpretability and transparency in malaria and typhoid diagnosis system. Despite its promising results, the system's performance is constrained by the quality of the dataset. Additionally, while LIME and GPT improve transparency, they may introduce complexities in real-time deployment due to computational demands and the need for internet service to maintain relevance and accuracy. The findings suggest that AI-driven diagnostic systems can significantly enhance healthcare delivery in environments with limited resources, and future works can explore the applicability of this framework to other medical conditions and datasets.

Medical image analysis using improved SAM-Med2D: segmentation and classification perspectives

Recently emerged SAM-Med2D represents a state-of-the-art advancement in medical image segmentation. Through fine-tuning the Large Visual Model, Segment Anything Model (SAM), on extensive medical datasets, it has achieved impressive results in cross-modal medical image segmentation. However, its reliance on interactive prompts may restrict its applicability under specific conditions. To address this limitation, we introduce SAM-AutoMed, which achieves automatic segmentation of medical images by replacing the original prompt encoder with an improved MobileNet v3 backbone. The performance on multiple datasets surpasses both SAM and SAM-Med2D. Current enhancements on the Large Visual Model SAM lack applications in the field of medical image classification. Therefore, we introduce SAM-MedCls, which combines the encoder of SAM-Med2D with our designed attention modules to construct an end-to-end medical image classification model. It performs well on datasets of various modalities, even achieving state-of-the-art results, indicating its potential to become a universal model for medical image classification.

SMedIR: secure medical image retrieval framework with ConvNeXt-based indexing and searchable encryption in the cloud

The security and privacy of medical images are crucial due to their sensitive nature and the potential for severe consequences from unauthorized modifications, including data breaches and inaccurate diagnoses. This paper introduces a method for lossless medical image retrieval from encrypted images stored on third-party clouds. The proposed approach employs a symmetric integrity-centric image encryption scheme, leveraging multiple chaotic maps and cryptographic hash techniques, to ensure lossless image reconstruction. Medical images are first encrypted by the image owners and converted into hashcodes encapsulating essential features using a deep hashing technique with the ConvNeXt network as the backbone in parallel. To ensure index privacy, these hashcodes are encrypted in a searchable manner. The encrypted medical images, along with a secure index, are subsequently uploaded to cloud storage. Authorized medical image users can request similar medical images for diagnostic purposes by submitting a query image, from which a search trapdoor is generated and sent to the cloud. The retrieval process involves a secure similar image search over the encrypted indexes, followed by decryption along with integrity verification of the retrieved images. The proposed method has been rigorously tested on three standard medical datasets, demonstrating an improvement of 5-20% in retrieval accuracy compared to standard baselines. Formal security analysis and experimental results indicate that the proposed scheme offers enhanced security and retrieval accuracy, making it an effective solution for the encrypted storage and secure retrieval of medical image data.

Federated learning based multi‐head attention framework for medical image classification

AbstractIn this study, we propose a novel Federated Learning Based Multi‐Head Attention (FBMA) framework for image classification problems considering the Independent and Identically Distributed (IID) and Non‐Independent and Identically Distributed (Non‐IID) medical data. The FBMA architecture integrates FL principles with the Multi‐Head Attention mechanism, optimizing the model performance and ensuring privacy. Using Multi‐Head Attention, the FBMA framework allows the model to selectively focus on important regions of the image for feature extraction, and using FL, FBMA leverages decentralized medical institutions to facilitate collaborative model training while maintaining data privacy. Through rigorous experimentation on medical image datasets: MedMNIST Dataset, MedicalMNIST Dataset, and LC25000 Dataset, each partitioned into Non‐IID data distribution, the proposed FBMA framework exhibits high‐performance metrics. The results highlight the efficacy of our proposed FBMA framework, indicating its potential for real‐world applications where image classification demands both high accuracy and data privacy.

Prototype-Guided Graph Reasoning Network for Few-Shot Medical Image Segmentation.

Few-shot semantic segmentation (FSS) is of tremendous potential for data-scarce scenarios, particularly in medical segmentation tasks with merely a few labeled data. Most of the existing FSS methods typically distinguish query objects with the guidance of support prototypes. However, the variances in appearance and scale between support and query objects from the same anatomical class are often exceedingly considerable in practical clinical scenarios, thus resulting in undesirable query segmentation masks. To tackle the aforementioned challenge, we propose a novel prototype-guided graph reasoning network (PGRNet) to explicitly explore potential contextual relationships in structured query images. Specifically, a prototype-guided graph reasoning module is proposed to perform information interaction on the query graph under the guidance of support prototypes to fully exploit the structural properties of query images to overcome intra-class variances. Moreover, instead of fixed support prototypes, a dynamic prototype generation mechanism is devised to yield a collection of dynamic support prototypes by mining rich contextual information from support images to further boost the efficiency of information interaction between support and query branches. Equipped with the proposed two components, PGRNet can learn abundant contextual representations for query images and is therefore more resilient to object variations. We validate our method on three publicly available medical segmentation datasets, namely CHAOS-T2, MS-CMRSeg, and Synapse. Experiments indicate that the proposed PGRNet outperforms previous FSS methods by a considerable margin and establishes a new state-of-the-art performance.