Meta-Radiology最新文献

Diffusion probabilistic models (DPMs) which employ explicit likelihood characterization and a gradual sampling process to synthesize data, have gained increasing research interest. Despite their huge computational burdens due to the large number of steps involved during sampling, DPMs are widely appreciated in various medical imaging tasks for their high-quality and diversity of generation. Magnetic resonance imaging (MRI) is an important medical imaging modality with excellent soft tissue contrast and superb spatial resolution, which possesses unique opportunities for DPMs. Although there is a recent surge of studies exploring DPMs in MRI, a survey paper of DPMs specifically designed for MRI applications is still lacking. This review article aims to help researchers in the MRI community to grasp the advances of DPMs in different applications. We first introduce the theory of two dominant kinds of DPMs, categorized according to whether the diffusion time step is discrete or continuous, and then provide a comprehensive review of emerging DPMs in MRI, including reconstruction, image generation, image translation, segmentation, anomaly detection, and further research topics. Finally, we discuss the general limitations as well as limitations specific to the MRI tasks of DPMs and point out potential areas that are worth further exploration.

Influenced by ChatGPT, artificial intelligence (AI) large models have witnessed a global upsurge in large model research and development. As people enjoy the convenience by this AI large model, more and more large models in subdivided fields are gradually being proposed, especially large models in radiology imaging field. This article first introduces the development history of large models, technical details, workflow, working principles of multimodal large models and working principles of video generation large models. Secondly, we summarize the latest research progress of AI large models in radiology education, radiology report generation, applications of unimodal and multimodal radiology. Finally, this paper also summarizes some of the challenges of large AI models in radiology, with the aim of better promoting the rapid revolution in the field of radiography.

Esophagus carcinoma (EC) ranks sixth in cancer-related mortality and seventh in terms of morbidity worldwide, and radical esophagectomy is considered as the basis of comprehensive treatment for locally advanced EC. Accurate preoperative determination of lymph node status is critical for treatment decision-making, assessment of survival time and life quality of patients after surgery. However, the rate of misdiagnosis and missed diagnosis of metastatic lymph nodes by traditional imaging methods is high. With the development of artificial intelligence technology and medical image digitization, medical image analysis methods based on artificial intelligence have brought new ideas to the diagnosis and research of lymph node metastasis secondary to EC. At present, texture analysis, radiomics and deep learning are the most widely used methods. These technologies extract and analyze quantitative features from traditional medical images to provide biological information such as tumor characteristics and heterogeneity to guide clinical practice. Therefore, this review mainly introduces and discusses the current status of imaging research on lymph node metastasis in patients with EC based on texture analysis, radiomics and deep learning, and prospects the important research directions in the future with a view to improving the diagnostic capability of lymph node metastasis in patients with EC in China.

Recent breakthroughs in artificial intelligence (AI) research include advancements in natural language processing (NLP) achieved by large language models (LLMs), and; in particular, generative pre–trained transformer (GPT) architectures. The latest GPT developed by OpenAI, GPT–4, has shown remarkable intelligence across various domains and tasks. It exhibits capabilities in abstraction, comprehension, vision, computer coding, mathematics, and more, suggesting it to be a significant step towards artificial general intelligence (AGI), a level of AI that possesses capabilities similar to human intelligence. This paper explores this AGI, its knowledge diffusive and societal influences, and its governance. In addition to coverage of the major associated topics studied in the literature, and making up for their loopholes, we scrutinize how GPT-4 can facilitate the diffusion of knowledge across different areas of science by promoting their interpretability and explainability (IE) to inexperts. Where applicable, the topics are also accompanied by their specific potential implications on medical imaging.

Aim and scope

This paper aims to propose a Multi-scale Capsule-weighted Fusion Classification Network (MCFCN), a classification model for automatic diagnosis of lung lesions by CT scanning.

Background

The automatic diagnosis of lung lesions based on chest CT scans plays a crucial role in assisting doctors to identify suspicious cases quickly and accurately. However, existing methods struggle to differentiate lesions with similar morphologies, and current feature extraction techniques lack the ability to effectively highlight small-scale targets in a large-scale environment, leading to incomplete extraction of subtle features and ultimately compromising the classification performance.

Method

The MCFCN employs a dynamic routing clustering algorithm to emphasize small-scale features, preventing feature loss. Additionally, a scale difference fusion network is utilized to extract precise position scaling parameters by incorporating weighted fusion of information from different scales.

Results

MCFCN achieves an accuracy of 99.41% for COVID-19 classification, 93.33% for CAP classification, and 100% for Normal classification, with an overall accuracy of 98.36%.

Conclusion

Experimental results on the target dataset demonstrate that MCFCN outperforms state-of-the-art methods. In the future, this model can be further explored and optimized to enhance its application value in clinical practice

Soft tissue sarcomas (STSs) represent a group of heterogeneous mesenchymal tumors of which are generally classified as per the histopathology. Despite being rare in incidence and prevalence, STSs are usually correlated with unfavorable prognosis and high mortality rate. Early and accurate diagnosis of STSs are critical in clinical management of STSs. Deep learning (DL) refers to a subtype of artificial intelligence that has been adopted to assist healthcare professionals to optimize personalized treatment for a given situation, particularly in image analysis. Recently, emerging studies have demonstrated that application of DL based on medical images could substantially improve the accuracy and efficiency of clinicians to the identification, diagnosis, treatment, and prognosis prediction of STSs, and thereby facilitating the clinical decision-making. Herein, we aimed to extensively summarize the recent applications of DL-based artificial intelligence in STSs from the aspects of data acquisition, algorithm, and model establishment. Besides, the reinforcement of the model by transfer learning and generative adversarial network (GAN) for data augmentation has also been elaborated. It is worth noting that high-quality data with accurate annotations, as well as optimized algorithmic performance are pivotal in the clinical application of DL in STSs.

Neurodegenerative diseases refer to degenerative diseases of the nervous system caused by neuronal degeneration and apoptosis. Usually, the onset of the disease is insidious, and the progression is slow, which can last for several years to decades. Clinical symptoms only appear in the later stages of pathological changes when the degree of nerve cell loss reaches or exceeds a certain threshold. Traditional electrophysiological and medical imaging techniques lack valuable indicators and markers. Therefore, early diagnosis and differentiation are very difficult. Radiomics is a new medical imaging technology merged in recent years, which can extract a large number of invisible features from raw image data with high throughput, and quantitatively analyze the pathological and physiological changes. It demonstrates important potential value in the diagnosis, grading, and prognosis evaluation of NDs. This review provides an overview of the research progress of radiomics in neurodegenerative diseases, emphasizing the process principles of radiomics and its application in the diagnosis, classification, and prediction of these diseases. This helps to deepen the understanding of neurodegenerative diseases and promote early diagnosis and treatment in clinical practice.

Magnetic resonance-guided focused ultrasound (MRgFUS) is a non-invasive technique for neuroregulation that offers several advantages, including non-invasiveness, no need for general anesthesia requirement, real-time target localization, and real-time temperature monitoring. Currently, the U.S. Food and Drug Administration has approved this technology for the treatment of essential tremor and Parkinson's disease, and its indications are continually expanding to encompass various intracranial diseases. In this article, we summarize clinical trials of high-intensity FUS in the treatment of intracranial diseases. Next, we introduce the preclinical and clinical studies on low-intensity FUS-induced blood-brain barrier opening and neuromodulation. Finally, we discuss the challenges and future directions of this technology. This review aims to guide future clinical trials and provide new perspectives for investigating the neural mechanisms of MRgFUS.

The epidemiological of lung cancer surgery patients is changing, resulting in changes in diagnosis and treatment. Changing the way and content of imaging reports in response to the above changes is necessary. This paper aims to review the problems involved in lung nodule screening, diagnosis, and treatment stages from the radiologist's perspective and suggest feasible solutions.

Aim

To assess the utility of different radiomics feature selection methods in predicting transarterial chemoembolization (TACE) response in hepatocellular carcinoma (HCC) patients.

Materials and methods

This study employed a dataset of 136 paired MR T1-weighted contrast-enhanced abdominal images with liver tumor masks before and after TACE. TACE response for each image pair was classified by European Association for the Study of the Liver (EASL) and modified Response Evaluation Criteria in Solid Tumors (mRECIST) guidelines. 100D feature vectors were generated for the paired tumor areas. Eighteen existing feature selection methods were employed to select the top-k features to train and test a non-linear support vector machine (SVM) with a Gaussian kernel. Five-cross validation was performed to identify the highest performing feature selection methods.

Results

For all benchmarks, a L0-based method selecting the top-5 or top-10 features achieved the highest performance. For images classified with EASL criteria that were analyzed with the L0-based method, the accuracy (ACC), area under curve (AUC), and balanced F score (F1-score) were 0.75 ​± ​0.06, 0.75 ​± ​0.09, and 0.80 ​± ​0.05, respectively. For images classified with mRECIST criteria that were analyzed with the L0-based method, the ACC, AUC, and F1-score were 0.75 ​± ​0.07, 0.71 ​± ​0.16, and 0.82 ​± ​0.04, respectively.

Conclusion

A L0-based method that selected the top-5/10 most important features predicted TACE response in HCC patients with the highest accuracy under both EASL and mRECIST criteria. This proof-of-concept investigation represents a step forward in the development of a reliable clinical decision-making tool for management of intermediate HCC patients undergoing TACE.