Annotation-efficient deep learning for automatic medical image segmentation

Detection of abnormal extraocular muscles in small datasets of computed tomography images using a three-dimensional variational autoencoder

We sought to establish an unsupervised algorithm with a three-dimensional (3D) variational autoencoder model (VAE) for the detection of abnormal extraocular muscles in small datasets of orbital computed tomography (CT) images. 334 CT images of normal orbits and 96 of abnormal orbits diagnosed as thyroid eye disease were used for training and validation; 24 normal and 11 abnormal orbits were used for the test. A 3D VAE was developed and trained. All images were preprocessed to emphasize extraocular muscles and to suppress background noise (e.g., high signal intensity from bones). The optimal cut-off value was identified through receiver operating characteristic (ROC) curve analysis. The ability of the model to detect muscles of abnormal size was assessed by visualization. The model achieved a sensitivity of 79.2%, specificity of 72.7%, accuracy of 77.1%, F1-score of 0.667, and AUROC of 0.801. Abnormal CT images correctly identified by the model showed differences in the reconstruction of extraocular muscles. The proposed model showed potential to detect abnormalities in extraocular muscles using a small dataset, similar to the diagnostic approach used by physicians. Unsupervised learning could serve as an alternative detection method for medical imaging studies in which annotation is difficult or impossible to perform.

Deep-learning based quantification model for hip bone marrow edema and synovitis in patients with spondyloarthritis based on magnetic resonance images

Objectives: Hip inflammation is one of the most common complications in patients with spondyloarthritis (SpA). Herein, we employed use of a deep learning-based magnetic resonance imaging (MRI) evaluation model to identify irregular and multiple inflammatory lesions of the hip. Methods: All of the SpA patients were enrolled at the Xijing Hospital. The erythrocyte sediment rate (ESR), C-reactive protein (CRP), hip function Harris score, and disease activity were evaluated by clinicians. Manual MRI annotations including bone marrow edema (BME) and effusion/synovitis, and a hip MRI scoring system (HIMRISS) assessment was performed by experienced musculoskeletal radiologists. The segmentation accuracies of four deep learning models, including U-Net, UNet++, Attention-Unet, and HRNet, were compared using five-fold cross-validation. The clinical agreement of U-Net was evaluated with clinical symptoms and HIMRISS results. Results: A total of 1945 MRI slices of STIR/T2WI sequences were obtained from 195 SpA patients with hip involvement. After the five-fold cross-validation, U-Net achieved an average segmentation accuracy of 88.48% for the femoral head and 69.36% for inflammatory lesions, which are higher than those obtained by the other three models. The UNet-score, which was calculated based on the same MRI slices as HIMRISS, was significantly correlated with the HIMRISS scores and disease activity indexes (p values <0.05). Conclusion: This deep-learning based automatic MRI evaluation model could achieve similar quantification performance as an expert radiologist, and it has the potential to improve the accuracy and efficiency of clinical diagnosis for SpA patients with hip involvement.

Artificial Intelligence (AI) in Breast Imaging: A Scientometric Umbrella Review

Artificial intelligence (AI), a rousing advancement disrupting a wide spectrum of applications with remarkable betterment, has continued to gain momentum over the past decades. Within breast imaging, AI, especially machine learning and deep learning, honed with unlimited cross-data/case referencing, has found great utility encompassing four facets: screening and detection, diagnosis, disease monitoring, and data management as a whole. Over the years, breast cancer has been the apex of the cancer cumulative risk ranking for women across the six continents, existing in variegated forms and offering a complicated context in medical decisions. Realizing the ever-increasing demand for quality healthcare, contemporary AI has been envisioned to make great strides in clinical data management and perception, with the capability to detect indeterminate significance, predict prognostication, and correlate available data into a meaningful clinical endpoint. Here, the authors captured the review works over the past decades, focusing on AI in breast imaging, and systematized the included works into one usable document, which is termed an umbrella review. The present study aims to provide a panoramic view of how AI is poised to enhance breast imaging procedures. Evidence-based scientometric analysis was performed in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guideline, resulting in 71 included review works. This study aims to synthesize, collate, and correlate the included review works, thereby identifying the patterns, trends, quality, and types of the included works, captured by the structured search strategy. The present study is intended to serve as a "one-stop center" synthesis and provide a holistic bird's eye view to readers, ranging from newcomers to existing researchers and relevant stakeholders, on the topic of interest.

Single-shot T2 mapping via multi-echo-train multiple overlapping-echo detachment planar imaging and multitask deep learning

BACKGROUND

Quantitative magnetic resonance imaging provides robust biomarkers in clinics. Nevertheless, the lengthy scan time reduces imaging throughput and increases the susceptibility of imaging results to motion. In this context, a single-shot T₂ mapping method based on multiple overlapping-echo detachment (MOLED) planar imaging was presented, but the relatively small echo time range limits its accuracy, especially in tissues with large T₂ .

PURPOSE

In this work we proposed a novel single-shot method, Multi-Echo-Train Multiple OverLapping-Echo Detachment (METMOLED) planar imaging, to accommodate a large range of T₂ quantification without additional measurements to rectify signal degeneration arisen from refocusing pulse imperfection.

METHODS

Multiple echo-train techniques were integrated into the MOLED sequence to capture larger TE information. Maps of T₂ , B₁ , and spin density were reconstructed synchronously from acquired METMOLED data via multitask deep learning. A typical U-Net was trained with 3000/600 synthetic data with geometric/brain patterns to learn the mapping relationship between METMOLED signals and quantitative maps. The refocusing pulse imperfection was settled through the inherent information of METMOLED data and auxiliary tasks.

RESULTS

Experimental results on the digital brain (structural similarity (SSIM) index = 0.975/0.991/0.988 for MOLED/METMOLED-2/METMOLED-3, hyphenated number denotes the number of echo-trains), physical phantom (the slope of linear fitting with reference T₂ map = 1.047/1.017/1.006 for MOLED/METMOLED-2/METMOLED-3), and human brain (Pearson's correlation coefficient (PCC) = 0.9581/0.9760/0.9900 for MOLED/METMOLED-2/METMOLED-3) demonstrated that the METMOLED improved the quantitative accuracy and the tissue details in contrast to the MOLED. These improvements were more pronounced in tissues with large T₂ and in application scenarios with high temporal resolution (PCC = 0.8692/0.9465/0.9743 for MOLED/METMOLED-2/METMOLED-3). Moreover, the METMOLED could rectify the signal deviations induced by the non-ideal slice profiles of refocusing pulses without additional measurements. A preliminary measurement also demonstrated that the METMOLED is highly repeatable (mean coefficient of variation (CV) = 1.65%).

CONCLUSIONS

METMOLED breaks the restriction of echo-train length to TE and implements unbiased T₂ estimates in an extensive range. Furthermore, it corrects the effect of refocusing pulse inaccuracy without additional measurements or signal post-processing, thus retaining its single-shot characteristic. This technique would be beneficial for accurate T₂ quantification.

Artificial intelligence in multi-parametric magnetic resonance imaging: A review

Multi-parametric magnetic resonance imaging (mpMRI) is an indispensable tool in the clinical workflow for the diagnosis and treatment planning of various diseases. Machine learning-based artificial intelligence (AI) methods, especially those adopting the deep learning technique, have been extensively employed to perform mpMRI image classification, segmentation, registration, detection, reconstruction, and super-resolution. The current availability of increasing computational power and fast-improving AI algorithms have empowered numerous computer-based systems for applying mpMRI to disease diagnosis, imaging-guided radiotherapy, patient risk and overall survival time prediction, and the development of advanced quantitative imaging technology for magnetic resonance fingerprinting. However, the wide application of these developed systems in the clinic is still limited by a number of factors, including robustness, reliability, and interpretability. This survey aims to provide an overview for new researchers in the field as well as radiologists with the hope that they can understand the general concepts, main application scenarios, and remaining challenges of AI in mpMRI. This article is protected by copyright. All rights reserved.

Co-Training Method Based on Semi-Decoupling Features for MOOC Learner Behavior Prediction

Facing the problem of massive unlabeled data and limited labeled samples, semi-supervised learning is favored, especially co-training. Standard co-training requires sufficiently redundant and conditionally independent dual views; however, in fact, few dual views exist that satisfy this condition. To solve this problem, we propose a co-training method based on semi-decoupling features, that is, semi-decoupling features based on a known single view and then constructing independent and redundant dual views: (1) take a small number of important features as shared features of the dual views according to the importance of the features; (2) separate the remaining features one by one or in small batches according to the correlation between the features to make “divergent” features of the dual views; (3) combine the shared features and the “divergent” features to construct dual views. In this paper, the experimental dataset was from the edX dataset jointly released by Harvard University and MIT; the evaluation metrics adopted F1, Precision, and Recall. The analysis methods included three experiments: multiple models, iterations, and hyperparameters. The experimental results show that the effect of this model on MOOC learner behavior prediction was better than the other models, and the best prediction result was obtained in iteration 2. These all verify the effectiveness and superiority of this algorithm and provide a scientific and feasible reference for the development of the future education industry.

Development of a radiomics model to diagnose pheochromocytoma preoperatively: a multicenter study with prospective validation

Background

Preoperative diagnosis of pheochromocytoma (PHEO) accurately impacts preoperative preparation and surgical outcome in PHEO patients. Highly reliable model to diagnose PHEO is lacking. We aimed to develop a magnetic resonance imaging (MRI)-based radiomic-clinical model to distinguish PHEO from adrenal lesions.

Methods

In total, 305 patients with 309 adrenal lesions were included and divided into different sets. The least absolute shrinkage and selection operator (LASSO) regression model was used for data dimension reduction, feature selection, and radiomics signature building. In addition, a nomogram incorporating the obtained radiomics signature and selected clinical predictors was developed by using multivariable logistic regression analysis. The performance of the radiomic-clinical model was assessed with respect to its discrimination, calibration, and clinical usefulness.

Results

Seven radiomics features were selected among the 1301 features obtained as they could differentiate PHEOs from other adrenal lesions in the training (area under the curve [AUC], 0.887), internal validation (AUC, 0.880), and external validation cohorts (AUC, 0.807). Predictors contained in the individualized prediction nomogram included the radiomics signature and symptom number (symptoms include headache, palpitation, and diaphoresis). The training set yielded an AUC of 0.893 for the nomogram, which was confirmed in the internal and external validation sets with AUCs of 0.906 and 0.844, respectively. Decision curve analyses indicated the nomogram was clinically useful. In addition, 25 patients with 25 lesions were recruited for prospective validation, which yielded an AUC of 0.917 for the nomogram.

Conclusion

We propose a radiomic-based nomogram incorporating clinically useful signatures as an easy-to-use, predictive and individualized tool for PHEO diagnosis.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-022-03233-w.