Search

In recent years, remarkable progress in zero-shot learning (ZSL) has been achieved by generative adversarial networks (GAN). To compensate for the lack of training samples in ZSL, a surge of GAN architectures have been developed by human experts through trial-and-error testing. Despite their efficacy, however, there is still no guarantee that these hand-crafted models can consistently achieve good performance across diversified datasets or scenarios. Accordingly, in this paper, we turn to neural architecture search (NAS) and make the first attempt to bring NAS techniques into the ZSL realm. Specifically, we propose a differentiable GAN architecture search method over a specifically designed search space for zero-shot learning, referred to as ZeroNAS. Considering the relevance and balance of the generator and discriminator, ZeroNAS jointly searches their architectures in a min-max player game via adversarial training. Extensive experiments conducted on four widely used benchmark datasets demonstrate that ZeroNAS is capable of discovering desirable architectures that perform favorably against state-of-the-art ZSL and generalized zero-shot learning (GZSL) approaches. Source code is at https://github.com/caixiay/ZeroNAS.

ImportanceSelection of antiseizure medications (ASMs) for epilepsy remains largely a trial-and-error approach. Under this approach, many patients have to endure sequential trials of ineffective treatments until the “right drugs” are prescribed.ObjectiveTo develop and validate a deep learning model using readily available clinical information to predict treatment success with the first ASM for individual patients.Design, Setting, and ParticipantsThis cohort study developed and validated a prognostic model. Patients were treated between 1982 and 2020. All patients were followed up for a minimum of 1 year or until failure of the first ASM. A total of 2404 adults with epilepsy newly treated at specialist clinics in Scotland, Malaysia, Australia, and China between 1982 and 2020 were considered for inclusion, of whom 606 (25.2%) were excluded from the final cohort because of missing information in 1 or more variables.ExposuresOne of 7 antiseizure medications.Main Outcomes and MeasuresWith the use of the transformer model architecture on 16 clinical factors and ASM information, this cohort study first pooled all cohorts for model training and testing. The model was trained again using the largest cohort and externally validated on the other 4 cohorts. The area under the receiver operating characteristic curve (AUROC), weighted balanced accuracy, sensitivity, and specificity of the model were all assessed for predicting treatment success based on the optimal probability cutoff. Treatment success was defined as complete seizure freedom for the first year of treatment while taking the first ASM. Performance of the transformer model was compared with other machine learning models.ResultsThe final pooled cohort included 1798 adults (54.5% female; median age, 34 years [IQR, 24-50 years]). The transformer model that was trained using the pooled cohort had an AUROC of 0.65 (95% CI, 0.63-0.67) and a weighted balanced accuracy of 0.62 (95% CI, 0.60-0.64) on the test set. The model that was trained using the largest cohort only had AUROCs ranging from 0.52 to 0.60 and a weighted balanced accuracy ranging from 0.51 to 0.62 in the external validation cohorts. Number of pretreatment seizures, presence of psychiatric disorders, electroencephalography, and brain imaging findings were the most important clinical variables for predicted outcomes in both models. The transformer model that was developed using the pooled cohort outperformed 2 of the 5 other models tested in terms of AUROC.Conclusions and RelevanceIn this cohort study, a deep learning model showed the feasibility of personalized prediction of response to ASMs based on clinical information. With improvement of performance, such as by incorporating genetic and imaging data, this model may potentially assist clinicians in selecting the right drug at the first trial.

Background It is unclear whether brain volumes are causally affected by Intraocular pressure (IOP) is highly correlated with blood pressure (BP).Methods The study included 8634 participants for IOP and 36069 participants for BP in observational analyses and 37410 participants for both IOP and BP in Mendelian Randomisation (MR) analyses from UK Biobank. IOP and BP were measured between 2006–2010. Brain volumes were measured using MRI between 2014–2019.Results Higher IOP was associated with smaller volumes of total brain (β (95% CI) for each 5-mmHg increment: -3.24 (-5.05, -1.44) ml) and grey matter (-1.10 (-2.17, -0.03) ml) independent of BP. Diastolic BP (β (95% CI) for each 10-mmHg increment: 0.13 (0.05, 0.21)) was associated with higher white matter hyperintensity (WMH) independent of antihypertensive medications. Associations between IOP and total brain and WMH volumes were stronger in younger individuals or those without hypertension. Associations between DBP/SBP and brain volumes were stronger in younger individuals, women, and lowly educated individuals. All MR analytic methods demonstrated a significant relationship between DBP and WMH (β (95% CI) for each 10-mmHg increment of DBP for inverse-variance weighting method: 0.019 (0.013, 0.026)). The β (95% CI) for grey matter volume (ml) associated with each 5-mmHg increment of IOP for inverse-variance weighting method was − 3.42 (-5.39, -1.45).Conclusions Higher IOP is casually linked to larger grey matter volume reduction while increased DBP casually linked to higher WMH load. Younger or lowly educated individuals deserve more scrutiny for the prevention of brain volume reduction potentially via IOP/DBP lowering.

Dermatologists often diagnose or rule out early melanoma by evaluating the follow-up dermoscopic images of skin lesions. However, existing algorithms for early melanoma diagnosis are developed using single time-point images of lesions. Ignoring the temporal, morphological changes of lesions can lead to misdiagnosis in borderline cases. In this study, we propose a framework for automated early melanoma diagnosis using sequential dermoscopic images. To this end, we construct our method in three steps. First, we align sequential dermoscopic images of skin lesions using estimated Euclidean transformations, extract the lesion growth region by computing image differences among the consecutive images, and then propose a spatio-temporal network to capture the dermoscopic changes from aligned lesion images and the corresponding difference images. Finally, we develop an early diagnosis module to compute probability scores of malignancy for lesion images over time. We collected 179 serial dermoscopic imaging data from 122 patients to verify our method. Extensive experiments show that the proposed model outperforms other commonly used sequence models. We also compared the diagnostic results of our model with those of seven experienced dermatologists and five registrars. Our model achieved higher diagnostic accuracy than clinicians (63.69% vs. 54.33%, respectively) and provided an earlier diagnosis of melanoma (60.7% vs. 32.7% of melanoma correctly diagnosed on the first follow-up images). These results demonstrate that our model can be used to identify melanocytic lesions that are at high-risk of malignant transformation earlier in the disease process and thereby redefine what is possible in the early detection of melanoma.

Prevailing deep methods for image recognition require massive labeled samples in each visual category for training. However, large amounts of data annotations are time-consuming, and some uncommon categories only have rare samples available. For this issue, we focus on more challenging few-shot learning (FSL) task, where just few labeled images are used in the training stage. Existing FSL models are constructed with various convolutional neural networks (CNNs), which are trained on an auxiliary base dataset and evaluated for new few-shot predictions on a novel dataset. The performance of these models is difficult to break through because of the domain shift between base and novel datasets and the monotonous network architectures. Considering that, we propose a novel automatic attribute consistent network called Auto-ACNet to overcome the above problems. On one hand, Auto-ACNet utilizes the attribute information about base and novel categories to guide the procedure of representation learning. It introduces the consistent and non-consistent subnets to capture the common and different attributes of image pair, which helps to mitigate the domain shift problem. On the other hand, the architecture of Auto-ACNet is searched with the popular neural architecture search (NAS) technique DARTS, for obtaining a superior FSL network automatically. And the DARTS’s search space is improved by adding the position-aware module to extract the attribute characteristics better. Extensive experimental results on two datasets indicate that the proposed Auto-ACNet achieves significant improvement over the state-of-the-art competitors in this literature.

Background: Little is known about the association between macronutrient intake and incident dementia. Objective: To identify an optimal range of macronutrient intake associated with reduced risk of dementia. Methods: Our analysis included 93,389 adults aged 60–75 years from the UK Biobank. Diet was assessed using a web-based 24-h recall questionnaire between 2009–2012. Dementia was ascertained using hospital inpatient, death records, and self-reported data up to January 2021. We calculated a macronutrient score based on associations between an individual’s macronutrient intake and incident dementia. Results: During a median follow-up of 8.7 years, 1,171 incident dementia cases were documented. We found U-shape relationships for carbohydrate, fat, and protein intake with incident dementia. Compared to individuals with optimal carbohydrate intake, those with high intake (HR (95%CI): 1.48(1.15–1.91)) but not low intake (1.19(0.89–1.57)) had a higher risk of dementia. In the multivariable analysis, a low-fat intake (HR (95%CI): 1.42(1.11–1.82)) was associated with a higher risk of all-cause dementia. After adjustment for covariates, a high (HR (95%CI): 1.41(1.09–1.83)) but not low protein intake (1.22(0.94–1.57)) was associated with an increased risk of dementia. Individuals in quintiles 3–5 of optimal macronutrient score had a lower risk of dementia compared with those in quintile 1 (HR (95%CI): 0.76(0.64–0.91) for quintile 3, 0.71(0.60–0.85) for quintile 4, 0.74(0.61–0.91) for quintile 5). The association between macronutrient score and incident dementia was significant across subgroups of age, gender, education, and smoking. Conclusion: Moderate intakes of carbohydrate, fat, and protein were associated with the lowest risk of incident dementia.

Background HIV and sexually transmitted infections (STIs) are major global public health concerns. Over 1 million curable STIs occur every day among people aged 15 years to 49 years worldwide. Insufficient testing or screening substantially impedes the elimination of HIV and STI transmission. Objective The aim of our study was to develop an HIV and STI risk prediction tool using machine learning algorithms. Methods We used clinic consultations that tested for HIV and STIs at the Melbourne Sexual Health Centre between March 2, 2015, and December 31, 2018, as the development data set (training and testing data set). We also used 2 external validation data sets, including data from 2019 as external “validation data 1” and data from January 2020 and January 2021 as external “validation data 2.” We developed 34 machine learning models to assess the risk of acquiring HIV, syphilis, gonorrhea, and chlamydia. We created an online tool to generate an individual’s risk of HIV or an STI. Results The important predictors for HIV and STI risk were gender, age, men who reported having sex with men, number of casual sexual partners, and condom use. Our machine learning–based risk prediction tool, named MySTIRisk, performed at an acceptable or excellent level on testing data sets (area under the curve [AUC] for HIV=0.78; AUC for syphilis=0.84; AUC for gonorrhea=0.78; AUC for chlamydia=0.70) and had stable performance on both external validation data from 2019 (AUC for HIV=0.79; AUC for syphilis=0.85; AUC for gonorrhea=0.81; AUC for chlamydia=0.69) and data from 2020-2021 (AUC for HIV=0.71; AUC for syphilis=0.84; AUC for gonorrhea=0.79; AUC for chlamydia=0.69). Conclusions Our web-based risk prediction tool could accurately predict the risk of HIV and STIs for clinic attendees using simple self-reported questions. MySTIRisk could serve as an HIV and STI screening tool on clinic websites or digital health platforms to encourage individuals at risk of HIV or an STI to be tested or start HIV pre-exposure prophylaxis. The public can use this tool to assess their risk and then decide if they would attend a clinic for testing. Clinicians or public health workers can use this tool to identify high-risk individuals for further interventions.

The need for comprehensive and automated screening methods for retinal image classification has long been recognized. Well-qualified doctors annotated images are very expensive and only a limited amount of data is available for various retinal diseases such as diabetic retinopathy (DR) and age-related macular degeneration (AMD). Some studies show that some retinal diseases such as DR and AMD share some common features like haemorrhages and exudation but most classification algorithms only train those disease models independently when the only single label for one image is available. Inspired by multi-task learning where additional monitoring signals from various sources is beneficial to train a robust model. We propose a method called synergic adversarial label learning (SALL) which leverages relevant retinal disease labels in both semantic and feature space as additional signals and train the model in a collaborative manner using knowledge distillation. Our experiments on DR and AMD fundus image classification task demonstrate that the proposed method can significantly improve the accuracy of the model for grading diseases by 5.91% and 3.69% respectively. In addition, we conduct additional experiments to show the effectiveness of SALL from the aspects of reliability and interpretability in the context of medical imaging application.

Background the Cardiovascular Risk Factors, Aging, and Incidence of Dementia (CAIDE) dementia risk score is a recognised tool for dementia risk stratification. However, its application is limited due to the requirements for multidimensional information and fasting blood draw. Consequently, an effective and non-invasive tool for screening individuals with high dementia risk in large population-based settings is urgently needed. Methods a deep learning algorithm based on fundus photographs for estimating the CAIDE dementia risk score was developed and internally validated by a medical check-up dataset included 271,864 participants in 19 province-level administrative regions of China, and externally validated based on an independent dataset included 20,690 check-up participants in Beijing. The performance for identifying individuals with high dementia risk (CAIDE dementia risk score ≥ 10 points) was evaluated by area under the receiver operating curve (AUC) with 95% confidence interval (CI). Results the algorithm achieved an AUC of 0.944 (95% CI: 0.939–0.950) in the internal validation group and 0.926 (95% CI: 0.913–0.939) in the external group, respectively. Besides, the estimated CAIDE dementia risk score derived from the algorithm was significantly associated with both comprehensive cognitive function and specific cognitive domains. Conclusions this algorithm trained via fundus photographs could well identify individuals with high dementia risk in a population setting. Therefore, it has the potential to be utilised as a non-invasive and more expedient method for dementia risk stratification. It might also be adopted in dementia clinical trials, incorporated as inclusion criteria to efficiently select eligible participants.

Retinopathy of prematurity (ROP) is currently one of the leading causes ofinfant blindness worldwide. In recent years, significant progress has been made incomputer-aided diagnostic methods based on deep learning. However, deeplearning is often data hungry, with its need for large amounts of annotated datafor model optimization, and in clinical scenarios, annotated data requires longhours of effort by experienced physicians. In contrast, a large number ofunlabelled images are relatively easy to obtain. In this paper, we propose a newsemi-supervised learning framework to reduce annotation costs by using partiallylabelled images and a large number of unlabelled images for automatic ROPstaging. Our semi-supervised deep learning framework consists of a studentmodel and a teacher model. We input images that have undergone differentperturbations into the student and teacher models. By encouraging consistencyof predicted output and inter-sample semantic structure between the student andteacher models, useful discriminative information is mined from the unlabelleddata to improve model classification performance. The teacher model parametersare updated by an exponential moving average algorithm. We have conductedextensive experiments on a real clinical dataset. Using only 30% of the labelledimages, our method achieves area under the receiver operating curve (AUC) of0.8630, Accuracy of 0.7895, Sensitivity of 0.7868, and Specificity of 0.8010. Theclassification performance of our method is close to that of using all 100%labelled images when the proportion of labelled images is 50%.

Usually, lesions are not isolated but are associated with the surrounding tissues. For example, the growth of a tumour can depend on or infiltrate into the surrounding tissues. Due to the pathological nature of the lesions, it is challenging to distinguish their boundaries in medical imaging. However, these uncertain regions may contain diagnostic information. Therefore, the simple binarization of lesions by traditional binary segmentation can result in the loss of diagnostic information. In this work, we introduce the image matting into the 3D scenes and use the alpha matte, i.e., a soft mask, to describe lesions in a 3D medical image. The traditional soft mask acted as a training trick to compensate for the easily mislabelled or under-labelled ambiguous regions. In contrast, 3D matting uses soft segmentation to characterize the uncertain regions more finely, which means that it retains more structural information for subsequent diagnosis and treatment. The current study of image matting methods in 3D is limited. To address this issue, we conduct a comprehensive study of 3D matting, including both traditional and deep-learning-based methods. We adapt four state-of-the-art 2D image matting algorithms to 3D scenes and further customize the methods for CT images to calibrate the alpha matte with the radiodensity. Moreover, we propose the first end-to-end deep 3D matting network and implement a solid 3D medical image matting benchmark. Its efficient counterparts are also proposed to achieve a good performance-computation balance. Furthermore, there is no high-quality annotated dataset related to 3D matting, slowing down the development of data-driven deep-learning-based methods. To address this issue, we construct the first 3D medical matting dataset. The validity of the dataset was verified through clinicians' assessments and downstream experiments., Accepted by Computers in Biology and Medicine. arXiv admin note: substantial text overlap with arXiv:2209.07843

AimsTo predict the vault and the EVO-implantable collamer lens (ICL) size by artificial intelligence (AI) and big data analytics.MethodsSix thousand two hundred and ninety-seven eyes implanted with an ICL from 3536 patients were included. The vault values were measured by the anterior segment analyzer (Pentacam HR). Permutation importance and Impurity-based feature importance are used to investigate the importance between the vault and input parameters. Regression models and classification models are applied to predict the vault. The ICL size is set as the target of the prediction, and the vault and the other input features are set as the new inputs for the ICL size prediction. Data were collected from 2015 to 2020. Random Forest, Gradient Boosting and XGBoost were demonstrated satisfying accuracy and mean area under the curve (AUC) scores in vault predicting and ICL sizing.ResultsIn the prediction of the vault, the Random Forest has the best results in the regression model (R2=0.315), then follows the Gradient Boosting (R2=0.291) and XGBoost (R2=0.285). The maximum classification accuracy is 0.828 in Random Forest, and the mean AUC is 0.765. The Random Forest predicts the ICL size with an accuracy of 82.2% and the Gradient Boosting and XGBoost, which are also compatible with 81.5% and 81.8% accuracy, respectively.ConclusionsRandom Forest, Gradient Boosting and XGBoost models are applicable for vault predicting and ICL sizing. AI may assist ophthalmologists in improving ICL surgery safety, designing surgical strategies, and predicting clinical outcomes.

Deep-learning-based segmentation methods have shown great success across many medical image applications. However, the custom training paradigms suffer from a well-known constraint of the requirement of pixel-wise annotations, which is labor-intensive, especially when they are required to learn new classes incrementally. Contemporary incremental learning focuses on dealing with catastrophic forgetting in image classification and object detection. However, this work aims to promote the performance of the current model to learn new classes with the help of the previous model in the context of incremental learning of instance segmentation. It enormously benefits the current model when the labeled data is limited because of the high labor intensity of manual labeling. In this paper, on the Diabetic Retinopathy (DR) lesion segmentation problem, a novel incremental segmentation paradigm is proposed to distill the knowledge of the previous model to improve the current model. Remarkably, we propose various approaches working on the class-based alignment of the probability maps of the current and the previous model, accounting for the difference between the background classes of the two models. The experimental evaluation of DR lesion segmentation shows the effectiveness of the proposed approaches.