A deep learning model for early risk prediction of heart failure with preserved ejection fraction by DNA methylation profiles combined with clinical features.

BACKGROUND: Heart failure with preserved ejection fraction (HFpEF), affected collectively by genetic and environmental factors, is the common subtype of chronic heart failure. Although the available risk assessment methods for HFpEF have achieved some progress, they were based on clinical or genetic features alone. Here, we have developed a deep learning framework, HFmeRisk, using both 5 clinical features and 25 DNA methylation loci to predict the early risk of HFpEF in the Framingham Heart Study Cohort. RESULTS: The framework incorporates Least Absolute Shrinkage and Selection Operator and Extreme Gradient Boosting-based feature selection, as well as a Factorization-Machine based neural network-based recommender system. Model discrimination and calibration were assessed using the AUC and Hosmer-Lemeshow test. HFmeRisk, including 25 CpGs and 5 clinical features, have achieved the AUC of 0.90 (95% confidence interval 0.88-0.92) and Hosmer-Lemeshow statistic was 6.17 (P = 0.632), which outperformed models with clinical characteristics or DNA methylation levels alone, published chronic heart failure risk prediction models and other benchmark machine learning models. Out of them, the DNA methylation levels of two CpGs were significantly correlated with the paired transcriptome levels (R < -0.3, P < 0.05). Besides, DNA methylation locus in HFmeRisk were associated with intercellular signaling and interaction, amino acid metabolism, transport and activation and the clinical variables were all related with the mechanism of occurrence of HFpEF. Together, these findings give new evidence into the HFmeRisk model. CONCLUSION: Our study proposes an early risk assessment framework for HFpEF integrating both clinical and epigenetic features, providing a promising path for clinical decision making.

Deep Learning for the Detection of Breast Cancers on Chest Computed Tomography.

BACKGROUND: Incidental breast cancers can be detected on chest computed tomography (CT) scans. With the use of deep learning, the sensitivity of incidental breast cancer detection on chest CT would improve. This study aimed to evaluate the performance of a deep learning algorithm to detect breast cancers on chest CT and to validate the results in the internal and external datasets. PATIENTS AND METHODS: This retrospective study collected 1170 preoperative chest CT scans after the diagnosis of breast cancer for algorithm development (n = 1070), internal test (n = 100), and external test (n = 100). A deep learning algorithm based on RetinaNet was developed and tested to detect breast cancer on chest CT. RESULTS: In the internal test set, the algorithm detected 96.5% of breast cancers with 13.5 false positives per case (FPs/case). In the external test set, the algorithm detected 96.1% of breast cancers with 15.6 FPs/case. When the candidate probability of 0.3 was used as the cutoff value, the sensitivities were 92.0% with 7.36 FPs/case for the internal test set and 93.0% with 8.85 FPs/case for the external test set. When the candidate probability of 0.4 was used as the cutoff value, the sensitivities were 88.5% with 5.24 FPs/case in the internal test set and 90.7% with 6.3 FPs/case in the external test set. CONCLUSION: The deep learning algorithm could sensitively detect breast cancer on chest CT in both the internal and external test sets.

Deep Learning Predicts Interval and Screening-detected Cancer from Screening Mammograms: A Case-Case-Control Study in 6369 Women.

Background The ability of deep learning (DL) models to classify women as at risk for either screening mammography-detected or interval cancer (not detected at mammography) has not yet been explored in the literature. Purpose To examine the ability of DL models to estimate the risk of interval and screening-detected breast cancers with and without clinical risk factors. Materials and Methods This study was performed on 25 096 digital screening mammograms obtained from January 2006 to December 2013. The mammograms were obtained in 6369 women without breast cancer, 1609 of whom developed screening-detected breast cancer and 351 of whom developed interval invasive breast cancer. A DL model was trained on the negative mammograms to classify women into those who did not develop cancer and those who developed screening-detected cancer or interval invasive cancer. Model effectiveness was evaluated as a matched concordance statistic (C statistic) in a held-out 26% (1669 of 6369) test set of the mammograms. Results The C statistics and odds ratios for comparing patients with screening-detected cancer versus matched controls were 0.66 (95% CI: 0.63, 0.69) and 1.25 (95% CI: 1.17, 1.33), respectively, for the DL model, 0.62 (95% CI: 0.59, 0.65) and 2.14 (95% CI: 1.32, 3.45) for the clinical risk factors with the Breast Imaging Reporting and Data System (BI-RADS) density model, and 0.66 (95% CI: 0.63, 0.69) and 1.21 (95% CI: 1.13, 1.30) for the combined DL and clinical risk factors model. For comparing patients with interval cancer versus controls, the C statistics and odds ratios were 0.64 (95% CI: 0.58, 0.71) and 1.26 (95% CI: 1.10, 1.45), respectively, for the DL model, 0.71 (95% CI: 0.65, 0.77) and 7.25 (95% CI: 2.94, 17.9) for the risk factors with BI-RADS density (b rated vs non-b rated) model, and 0.72 (95% CI: 0.66, 0.78) and 1.10 (95% CI: 0.94, 1.29) for the combined DL and clinical risk factors model. The P values between the DL, BI-RADS, and combined model's ability to detect screen and interval cancer were .99, .002, and .03, respectively. Conclusion The deep learning model outperformed in determining screening-detected cancer risk but underperformed for interval cancer risk when compared with clinical risk factors including breast density. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Bae and Kim in this issue.

A hybrid CNN-LSTM model for pre-miRNA classification.

miRNAs (or microRNAs) are small, endogenous, and noncoding RNAs construct of about 22 nucleotides. Cumulative evidence from biological experiments shows that miRNAs play a fundamental and important role in various biological processes. Therefore, the classification of miRNA is a critical problem in computational biology. Due to the short length of mature miRNAs, many researchers are working on precursor miRNAs (pre-miRNAs) with longer sequences and more structural features. Pre-miRNAs can be divided into two groups as mirtrons and canonical miRNAs in terms of biogenesis differences. Compared to mirtrons, canonical miRNAs are more conserved and easier to be identified. Many existing pre-miRNA classification methods rely on manual feature extraction. Moreover, these methods focus on either sequential structure or spatial structure of pre-miRNAs. To overcome the limitations of previous models, we propose a nucleotide-level hybrid deep learning method based on a CNN and LSTM network together. The prediction resulted in 0.943 (%95 CI ± 0.014) accuracy, 0.935 (%95 CI ± 0.016) sensitivity, 0.948 (%95 CI ± 0.029) specificity, 0.925 (%95 CI ± 0.016) F1 Score and 0.880 (%95 CI ± 0.028) Matthews Correlation Coefficient. When compared to the closest results, our proposed method revealed the best results for Acc., F1 Score, MCC. These were 2.51%, 1.00%, and 2.43% higher than the closest ones, respectively. The mean of sensitivity ranked first like Linear Discriminant Analysis. The results indicate that the hybrid CNN and LSTM networks can be employed to achieve better performance for pre-miRNA classification. In future work, we study on investigation of new classification models that deliver better performance in terms of all the evaluation criteria.

Turning Back the Clock: Artificial Intelligence Recognition of Age Reduction after Face-Lift Surgery Correlates with Patient Satisfaction.

BACKGROUND: Patients desire face-lifting procedures primarily to appear younger, more refreshed, and attractive. Because there are few objective studies assessing the success of face-lift surgery, the authors used artificial intelligence, in the form of convolutional neural network algorithms alongside FACE-Q patient-reported outcomes, to evaluate perceived age reduction and patient satisfaction following face-lift surgery. METHODS: Standardized preoperative and postoperative (1 year) images of 50 consecutive patients who underwent face-lift procedures (platysmaplasty, superficial musculoaponeurotic system-ectomy, cheek minimal access cranial suspension malar lift, or fat grafting) were used by four neural networks (trained to identify age based on facial features) to estimate age reduction after surgery. In addition, FACE-Q surveys were used to measure patient-reported facial aesthetic outcome. Patient satisfaction was compared to age reduction. RESULTS: The neural network preoperative age accuracy score demonstrated that all four neural networks were accurate in identifying ages (mean score, 100.8). Patient self-appraisal age reduction reported a greater age reduction than neural network age reduction after a face lift (-6.7 years versus -4.3 years). FACE-Q scores demonstrated a high level of patient satisfaction for facial appearance (75.1 ± 8.1), quality of life (82.4 ± 8.3), and satisfaction with outcome (79.0 ± 6.3). Finally, there was a positive correlation between neural network age reduction and patient satisfaction. CONCLUSION: Artificial intelligence algorithms can reliably estimate the reduction in apparent age after face-lift surgery; this estimated age reduction correlates with patient satisfaction. CLINICAL QUESTION/LEVEL OF EVIDENCE: Diagnostic, IV.

Using deep learning to identify recent positive selection in malaria parasite sequence data.

BACKGROUND: Malaria, caused by Plasmodium parasites, is a major global public health problem. To assist an understanding of malaria pathogenesis, including drug resistance, there is a need for the timely detection of underlying genetic mutations and their spread. With the increasing use of whole-genome sequencing (WGS) of Plasmodium DNA, the potential of deep learning models to detect loci under recent positive selection, historically signals of drug resistance, was evaluated. METHODS: A deep learning-based approach (called "DeepSweep") was developed, which can be trained on haplotypic images from genetic regions with known sweeps, to identify loci under positive selection. DeepSweep software is available from https://github.com/WDee/Deepsweep . RESULTS: Using simulated genomic data, DeepSweep could detect recent sweeps with high predictive accuracy (areas under ROC curve > 0.95). DeepSweep was applied to Plasmodium falciparum (n = 1125; genome size 23 Mbp) and Plasmodium vivax (n = 368; genome size 29 Mbp) WGS data, and the genes identified overlapped with two established extended haplotype homozygosity methods (within-population iHS, across-population Rsb) (~ 60-75% overlap of hits at P < 0.0001). DeepSweep hits included regions proximal to known drug resistance loci for both P. falciparum (e.g. pfcrt, pfdhps and pfmdr1) and P. vivax (e.g. pvmrp1). CONCLUSION: The deep learning approach can detect positive selection signatures in malaria parasite WGS data. Further, as the approach is generalizable, it may be trained to detect other types of selection. With the ability to rapidly generate WGS data at low cost, machine learning approaches (e.g. DeepSweep) have the potential to assist parasite genome-based surveillance and inform malaria control decision-making.

Deep learning-based identification of acute ischemic core and deficit from non-contrast CT and CTA.

The accurate identification of irreversible infarction and salvageable tissue is important in planning the treatments for acute ischemic stroke (AIS) patients. Computed tomographic perfusion (CTP) can be used to evaluate the ischemic core and deficit, covering most of the territories of anterior circulation, but many community hospitals and primary stroke centers do not have the capability to perform CTP scan in emergency situation. This study aimed to identify AIS lesions from widely available non-contrast computed tomography (NCCT) and CT angiography (CTA) using deep learning. A total of 345AIS patients from our emergency department were included. A multi-scale 3D convolutional neural network (CNN) was used as the predictive model with inputs of NCCT, CTA, and CTA+ (8 s delay after CTA) images. An external cohort with 108 patients was included to further validate the generalization performance of the proposed model. Strong correlations with CTP-RAPID segmentations (r = 0.84 for core, r = 0.83 for deficit) were observed when NCCT, CTA, and CTA+ images were all used in the model. The diagnostic decisions according to DEFUSE3 showed high accuracy when using NCCT, CTA, and CTA+ (0.90±0.04), followed by the combination of NCCT and CTA (0.87±0.04), CTA-alone (0.76±0.06), and NCCT-alone (0.53±0.09).

Inter-vendor performance of deep learning in segmenting acute ischemic lesions on diffusion-weighted imaging: a multicenter study.

There is little evidence on the applicability of deep learning (DL) in the segmentation of acute ischemic lesions on diffusion-weighted imaging (DWI) between magnetic resonance imaging (MRI) scanners of different manufacturers. We retrospectively included DWI data of patients with acute ischemic lesions from six centers. Dataset A (n = 2986) and B (n = 3951) included data from Siemens and GE MRI scanners, respectively. The datasets were split into the training (80%), validation (10%), and internal test (10%) sets, and six neuroradiologists created ground-truth masks. Models A and B were the proposed neural networks trained on datasets A and B. The models subsequently fine-tuned across the datasets using their validation data. Another radiologist performed the segmentation on the test sets for comparisons. The median Dice scores of models A and B were 0.858 and 0.857 for the internal tests, which were non-inferior to the radiologist's performance, but demonstrated lower performance than the radiologist on the external tests. Fine-tuned models A and B achieved median Dice scores of 0.832 and 0.846, which were non-inferior to the radiologist's performance on the external tests. The present work shows that the inter-vendor operability of deep learning for the segmentation of ischemic lesions on DWI might be enhanced via transfer learning; thereby, their clinical applicability and generalizability could be improved.

CEM500K, a large-scale heterogeneous unlabeled cellular electron microscopy image dataset for deep learning.

Automated segmentation of cellular electron microscopy (EM) datasets remains a challenge. Supervised deep learning (DL) methods that rely on region-of-interest (ROI) annotations yield models that fail to generalize to unrelated datasets. Newer unsupervised DL algorithms require relevant pre-training images, however, pre-training on currently available EM datasets is computationally expensive and shows little value for unseen biological contexts, as these datasets are large and homogeneous. To address this issue, we present CEM500K, a nimble 25 GB dataset of 0.5 × 106 unique 2D cellular EM images curated from nearly 600 three-dimensional (3D) and 10,000 two-dimensional (2D) images from >100 unrelated imaging projects. We show that models pre-trained on CEM500K learn features that are biologically relevant and resilient to meaningful image augmentations. Critically, we evaluate transfer learning from these pre-trained models on six publicly available and one newly derived benchmark segmentation task and report state-of-the-art results on each. We release the CEM500K dataset, pre-trained models and curation pipeline for model building and further expansion by the EM community. Data and code are available at https://www.ebi.ac.uk/pdbe/emdb/empiar/entry/10592/ and https://git.io/JLLTz.

Evaluation of a deep learning-based computer-aided detection algorithm on chest radiographs: Case-control study.

ABSTRACT: Along with recent developments in deep learning techniques, computer-aided diagnosis (CAD) has been growing rapidly in the medical imaging field. In this work, we evaluate the deep learning-based CAD algorithm (DCAD) for detecting and localizing 3 major thoracic abnormalities visible on chest radiographs (CR) and to compare the performance of physicians with and without the assistance of the algorithm. A subset of 244 subjects (60% abnormal CRs) was evaluated. Abnormal findings included mass/nodules (55%), consolidation (21%), and pneumothorax (24%). Observer performance tests were conducted to assess whether the performance of physicians could be enhanced with the algorithm. The area under the receiver operating characteristic (ROC) curve (AUC) and the area under the jackknife alternative free-response ROC (JAFROC) were measured to evaluate the performance of the algorithm and physicians in image classification and lesion detection, respectively. The AUCs for nodule/mass, consolidation, and pneumothorax were 0.9883, 1.000, and 0.9997, respectively. For the image classification, the overall AUC of the pooled physicians was 0.8679 without DCAD and 0.9112 with DCAD. Regarding lesion detection, the pooled observers exhibited a weighted JAFROC figure of merit (FOM) of 0.8426 without DCAD and 0.9112 with DCAD. DCAD for CRs could enhance physicians' performance in the detection of 3 major thoracic abnormalities.

Mixed-data deep learning in repeated predictions of general medicine length of stay: a derivation study.

The accurate prediction of likely discharges and estimates of length of stay (LOS) aid in effective hospital administration and help to prevent access block. Machine learning (ML) may be able to help with these tasks. For consecutive patients admitted under General Medicine at the Royal Adelaide Hospital over an 8-month period, daily ward round notes and relevant discrete data fields were collected from the electronic medical record. These data were then split into training and testing sets (7-month/1-month train/test split) prior to use in ML analyses aiming to predict discharge within the next 2 days, discharge within the next 7 days and an estimated date of discharge (EDD). Artificial neural networks and logistic regression were effective at predicting discharge within 48 h of a given ward round note. These models achieved an area under the receiver operator curve (AUC) of 0.80 and 0.78, respectively. Prediction of discharge within 7 days of a given note was less accurate, with artificial neural network returning an AUC of 0.68 and logistic regression an AUC of 0.61. The generation of an exact EDD remains inaccurate. This study has shown that repeated estimates of LOS using daily ward round notes and mixed-data inputs are effective in the prediction of general medicine discharges in the next 48 h. Further research may seek to prospectively and externally validate models for prediction of upcoming discharge, as well as combination human-ML approaches for generating EDDs.

Can a Deep-learning Model for the Automated Detection of Vertebral Fractures Approach the Performance Level of Human Subspecialists?

BACKGROUND: Vertebral fractures are the most common osteoporotic fractures in older individuals. Recent studies suggest that the performance of artificial intelligence is equal to humans in detecting osteoporotic fractures, such as fractures of the hip, distal radius, and proximal humerus. However, whether artificial intelligence performs as well in the detection of vertebral fractures on plain lateral spine radiographs has not yet been reported. QUESTIONS/PURPOSES: (1) What is the accuracy, sensitivity, specificity, and interobserver reliability (kappa value) of an artificial intelligence model in detecting vertebral fractures, based on Genant fracture grades, using plain lateral spine radiographs compared with values obtained by human observers? (2) Do patients' clinical data, including the anatomic location of the fracture (thoracic or lumbar spine), T-score on dual-energy x-ray absorptiometry, or fracture grade severity, affect the performance of an artificial intelligence model? (3) How does the artificial intelligence model perform on external validation? METHODS: Between 2016 and 2018, 1019 patients older than 60 years were treated for vertebral fractures in our institution. Seventy-eight patients were excluded because of missing CT or MRI scans (24% [19]), poor image quality in plain lateral radiographs of spines (54% [42]), multiple myeloma (5% [4]), and prior spine instrumentation (17% [13]). The plain lateral radiographs of 941 patients (one radiograph per person), with a mean age of 76 ± 12 years, and 1101 vertebral fractures between T7 and L5 were retrospectively evaluated for training (n = 565), validating (n = 188), and testing (n = 188) of an artificial intelligence deep-learning model. The gold standard for diagnosis (ground truth) of a vertebral fracture is the interpretation of the CT or MRI reports by a spine surgeon and a radiologist independently. If there were any disagreements between human observers, the corresponding CT or MRI images would be rechecked by them together to reach a consensus. For the Genant classification, the injured vertebral body height was measured in the anterior, middle, and posterior third. Fractures were classified as Grade 1 (< 25%), Grade 2 (26% to 40%), or Grade 3 (> 40%). The framework of the artificial intelligence deep-learning model included object detection, data preprocessing of radiographs, and classification to detect vertebral fractures. Approximately 90 seconds was needed to complete the procedure and obtain the artificial intelligence model results when applied clinically. The accuracy, sensitivity, specificity, interobserver reliability (kappa value), receiver operating characteristic curve, and area under the curve (AUC) were analyzed. The bootstrapping method was applied to our testing dataset and external validation dataset. The accuracy, sensitivity, and specificity were used to investigate whether fracture anatomic location or T-score in dual-energy x-ray absorptiometry report affected the performance of the artificial intelligence model. The receiver operating characteristic curve and AUC were used to investigate the relationship between the performance of the artificial intelligence model and fracture grade. External validation with a similar age population and plain lateral radiographs from another medical institute was also performed to investigate the performance of the artificial intelligence model. RESULTS: The artificial intelligence model with ensemble method demonstrated excellent accuracy (93% [773 of 830] of vertebrae), sensitivity (91% [129 of 141]), and specificity (93% [644 of 689]) for detecting vertebral fractures of the lumbar spine. The interobserver reliability (kappa value) of the artificial intelligence performance and human observers for thoracic and lumbar vertebrae were 0.72 (95% CI 0.65 to 0.80; p < 0.001) and 0.77 (95% CI 0.72 to 0.83; p < 0.001), respectively. The AUCs for Grades 1, 2, and 3 vertebral fractures were 0.919, 0.989, and 0.990, respectively. The artificial intelligence model with ensemble method demonstrated poorer performance for discriminating normal osteoporotic lumbar vertebrae, with a specificity of 91% (260 of 285) compared with nonosteoporotic lumbar vertebrae, with a specificity of 95% (222 of 234). There was a higher sensitivity 97% (60 of 62) for detecting osteoporotic (dual-energy x-ray absorptiometry T-score ≤ -2.5) lumbar vertebral fractures, implying easier detection, than for nonosteoporotic vertebral fractures (83% [39 of 47]). The artificial intelligence model also demonstrated better detection of lumbar vertebral fractures compared with detection of thoracic vertebral fractures based on the external dataset using various radiographic techniques. Based on the dataset for external validation, the overall accuracy, sensitivity, and specificity on bootstrapping method were 89%, 83%, and 95%, respectively. CONCLUSION: The artificial intelligence model detected vertebral fractures on plain lateral radiographs with high accuracy, sensitivity, and specificity, especially for osteoporotic lumbar vertebral fractures (Genant Grades 2 and 3). The rapid reporting of results using this artificial intelligence model may improve the efficiency of diagnosing vertebral fractures. The testing model is available at http://140.113.114.104/vght_demo/corr/. One or multiple plain lateral radiographs of the spine in the Digital Imaging and Communications in Medicine format can be uploaded to see the performance of the artificial intelligence model. LEVEL OF EVIDENCE: Level II, diagnostic study.

Prediction of age and sex from paranasal sinus images using a deep learning network.

ABSTRACT: This study was conducted to develop a convolutional neural network (CNN)-based model to predict the sex and age of patients by identifying unique unknown features from paranasal sinus (PNS) X-ray images.We employed a retrospective study design and used anonymized patient imaging data. Two CNN models, adopting ResNet-152 and DenseNet-169 architectures, were trained to predict sex and age groups (20-39, 40-59, 60+ years). The area under the curve (AUC), algorithm accuracy, sensitivity, and specificity were assessed. Class-activation map (CAM) was used to detect deterministic areas. A total of 4160 PNS X-ray images were collected from 4160 patients. The PNS X-ray images of patients aged ≥20 years were retrieved from the picture archiving and communication database system of our institution. The classification performances in predicting the sex (male vs female) and 3 age groups (20-39, 40-59, 60+ years) for each established CNN model were evaluated.For sex prediction, ResNet-152 performed slightly better (accuracy = 98.0%, sensitivity = 96.9%, specificity = 98.7%, and AUC = 0.939) than DenseNet-169. CAM indicated that maxillary sinuses (males) and ethmoid sinuses (females) were major factors in identifying sex. Meanwhile, for age prediction, the DenseNet-169 model was slightly more accurate in predicting age groups (77.6 ±â€Š1.5% vs 76.3 ±â€Š1.1%). CAM suggested that the maxillary sinus and the periodontal area were primary factors in identifying age groups.Our deep learning model could predict sex and age based on PNS X-ray images. Therefore, it can assist in reducing the risk of patient misidentification in clinics.

Artificial intelligence in dermatopathology: Diagnosis, education, and research.

Artificial intelligence (AI) utilizes computer algorithms to carry out tasks with human-like intelligence. Convolutional neural networks, a type of deep learning AI, can classify basal cell carcinoma, seborrheic keratosis, and conventional nevi, highlighting the potential for deep learning algorithms to improve diagnostic workflow in dermatopathology of highly routine diagnoses. Additionally, convolutional neural networks can support the diagnosis of melanoma and may help predict disease outcomes. Capabilities of machine learning in dermatopathology can extend beyond clinical diagnosis to education and research. Intelligent tutoring systems can teach visual diagnoses in inflammatory dermatoses, with measurable cognitive effects on learners. Natural language interfaces can instruct dermatopathology trainees to produce diagnostic reports that capture relevant detail for diagnosis in compliance with guidelines. Furthermore, deep learning can power computation- and population-based research. However, there are many limitations of deep learning that need to be addressed before broad incorporation into clinical practice. The current potential of AI in dermatopathology is to supplement diagnosis, and dermatopathologist guidance is essential for the development of useful deep learning algorithms. Herein, the recent progress of AI in dermatopathology is reviewed with emphasis on how deep learning can influence diagnosis, education, and research.

Data Homogeneity Effect in Deep Learning-Based Prediction of Type 1 Diabetic Retinopathy.

This study is aimed at evaluating a deep transfer learning-based model for identifying diabetic retinopathy (DR) that was trained using a dataset with high variability and predominant type 2 diabetes (T2D) and comparing model performance with that in patients with type 1 diabetes (T1D). The Kaggle dataset, which is a publicly available dataset, was divided into training and testing Kaggle datasets. In the comparison dataset, we collected retinal fundus images of T1D patients at Chang Gung Memorial Hospital in Taiwan from 2013 to 2020, and the images were divided into training and testing T1D datasets. The model was developed using 4 different convolutional neural networks (Inception-V3, DenseNet-121, VGG1, and Xception). The model performance in predicting DR was evaluated using testing images from each dataset, and area under the curve (AUC), sensitivity, and specificity were calculated. The model trained using the Kaggle dataset had an average (range) AUC of 0.74 (0.03) and 0.87 (0.01) in the testing Kaggle and T1D datasets, respectively. The model trained using the T1D dataset had an AUC of 0.88 (0.03), which decreased to 0.57 (0.02) in the testing Kaggle dataset. Heatmaps showed that the model focused on retinal hemorrhage, vessels, and exudation to predict DR. In wrong prediction images, artifacts and low-image quality affected model performance. The model developed with the high variability and T2D predominant dataset could be applied to T1D patients. Dataset homogeneity could affect the performance, trainability, and generalization of the model.

A deep-learning system for the assessment of cardiovascular disease risk via the measurement of retinal-vessel calibre.

Retinal blood vessels provide information on the risk of cardiovascular disease (CVD). Here, we report the development and validation of deep-learning models for the automated measurement of retinal-vessel calibre in retinal photographs, using diverse multiethnic multicountry datasets that comprise more than 70,000 images. Retinal-vessel calibre measured by the models and by expert human graders showed high agreement, with overall intraclass correlation coefficients of between 0.82 and 0.95. The models performed comparably to or better than expert graders in associations between measurements of retinal-vessel calibre and CVD risk factors, including blood pressure, body-mass index, total cholesterol and glycated-haemoglobin levels. In retrospectively measured prospective datasets from a population-based study, baseline measurements performed by the deep-learning system were associated with incident CVD. Our findings motivate the development of clinically applicable explainable end-to-end deep-learning systems for the prediction of CVD on the basis of the features of retinal vessels in retinal photographs.

Differentiation Between Anteroposterior and Posteroanterior Chest X-Ray View Position With Convolutional Neural Networks.

PURPOSE: Detection and validation of the chest X-ray view position with use of convolutional neural networks to improve meta-information for data cleaning within a hospital data infrastructure. MATERIAL AND METHODS: Within this paper we developed a convolutional neural network which automatically detects the anteroposterior and posteroanterior view position of a chest radiograph. We trained two different network architectures (VGG variant and ResNet-34) with data published by the RSNA (26 684 radiographs, class distribution 46 % AP, 54 % PA) and validated these on a self-compiled dataset with data from the University Hospital Essen (4507, radiographs, class distribution 55 % PA, 45 % AP) labeled by a human reader. For visualization and better understanding of the network predictions, a Grad-CAM was generated for each network decision. The network results were evaluated based on the accuracy, the area under the curve (AUC), and the F1-score against the human reader labels. Also a final performance comparison between model predictions and DICOM labels was performed. RESULTS: The ensemble models reached accuracy and F1-scores greater than 95 %. The AUC reaches more than 0.99 for the ensemble models. The Grad-CAMs provide insight as to which anatomical structures contributed to a decision by the networks which are comparable with the ones a radiologist would use. Furthermore, the trained models were able to generalize over mislabeled examples, which was found by comparing the human reader labels to the predicted labels as well as the DICOM labels. CONCLUSION: The results show that certain incorrectly entered meta-information of radiological images can be effectively corrected by deep learning in order to increase data quality in clinical application as well as in research. KEY POINTS: · The predictions for both view positions are accurate with respect to external validation data.. · The networks based their decisions on anatomical structures and key points that were in-line with prior knowledge and human understanding.. · Final models were able to detect labeling errors within the test dataset.. CITATION FORMAT: · Hosch R, Kroll L, Nensa F et al. Differentiation Between Anteroposterior and Posteroanterior Chest X-Ray View Position With Convolutional Neural Networks. Fortschr Röntgenstr 2021; 193: 168 - 176.

Hierarchical Analysis of Factors Associated with T Staging of Gastric Cancer by Endoscopic Ultrasound.

BACKGROUND: Size, ulcer, differentiation, and location are known to be factors affecting the T stage accuracy of EUS in gastric cancer. However, whether an interaction exists among recognized variables is poorly understood. The aim of this study was to identify the combinatorial characteristics of group with high overestimation rate to determine which group should be considered carefully for EUS-based treatment plans. METHODS: We retrospectively analyzed early gastric cancer patients who underwent EUS from 2005 to 2016. The accuracy of EUS T stage and factors affecting over-/underestimation were examined by using decision tree analysis, the CHAID method. RESULTS: The most significant factor affecting the accuracy of the EUS T stage was the size. The rate of overestimation was higher in lesions > 3 cm (37.2% vs. 28.8% vs. 17.1%, p < 0.001). In lesions > 3 cm, the rate of overestimation was higher in lesions with an ulcer (62.1% vs. 35.0%, p < 0.001). Moreover, for lesions ≤ 3 cm, the accuracy of the EUS T stage was more affected by differentiation and location. The rate of overestimation was higher in undifferentiated-type lesions ≤ 2 cm (24.5% vs. 13.9%, p < 0.001) and 2-3 cm (33.3% vs. 25.7%, p = 0.011). In the differentiated type, the location affected the accuracy of the EUS T stage. CONCLUSION: In this hierarchical analysis, the rate of overestimation was higher in lesions > 3 cm with ulcer, lesions > 3 cm irrespective of ulcer, and undifferentiated-type lesions measuring 2-3 cm.

Determination of disease severity in COVID-19 patients using deep learning in chest X-ray images.

PURPOSE: Chest X-ray plays a key role in diagnosis and management of COVID-19 patients and imaging features associated with clinical elements may assist with the development or validation of automated image analysis tools. We aimed to identify associations between clinical and radiographic features as well as to assess the feasibility of deep learning applied to chest X-rays in the setting of an acute COVID-19 outbreak. METHODS: A retrospective study of X-rays, clinical, and laboratory data was performed from 48 SARS-CoV-2 RT-PCR positive patients (age 60±17 years, 15 women) between February 22 and March 6, 2020 from a tertiary care hospital in Milan, Italy. Sixty-five chest X-rays were reviewed by two radiologists for alveolar and interstitial opacities and classified by severity on a scale from 0 to 3. Clinical factors (age, symptoms, comorbidities) were investigated for association with opacity severity and also with placement of central line or endotracheal tube. Deep learning models were then trained for two tasks: lung segmentation and opacity detection. Imaging characteristics were compared to clinical datapoints using the unpaired student's t-test or Mann-Whitney U test. Cohen's kappa analysis was used to evaluate the concordance of deep learning to conventional radiologist interpretation. RESULTS: Fifty-six percent of patients presented with alveolar opacities, 73% had interstitial opacities, and 23% had normal X-rays. The presence of alveolar or interstitial opacities was statistically correlated with age (P = 0.008) and comorbidities (P = 0.005). The extent of alveolar or interstitial opacities on baseline X-ray was significantly associated with the presence of endotracheal tube (P = 0.0008 and P = 0.049) or central line (P = 0.003 and P = 0.007). In comparison to human interpretation, the deep learning model achieved a kappa concordance of 0.51 for alveolar opacities and 0.71 for interstitial opacities. CONCLUSION: Chest X-ray analysis in an acute COVID-19 outbreak showed that the severity of opacities was associated with advanced age, comorbidities, as well as acuity of care. Artificial intelligence tools based upon deep learning of COVID-19 chest X-rays are feasible in the acute outbreak setting.