Improving difficult direct laryngoscopy prediction using deep learning and minimal image analysis: a single-center prospective study.

Accurate prediction of difficult direct laryngoscopy (DDL) is essential to ensure optimal airway management and patient safety. The present study proposed an AI model that would accurately predict DDL using a small number of bedside pictures of the patient's face and neck taken simply with a smartphone. In this prospective single-center study, adult patients scheduled for endotracheal intubation under general anesthesia were included. Patient pictures were obtained in frontal, lateral, frontal-neck extension, and open mouth views. DDL prediction was performed using a deep learning model based on the EfficientNet-B5 architecture, incorporating picture view information through multitask learning. We collected 18,163 pictures from 3053 patients. After under-sampling to achieve a 1:1 image ratio of DDL to non-DDL, the model was trained and validated with a dataset of 6616 pictures from 1283 patients. The deep learning model achieved a receiver operating characteristic area under the curve of 0.81-0.88 and an F1-score of 0.72-0.81 for DDL prediction. Including picture view information improved the model's performance. Gradient-weighted class activation mapping revealed that neck and chin characteristics in frontal and lateral views are important factors in DDL prediction. The deep learning model we developed effectively predicts DDL and requires only a small set of patient pictures taken with a smartphone. The method is practical and easy to implement.

Robust Long-Term Vehicle Trajectory Prediction Using Link Projection and a Situation-Aware Transformer.

The trajectory prediction of a vehicle emerges as a pivotal component in Intelligent Transportation Systems. On urban roads where external factors such as intersections and traffic control devices significantly affect driving patterns along with the driver's intrinsic habits, the prediction task becomes much more challenging. Furthermore, long-term forecasting of trajectories accumulates prediction errors, leading to substantially inaccurate predictions that may deviate from the actual road. As a solution to these challenges, we propose a long-term vehicle trajectory prediction method that is robust to error accumulation and prevents off-road predictions. In this study, the Transformer model is utilized to analyze and forecast vehicle trajectories. In addition, we propose an extra encoding network to precisely capture the effect of the external factors on the driving pattern by producing an abstract representation of the situation nearby the driver. To avoid off-road predictions, we propose a post-processing method, called link projection, which projects predictions onto the road geometry. Moreover, to overcome the limitations of Euclidean distance-based evaluation metrics in evaluating the accuracy of the entire trajectory, we propose a new metric called area-between-curves (ABC). It measures the similarity between two trajectories, and thus the accordance between the two can be effectively evaluated. Extensive evaluations are conducted using real-world datasets against widely-used methods to demonstrate the effectiveness of the proposed approach. The results show that the proposed approach outperforms the conventional deep learning models by up to 65.74% (RMSE), 60.13% (MAE) and 91.45% (ABC).

Convolutional neural network model for automatic recognition and classification of pancreatic cancer cell based on analysis of lipid droplet on unlabeled sample by 3D optical diffraction tomography.

INTRODUCTION: Pancreatic cancer cells generally accumulate large numbers of lipid droplets (LDs), which regulate lipid storage. To promote rapid diagnosis, an automatic pancreatic cancer cell recognition system based on a deep convolutional neural network was proposed in this study using quantitative images of LDs from stain-free cytologic samples by optical diffraction tomography. METHODS: We retrieved 3D refractive index tomograms and reconstructed 37 optical images of one cell. From the four cell lines, the obtained fields were separated into training and test datasets with 10,397 and 3,478 images, respectively. Furthermore, we adopted several machine learning techniques based on a single image-based prediction model to improve the performance of the computer-aided diagnostic system. RESULTS: Pancreatic cancer cells had a significantly lower total cell volume and dry mass than did normal pancreatic cells and were accompanied by greater numbers of lipid droplets (LDs). When evaluating multitask learning techniques utilizing the EfficientNet-b3 model through confusion matrices, the overall 2-category accuracy for cancer classification reached 96.7 %. Simultaneously, the overall 4-category accuracy for individual cell line classification achieved a high accuracy of 96.2 %. Furthermore, when we added the core techniques one by one, the overall performance of the proposed technique significantly improved, reaching an area under the curve (AUC) of 0.997 and an accuracy of 97.06 %. Finally, the AUC reached 0.998 through the ablation study with the score fusion technique. DISCUSSION: Our novel training strategy has significant potential for automating and promoting rapid recognition of pancreatic cancer cells. In the near future, deep learning-embedded medical devices will substitute laborious manual cytopathologic examinations for sustainable economic potential.

Improving Estimation of Layer Thickness and Identification of Slicer for 3D Printing Forensics.

This study emphasizes the significance of estimating the layer thickness and identifying slicer programs in the realm of 3D printing forensics. With the progress in 3D printing technology, precise estimation of the layer thickness has become crucial. However, previous research on layer thickness estimation has mainly treated the problem as a classification task, which is inadequate for continuous layer thickness parameters. Furthermore, previous studies have concentrated on hardware-based printer identification, but the identification of slicer programs through 3D objects is a vital aspect of the software domain and can provide valuable clues for 3D printing forensics. In this study, a regression-based approach utilizing a vision transformer model was proposed. Experiments conducted on the SI3DP++ dataset demonstrated that the proposed model could handle a broad range of data and outperform the current classification models. Additionally, this study proposed a new research direction by introducing slicer program identification, which significantly contributes to the field of 3D printing forensics.

Development and Validation of Deep Learning-Based Algorithms for Predicting Lumbar Herniated Nucleus Pulposus Using Lumbar X-rays.

Lumbar herniated nucleus pulposus (HNP) is difficult to diagnose using lumbar radiography. HNP is typically diagnosed using magnetic resonance imaging (MRI). This study developed and validated an artificial intelligence model that predicts lumbar HNP using lumbar radiography. A total of 180,271 lumbar radiographs were obtained from 34,661 patients in the form of lumbar X-ray and MRI images, which were matched together and labeled accordingly. The data were divided into a training set (31,149 patients and 162,257 images) and a test set (3512 patients and 18,014 images). Training data were used for learning using the EfficientNet-B5 model and four-fold cross-validation. The area under the curve (AUC) of the receiver operating characteristic (ROC) for the prediction of lumbar HNP was 0.73. The AUC of the ROC for predicting lumbar HNP in L (lumbar) 1-2, L2-3, L3-4, L4-5, and L5-S (sacrum)1 levels were 0.68, 0.68, 0.63, 0.67, and 0.72, respectively. Finally, an HNP prediction model was developed, although it requires further improvements.

Efficacy of an artificial neural network algorithm based on thick-slab magnetic resonance cholangiopancreatography images for the automated diagnosis of common bile duct stones.

BACKGROUND AND AIM: Magnetic resonance cholangiopancreatography (MRCP) can accurately diagnose common bile duct (CBD) stones but is laborious to interpret. We developed an artificial neural network (ANN) capable of automatically assisting physicians with the diagnosis of CBD stones. This study aimed to evaluate the ANN's diagnostic performance for detecting CBD stones in thick-slab MRCP images and identify clinical factors predictive of accurate diagnosis. METHODS: The presence of CBD stones was confirmed via direct visualization through endoscopic retrograde cholangiopancreatography (ERCP). The absence of CBD stones was confirmed by either a negative endoscopic ultrasound accompanied by clinical improvements or negative findings on ERCP. Our base networks were constructed using state-of-the-art EfficientNet-B5 neural network models, which are widely used for image classification. RESULTS: In total, 3156 images were collected from 789 patients. Of these, 2628 images from 657 patients were used for training. An additional 1924 images from 481 patients were prospectively collected for validation. Across the entire prospective validation cohort, the ANN achieved a sensitivity, specificity, positive predictive value, negative predictive value, and overall accuracy of 93.03%, 97.05%, 97.01%, 93.12%, and 95.01%, respectively. Similarly, a radiologist achieved a sensitivity, specificity, positive predictive value, negative predictive value, and overall accuracy 91.16%, 93.25%, 93.22%, 90.20%, and 91.68%, respectively. In multivariate analysis, only bile duct diameter > 10 mm (odds ratio = 2.45, 95% confidence interval [1.08-6.07], P = 0.040) was related to ANN diagnostic accuracy. CONCLUSION: Our ANN algorithm automatically and quickly diagnoses CBD stones in thick-slab MRCP images, therein aiding physicians with optimizing clinical practice, such as whether to perform ERCP.

Convolutional Neural Network Architecture for Recovering Watermark Synchronization.

In this paper, we propose a convolutional neural network-based template architecture that compensates for the disadvantages of existing watermarking techniques that are vulnerable to geometric distortion. The proposed template consists of a template generation network, a template extraction network, and a template matching network. The template generation network generates a template in the form of noise and the template is inserted into certain pre-defined spatial locations of the image. The extraction network detects spatial locations where the template is inserted in the image. Finally, the template matching network estimates the parameters of the geometric distortion by comparing the shape of spatial locations where the template was inserted with the locations where the template was detected. It is possible to recover an image in its original geometrical form using the estimated parameters, and as a result, watermarks applied using existing watermarking techniques that are vulnerable to geometric distortion can be decoded normally.

Exposing Digital Image Forgeries by Detecting Contextual Abnormality Using Convolutional Neural Networks.

Traditionally, digital image forensics mainly focused on the low-level features of an image, such as edges and texture, because these features include traces of the image's modification history. However, previous methods that employed low-level features are highly vulnerable, even to frequently used image processing techniques such as JPEG and resizing, because these techniques add noise to the low-level features. In this paper, we propose a framework that uses deep neural networks to detect image manipulation based on contextual abnormality. The proposed method first detects the class and location of objects using a well-known object detector such as a region-based convolutional neural network (R-CNN) and evaluates the contextual scores according to the combination of objects, the spatial context of objects and the position of objects. Thus, the proposed forensics can detect image forgery based on contextual abnormality as long as the object can be identified even if noise is applied to the image, contrary to methods that employ low-level features, which are vulnerable to noise. Our experiments showed that our method is able to effectively detect contextual abnormality in an image.