The temporal bone is a part of the lateral skull surface that contains organs responsible for hearing and balance. Mastering surgery of the temporal bone is challenging because of this complex and microscopic three-dimensional anatomy. Segmentation of intra-temporal anatomy based on computed tomography (CT) images is necessary for applications such as surgical training and rehearsal, amongst others. However, temporal bone segmentation is challenging due to the similar intensities and complicated anatomical relationships among critical structures, undetectable small structures on standard clinical CT, and the amount of time required for manual segmentation. This paper describes a single multi-class deep learning-based pipeline as the first fully automated algorithm for segmenting multiple temporal bone structures from CT volumes, including the sigmoid sinus, facial nerve, inner ear, malleus, incus, stapes, internal carotid artery and internal auditory canal. The proposed fully convolutional network, PWD-3DNet, is a patch-wise densely connected (PWD) three-dimensional (3D) network. The accuracy and speed of the proposed algorithm was shown to surpass current manual and semi-automated segmentation techniques. The experimental results yielded significantly high Dice similarity scores and low Hausdorff distances for all temporal bone structures with an average of 86% and 0.755 millimeter (mm), respectively. We illustrated that overlapping in the inference sub-volumes improves the segmentation performance. Moreover, we proposed augmentation layers by using samples with various transformations and image artefacts to increase the robustness of PWD-3DNet against image acquisition protocols, such as smoothing caused by soft tissue scanner settings and larger voxel sizes used for radiation reduction. The proposed algorithm was tested on low-resolution CTs acquired by another center with different scanner parameters than the ones used to create the algorithm and shows potential for application beyond the particular training data used in the study.

Machine learning and computer vision have proven to be valuable tools for farmers to streamline their resource utilization to lead to more sustainable and efficient agricultural production. These techniques have been applied to strawberry cultivation in the past with limited success. To build on this past work, in this study, two separate sets of strawberry images, along with their associated diseases, were collected and subjected to resizing and augmentation. Subsequently, a combined dataset consisting of nine classes was utilized to fine-tune three distinct pretrained models: vision transformer (ViT), MobileNetV2, and ResNet18. To address the imbalanced class distribution in the dataset, each class was assigned weights to ensure nearly equal impact during the training process. To enhance the outcomes, new images were generated by removing backgrounds, reducing noise, and flipping them. The performances of ViT, MobileNetV2, and ResNet18 were compared after being selected. Customization specific to the task was applied to all three algorithms, and their performances were assessed. Throughout this experiment, none of the layers were frozen, ensuring all layers remained active during training. Attention heads were incorporated into the first five and last five layers of MobileNetV2 and ResNet18, while the architecture of ViT was modified. The results indicated accuracy factors of 98.4%, 98.1%, and 97.9% for ViT, MobileNetV2, and ResNet18, respectively. Despite the data being imbalanced, the precision, which indicates the proportion of correctly identified positive instances among all predicted positive instances, approached nearly 99% with the ViT. MobileNetV2 and ResNet18 demonstrated similar results. Overall, the analysis revealed that the vision transformer model exhibited superior performance in strawberry ripeness and disease classification. The inclusion of attention heads in the early layers of ResNet18 and MobileNet18, along with the inherent attention mechanism in ViT, improved the accuracy of image identification. These findings offer the potential for farmers to enhance strawberry cultivation through passive camera monitoring alone, promoting the health and well-being of the population.

Autonomous vehicles (AVs) represent a transformative advance in automotive technology, promising increased safety and efficiency by reducing human error. However, integrating human factors remains a critical challenge, especially during takeover scenarios where the human driver must re-assume control of the vehicle. This review paper focuses on the engineering and human-centred design of takeover requests (TORs) within Level 3 autonomous vehicles , emphasizing the importance of seamless transitions between automated driving and manual control. We explore the concept of the Operational Design Domain (ODD), which dictates the specific conditions under which an AV may safely operate, and contextualize its role. Through a comprehensive analysis, we highlight how monitoring both the internal and external environment, and improving human-machine interfaces through the design of takeover requests (TOR), play pivotal roles in ensuring that transitions are safe and efficient. We argue for the necessity of integrating detailed human factors and ergonomic considerations to foster a human-centred approach in AV design. We aim to establish a symbiotic relationship between human drivers and autonomous systems, ensuring that AVs not only function optimally within their designated ODD, but also maintain high safety standards during critical takeover moments.

Background The use and detailed costs of services provided for people with advanced melanoma (amel) are not well known. We conducted an analysis to determine the use of health care services and the associated costs delineated by relevant attributable costs, which we defined for subjects in the province of Ontario. Methods Through the Ontario Cancer Data Linkage Project, a cohort of amel patients with diagnoses between 31 August 2005 and 2012 (follow-up to 2013) and with valid International Classification of Diseases (9th revision, Clinical Modification) 172 codes and histology codes was identified. A cohort of individuals with amel having a combination of at least 1 palliative, 1 medical oncology, and 1 hospitalization code was generated. The health system services used by this population were clustered into hospitalization, palliation, physician medical visits, medication, homecare, laboratory, diagnostics, and other resources. Overall rates of use and disaggregated costs were determined by phase of care for the entire cohort. Results The mean age for the 2748 individuals in the cohort was 67 years. The greater proportion of the patients were men (65.6%) and were more than 65 years of age (>50%). In this advanced cohort, fewer than 45% of patients were alive 3 years after the malignant melanoma diagnosis. The average annual cost per patient over the time horizon was $6,551. At $15,830, year 1 after diagnosis was the most expensive, followed by year 2, at $8,166. Conclusions Our data provide a baseline for the costs associated with amel treatment. Future studies will include newer agents and comparative effectiveness research for personalized therapies.

With the emergence of intelligent automotive systems, ensuring road safety has become paramount, necessitating the incorporation of sensors such as in-cabin cameras to comprehend drivers’ emotions. While emotion recognition through facial expression has demonstrated remarkable success in controlled laboratory settings, it often struggles to capture the complexities of real-world driving scenarios due to the lack of representative datasets. Facial Action Coding System (FACS) breaks down facial expressions into a combination of Action Units (AUs) and offers flexibility for more diverse facial expression categorization. In this study, we present an innovative framework designed to offer a solution to adapt machine learning (ML) models, including Random Forest (RF), Gradient Boosting (GB), and Long Short-Term Memory (LSTM) as a deep-learning model, originally trained on facial expression recognition (FER) datasets, to real-time monitoring of drivers’ emotional states. Notably, the LSTM model, trained on the CK+ dataset and tested on KMU-FED dataset, achieved the highest accuracy of 52.45%, while the best performance of the Random Forest model reached 50.12% accuracy on the same datasets. The goal of this study is to utilize AUs as descriptors in our framework, aiming to rectify inconsistencies across different datasets. This approach ensures enhanced precision in recognizing drivers’ emotions within the vehicle cabin. Additionally, acknowledging the deployment limitations inherent in on-vehicle systems, our framework is designed to develop compact models, thereby reducing the computational load.