In vitro fertilisation (IVF) is a widely adopted assisted reproductive technology that facilitates embryo development in cases of infertility. A critical step in this process is identifying which embryos successfully develop into blastocysts. The process needs continuous video monitoring, which is time-consuming and expensive. In this paper, we show a new deep learning method that predicts blastocyst formation by using just one image per day from each embryo, not full videos. Our method uses DINOv2 to extract detailed features from embryo images and a lightweight Video Vision Transformer(ViViT) version to analyze and classify the image sequences. Our experiments show that the proposed model can predict blastocyst and non-blastocyst formation with 95.7% accuracy. The proposed approach minimizes the reliance on costly continuous videomonitoring, indicating that more affordable equipment could be used without compromising precision. This improves both the practicality and cost-efficiency of implementation in IVF clinics, supporting embryologists in making more confident embryoselection decisions.

In the realm of stock market prediction, traditional supervised learning approaches often struggle with the vast and diverse nature of financial data, coupled with privacy concerns. This paper explores a novel methodology that combines unsupervised learning techniques with federated learning system to enhance stock market prediction models. We present a comprehensive system where local models, trained using unsupervised methods, contribute to a global model through federated aggregation. By leveraging federated learning, our approach allows multiple financial institutions to collaboratively train models on their decentralized data while preserving data privacy. This approach addresses the challenges of data heterogeneity and communication efficiency, providing a robust and scalable solution for advanced stock market forecasting. Our experiments demonstrate that integrating unsupervised learning with federated learning not only improves predictive accuracy but also enhances the model’s ability to identify emerging market trends and anomalies. Finally, we compare our distributed data model with other machine learning models that use local data.

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder characterized by the gradual decline in cognitive functions, particularly memory and reasoning. Early detection, especially during cognitive impairment (MCI) stage, is crucial for timely intervention and management. Enhanced diagnostic methods are essential for facilitating early identification and improving patient outcomes. This study presents a robust deep learning framework for the early detection of Alzheimer’s disease. It employs transfer learning and hyperparameter-tuning of InceptionResnetV2, InceptionV3, Xception architectures to enhance feature extraction by leveraging their pre-trained capabilities. An ensemble voting mechanism has been integrated to combine predictions from different models, optimizing both accuracy and robustness. The proposed ensemble voting approach demonstrated exceptional performance, achieving 98.96% accuracy and 100% precision for predicting classes Mildly Demented and Moderately Demented. It outperformed baseline and state-of-the-art models, highlighting its potential as a reliable tool for early diagnosis and intervention.

Context is an important factor in computer vision as it offers valuable information to clarify and analyze visual data. Utilizing the contextual information inherent in an image or a video can improve the precision and effectiveness of object detectors. For example, where recognizing an isolated object might be challenging, context information can improve comprehension of the scene. This study explores the impact of various context-based approaches to object detection. Initially, we investigate the role of context in object detection and survey it from several perspectives. We then review and discuss the most recent context-based object detection approaches and compare them. Finally, we conclude by addressing research questions and identifying gaps for further studies. More than 265 publications are included in this survey, covering different aspects of context in different categories of object detection, including general object detection, video object detection, small object detection, camouflaged object detection, zero-shot, one-shot, and few-shot object detection. This literature review presents a comprehensive overview of the latest advancements in context-based object detection, providing valuable contributions such as a thorough understanding of contextual information and effective methods for integrating various context types into object detection, thus benefiting researchers.

In vitro fertilization (IVF) is of great aid to couples who are struggling to conceive. The IVF clinics, where couples undergo fertility treatments, require a carefully controlled environment to ensure the effectiveness of the procedures. In recent years, IVF has seen significant progress, thanks to new technologies and methods that improve success rates and expand options for infertile couples. One notable advancement involves combining pre-implantation genetic testing (PGT) with time-lapse imaging technology, which allows continuous monitoring of embryo development with minimal disturbance. This innovation improves the selection of healthy embryos for transfer, increasing success rates and reducing the risk of multiple pregnancies. However, maintaining a stable environment remains a key challenge. Fluctuations in temperature, humidity, air quality, and particulate matter can affect IVF success rates by disrupting the embryo’s delicate environment and potentially causing implantation failure. We discuss in this position paper our approach to alleviate such environmental problems in our project EIVF-AI funded by the Swedish funding agency Vinnova.

This study explores the use of machine learning to enhance patient mobility and caregiver ergonomics by optimizing the use of mobility aids. Traditional manual assessments can be subjective and inaccurate, so this research develops a data-driven model for object detection and human activity recognition. A computer vision dataset was created using video recordings of controlled caregiving scenarios. The study leverages advanced machine learning models, including YOLO for object detection, pose estimation, ResNet-18 for frame classification, Inception-v4 for feature extraction, and LSTM for sequence modeling. The findings provide valuable insights into integrating machine learning into mobility aids, improving both patient outcomes and caregiver well-being.

The COVID-19 pandemic has underscored the critical challenges faced by healthcare systems worldwide, particularly in meeting the escalating demand for resources such as ICU beds, specialized care, and medical equipment. This shortfall has resulted in significant loss of life, highlighting the urgent need for accurate and timely diagnosis to optimize patient outcomes and reduce healthcare costs. In response to these challenges, our research focuses on developing a machine learning system capable of predicting whether patients will require ICU admission or can be managed remotely at home during peak periods of demand. Leveraging a novel two-dimensional reduction approach that combines evolutionary algorithms, Pattern Discovery, and machine learning techniques, we aim to streamline patient-collected data to train predictive models capable of forecasting ICU needs and remote care requirements. By providing healthcare systems with the ability to anticipate patient needs during critical phases of the pandemic, our predictive model empowers healthcare providers to allocate resources more effectively, optimize patient care delivery, and mitigate the impact of healthcare crises. The results of our experimental evaluation demonstrate the promising potential of our approach in addressing the pressing challenges posed by the COVID-19 pandemic and similar public health emergencies.

Object detection is a critical task in computer vision with wide-ranging applications, from autonomous driving tosurveillance systems. Despite notable progress, challenges such as detecting small objects, managing occlusions,and effectively integrating multiscale features persist. We propose RetinaGate, a novel object detection architec-ture that introduces a Gated Feature Pyramid Network (G-FPN) to adaptively fuse multi-scale features, enhancedby Squeeze-and-Excitation-based channel attention for improved accuracy. As a plug-and-play module, G-FPNcan be seamlessly integrated into existing detection models to enhance their accuracy. These enhancementsstrengthen the model’s capacity to capture fine-grained details and leverage contextual information more effec-tively. Experimental results on three benchmark datasets demonstrate that RetinaGate outperforms the baselineRetinaNet in terms of detection accuracy, particularly in challenging detection scenarios such as underwater.

This study aims to integrate embryo images and environmental laboratory factors to predict embryoquality using a complex machine learning model. A challenge was data misalignment, which was solvedby using a Random Forest Regressor to synthesise data for a complete dataset. The fine-tuned InceptionV3 model with the added attention mechanisms inherent to transformers was used for multitask learningto predict three different scores for embryo quality: cell expansion (EXP), inner cell mass (ICP) andtrophectoderm (TE). The model achieved an accuracy and F1 Score of 92.76%, 92.74% for EXP, 72.63%,59.52% for TE and 63.69%, 10.77% for ICM prediction. These results indicate a great performance for 2 ofthe three scores and build a basis for a reliable model for prediction of embryo quality.

Falls frequently present substantial safety hazards to those who are alone, particularly the elderly. Deploying a rapid and proficient method for detecting falls is a highly effective approach to tackle this concealed peril. The majority of existing fall detection methods rely on either visual data or wearable devices, both of which have drawbacks. This research presents a multimodal approach that integrates video and audio modalities to address the issue of fall detection systems and enhances the accuracy of fall detection in challenging environmental conditions. This multimodal approach, which leverages the benefits of attention mechanism in both video and audio streams, utilizes features from both modalities through feature-level fusion to detect falls in unfavorable conditions where visual systems alone are unable to do so. We assessed the performance of our multimodal fall detection model using Le2i and UP-Fall datasets. Additionally, we compared our findings with other fall detection methods. The outstanding results of our multimodal model indicate its superior performance compared to single fall detection models.