Segmental longitudinal strain (SLS) of the left ventricle (LV) is an important prognostic indicator for evaluating regional LV dysfunction, in particular for diagnosing and managing myocardial ischemia. Current techniques for strain estimation require significant manual intervention and expertise, limiting their efficiency and making them too resource-intensive for monitoring purposes. This study introduces the first automated pipeline, autoStrain, for SLS estimation in transesophageal echocardiography (TEE) using deep learning (DL) methods for motion estimation. We present a comparative analysis of two DL approaches: TeeFlow, based on the RAFT optical flow model for dense frame-to-frame predictions, and TeeTracker, based on the CoTracker point trajectory model for sparse long-sequence predictions. As ground truth motion data from real echocardiographic sequences are hardly accessible, we took advantage of a unique simulation pipeline (SIMUS) to generate a highly realistic synthetic TEE (synTEE) dataset of 80 patients with ground truth myocardial motion to train and evaluate both models. Our evaluation shows that TeeTracker outperforms TeeFlow in accuracy, achieving a mean distance error in motion estimation of 0.65 $\pm$ 0.20 mm on a synTEE test dataset. Clinical validation on 16 patients further demonstrated that SLS estimation with our autoStrain pipeline aligned with clinical references, achieving a mean difference (95% limits of agreement) of 1.09% (-8.90% to 11.09%). Incorporation of simulated ischemia in the synTEE data improved the accuracy of the models in quantifying abnormal deformation. Our findings indicate that integrating AI-driven motion estimation with TEE can significantly enhance the precision and efficiency of cardiac function assessment in clinical settings.

Semantic segmentation is essential for analyzing high-definition remote sensing images (HRSIs) because it allows the precise classification of objects and regions at the pixel level. However, remote sensing data present challenges owing to geographical location, weather, and environmen-tal variations, making it difficult for semantic segmentation models to generalize across diverse scenarios. Ex-isting methods are often limited to specific data domains and require expert annotators and specialized equipment for semantic labeling. In this study, we propose a novel unsupervised domain adaptation technique for remote sensing semantic segmentation by utilizing geographical coor-dinates that are readily accessible in remote sensing se-tups as metadata in a dataset. To bridge the domain gap, we propose a novel approach that considers the combination of an image's location-encoding trait and the spheri-cal nature of Earth's surface. Our proposed SegDesicNet module regresses the GRID positional encoding of the geo-coordinates projected over the unit sphere to obtain the domain loss. Our experimental results demonstrate that the proposed SegDesicNet outperforms state-of-the-art do-main adaptation methods in remote sensing image segmentation, achieving an improvement of approximately 6% in the mean intersection over union (MIoU) with a ~ 27% drop in parameter count on benchmarked subsets of the publicly available FLAIR #1 dataset. We also benchmarked our method performance on the custom split of the ISPRS Potsdam dataset. Our algorithm seeks to reduce the modeling disparity between artificial neural networks and human comprehension of the physical world, making the technol-ogy more human-centric and scalable.

This study introduces a robust vehicle localization framework designed for GNSS-limited environments. The proposed approach dynamically integrates georeferenced snow poles—fixed markers used to delineate road boundaries in winter with LiDAR-based odometry to enhance vehicle positioning and navigation. By alternating between GNSS data and LiDAR-based localization depending on GNSS signal availability, the framework addresses the challenges of GNSS-denied environments while leveraging sparse GNSS signals when available. A newly developed dataset of 360-degree snow pole images, captured using an Ouster OS2-128 LiDAR sensor, demonstrates the system’s applicability for autonomous driving. The method achieves a median localization error of$8.39 \, \text {m}$in GNSS-denied conditions, significantly outperforming techniques like FastReg ($35.68 \, \text {m}$), and progressively improves to sub-meter accuracy as GNSS availability increases. The open-source pipeline, to be made available onhttps://github.com/bdps1989/Snow-pole-based-vehicle-localization, offers a scalable, reliable, and near real-time solution for autonomous navigation in Nordic winter conditions, advancing research in localization under adverse environments.

The fast development of Large Language Models (LLMs) has made transformative applications in several fields attainable or possible. However, language models must often be more effective in specialized areas, especially health and prevention. This paper presents a novel method for fine-tuning LLMs using activity-related data to optimize them for the applicability of such models in health and wellbeing applications. We thus empirically evaluate this approach on the Cardiac Exercise Research corpus for fine-tuning the LLMs. The fine-tuning utilizes Quantized Low-Rank Adaptation (QLoRA) to ensure the models’ size remains small while maintaining high performance and accuracy to keep the semantic understanding and relevance with health-related queries. Our results in answering domain-related prompts showed an improved user satisfaction and sentiment scores, providing strong confidence in the method’s effectiveness. This study highlights the potential of domain-specific LLMs in advancing personalized healthcare. It instills a sense of optimism about the future of healthcare and the seamless integration of AI within health prevention and well-being domains.

The integration of Light Detection and Ranging (LiDAR) and Global Navigation Satellite System (GNSS) technologies marks a significant advancement in the fields of autonomous driving and intelligent transportation systems. This research introduces a methodology for geolocalizing road objects, specifically poles, by leveraging the detailed spatial data from LiDAR combined with the location capabilities of GNSS, while carefully accounting for these sensor offsets. Our approach takes advantage of the synergy between LiDAR’s exceptional spatial resolution and GNSS’s global positioning capability. This precision is crucial for the navigation systems of autonomous vehicles. By processing LiDAR data to detect objects and calculate their positions relative to the sensor, and then transforming these positions into global coordinates using inverse geodesic calculations, we present a methodology that can perform object geolocation in various environments. This paper details the development of the methodology, the challenges encountered, and the solutions devised, showcasing the approach’s performance through experimental results and suggests future directions for further research.

We address the task of in-the-wild human figure synthesis, where the primary goal is to synthesize a full body given any region in any image. In-the-wild human figure synthesis has long been a challenging and under-explored task, where current methods struggle to handle extreme poses, occluding objects, and complex backgrounds.Our main contribution is TriA-GAN, a keypoint-guided GAN that can synthesize Anyone, Anywhere, in Any given pose. Key to our method is projected GANs combined with a well-crafted training strategy, where our simple generator architecture can successfully handle the challenges of in-the-wild full-body synthesis. We show that TriA-GAN significantly improves over previous in-the-wild full-body synthesis methods, all while requiring less conditional information for synthesis (keypoints vs. DensePose). Finally, we show that the latent space of TriA-GAN is compatible with standard unconditional editing techniques, enabling text-guided editing of generated human figures.

Generative Adversarial Networks (GANs) are widely adopted for anonymization of human figures. However, current state-of-the-art limits anonymization to the task of face anonymization. In this paper, we propose a novel anonymization framework (DeepPrivacy2) for realistic anonymization of human figures and faces. We introduce a new large and diverse dataset for full-body synthesis, which significantly improves image quality and diversity of generated images. Furthermore, we propose a style-based GAN that produces high-quality, diverse, and editable anonymizations. We demonstrate that our full-body anonymization framework provides stronger privacy guarantees than previously proposed methods. Source code and appendix is available at: github.com/hukkelas/deep_privacy2.

Recent work on image anonymization has shown that generative adversarial networks (GANs) can generate near-photorealistic faces to anonymize individuals. However, scaling up these networks to the entire human body has remained a challenging and yet unsolved task. We propose a new anonymization method that generates realistic humans for in-the-wild images. A key part of our design is to guide adversarial nets by dense pixel-to-surface correspondences between an image and a canonical 3D surface. We introduce Variational Surface-Adaptive Modulation (V-SAM) that embeds surface information throughout the generator. Combining this with our novel discriminator surface supervision loss, the generator can synthesize high quality humans with diverse appearances in complex and varying scenes. We demonstrate that surface guidance significantly improves image quality and diversity of samples, yielding a highly practical generator. Finally, we show that our method preserves data usability without infringing privacy when collecting image datasets for training computer vision models. Source code and appendix is available at: github.com/hukkelas/full_body_anonymization

Perioperative monitoring of cardiac function is beneficial for early detection of cardiovascular complications. The standard of care for cardiac monitoring performed by trained cardiologists and anesthesiologists involves a manual and qualitative evaluation of ultrasound imaging, which is a time-demanding and resource-intensive process with intraobserver- and interobserver variability. In practice, such measures can only be performed a limited number of times during the intervention. To overcome these difficulties, this study presents a robust method for automatic and quantitative monitoring of cardiac function based on 3D transesophageal echocardiography (TEE) B-mode ultrasound recordings of the left ventricle (LV). Such an assessment obtains consistent measurements and can produce a near real-time evaluation of ultrasound imagery. Hence, the presented method is time-saving and results in increased accessibility. The mitral annular plane systolic excursion (MAPSE), characterizing global LV function, is estimated by landmark detection and cardiac view classification of two-dimensional images extracted along the long-axis of the ultrasound volume. MAPSE estimation directly from 3D TEE recordings is beneficial since it removes the need for manual acquisition of cardiac views, hence decreasing the need for interference by physicians. Two convolutional neural networks (CNNs) were trained and tested on acquired ultrasound data of 107 patients, and MAPSE estimates were compared to clinically obtained references in a blinded study including 31 patients. The proposed method for automatic MAPSE estimation had low bias and low variability in comparison to clinical reference measures. The method accomplished a mean difference for MAPSE estimates of (-0.16±1.06) mm. Thus, the results did not show significant systematic errors. The obtained bias and variance of the method were comparable to inter-observer variability of clinically obtained MAPSE measures on 2D TTE echocardiography. The novel pipeline proposed in this study has the potential to enhance cardiac monitoring in perioperative- and intensive care settings.

We present SIT-SR 3D, a novel self-supervised method for 3D single image super-resolution (SISR). Scaling 2D SISR networks to 3D SISR requires code redesign, high computing resources, and 3D ground-truth. However, we circumvent this by (1) using a pre-trained 2D SISR for indirect supervision and (2) using a novel consistency loss to learn frame interpolation. Any pre-trained state of the art 2D SISR method can replace the 2D SISR used in SIT-SR 3D, thus transferring the merits of 2D to 3D and ensuring modularity. We trained two end-to-end 3D baselines in a supervised setting; a 3D RRDBNet trained only with L1 loss and a 3D ESRGAN trained with adversarial and perceptual loss. We show that the proposed pipeline's self-supervised version is qualitatively better than the baselines. When trained in a supervised setting, SIT-SR 3D achieves better PSNR than its counterparts. Furthermore, our pipeline uses fewer parameters compared to the baselines. We demonstrate our results on an open-source digital rock CT dataset. Our code and pre-trained models will be made publicly available.

Deep learning-based methods, in particular, convolutional neural networks and fully convolutional networks are now widely used in the medical image analysis domain. The scope of this review focuses on the analysis using deep learning of focal liver lesions, with a special interest in hepatocellular carcinoma and metastatic cancer; and structures like the parenchyma or the vascular system. Here, we address several neural network architectures used for analyzing the anatomical structures and lesions in the liver from various imaging modalities such as computed tomography, magnetic resonance imaging and ultrasound. Image analysis tasks like segmentation, object detection and classification for the liver, liver vessels and liver lesions are discussed. Based on the qualitative search, 91 papers were filtered out for the survey, including journal publications and conference proceedings. The papers reviewed in this work are grouped into eight categories based on the methodologies used. By comparing the evaluation metrics, hybrid models performed better for both the liver and the lesion segmentation tasks, ensemble classifiers performed better for the vessel segmentation tasks and combined approach performed better for both the lesion classification and detection tasks. The performance was measured based on the Dice score for the segmentation, and accuracy for the classification and detection tasks, which are the most commonly used metrics.

Real-time 3D Echocardiography (RT3DE) has been proven to be an accurate tool for left ventricular (LV) volume assessment. However, identification of the LV endocardium remains a challenging task, mainly because of the low tissue/blood contrast of the images combined with typical artifacts. Several semi and fully automatic algorithms have been proposed for segmenting the endocardium in RT3DE data in order to extract relevant clinical indices, but a systematic and fair comparison between such methods has so far been impossible due to the lack of a publicly available common database. Here, we introduce a standardized evaluation framework to reliably evaluate and compare the performance of the algorithms developed to segment the LV border in RT3DE. A database consisting of 45 multivendor cardiac ultrasound recordings acquired at different centers with corresponding reference measurements from three experts are made available. The algorithms from nine research groups were quantitatively evaluated and compared using the proposed online platform. The results showed that the best methods produce promising results with respect to the experts' measurements for the extraction of clinical indices, and that they offer good segmentation precision in terms of mean distance error in the context of the experts' variability range. The platform remains open for new submissions.

The goal is to create an assistant for ultrasound- guided femoral nerve block. By segmenting and visualizing the important structures such as the femoral artery, we hope to improve the success of these procedures. This article is the first step towards this goal and presents novel real-time methods for identifying and reconstructing the femoral artery, and registering a model of the surrounding anatomy to the ultrasound images. The femoral artery is modelled as an ellipse. The artery is first detected by a novel algorithm which initializes the artery tracking. This algorithm is completely automatic and requires no user interaction. Artery tracking is achieved with a Kalman filter. The 3D artery is reconstructed in real-time with a novel algorithm and a tracked ultrasound probe. A mesh model of the surrounding anatomy was created from a CT dataset. Registration of this model is achieved by landmark registration using the centerpoints from the artery tracking and the femoral artery centerline of the model. The artery detection method was able to automatically detect the femoral artery and initialize the tracking in all 48 ultrasound sequences. The tracking algorithm achieved an average dice similarity coefficient of 0.91, absolute distance of 0.33 mm, and Hausdorff distance 1.05 mm. The mean registration error was 2.7 mm, while the average maximum error was 12.4 mm. The average runtime was measured to be 38, 8, 46 and 0.2 milliseconds for the artery detection, tracking, reconstruction and registration methods respectively.

Segmentation of anatomical structures, from modalities like computed tomography (CT), magnetic resonance imaging (MRI) and ultrasound, is a key enabling technology for medical applications such as diagnostics, planning and guidance. More efficient implementations are necessary, as most segmentation methods are computationally expensive, and the amount of medical imaging data is growing. The increased programmability of graphic processing units (GPUs) in recent years have enabled their use in several areas. GPUs can solve large data parallel problems at a higher speed than the traditional CPU, while being more affordable and energy efficient than distributed systems. Furthermore, using a GPU enables concurrent visualization and interactive segmentation, where the user can help the algorithm to achieve a satisfactory result. This review investigates the use of GPUs to accelerate medical image segmentation methods. A set of criteria for efficient use of GPUs are defined and each segmentation method is rated accordingly. In addition, references to relevant GPU implementations and insight into GPU optimization are provided and discussed. The review concludes that most segmentation methods may benefit from GPU processing due to the methods' data parallel structure and high thread count. However, factors such as synchronization, branch divergence and memory usage can limit the speedup.

In neurosurgery, information of blood flow is important to identify and avoid damage to important vessels. Three-dimensional intraoperative ultrasound color-Doppler imaging has proven useful in this respect. However, due to Doppler angle-dependencies and the complexity of the vascular architecture, clinical valuable 3-D information of flow direction and velocity is currently not available. In this work, we aim to correct for angle-dependencies in 3-D flow images based on a geometric model of the neurovascular tree generated on-the-fly from free-hand 2-D imaging and an accurate position sensor system. The 3-D vessel model acts as a priori information of vessel orientation used to angle-correct the Doppler measurements, as well as provide an estimate of the average flow direction. Based on the flow direction we were also able to do aliasing correction to approximately double the measurable velocity range. In vitro experiments revealed a high accuracy and robustness for estimating the mean direction of flow. Accurate angle-correction of axial velocities were possible given a sufficient beam-to-flow angle for at least parts of a vessel segment . In vitro experiments showed an absolute relative bias of 9.5% for a challenging low-flow scenario. The method also showed promising results in vivo, improving the depiction of flow in the distal branches of intracranial aneurysms and the feeding arteries of an arteriovenous malformation. Careful inspection by an experienced surgeon confirmed the correct flow direction for all in vivo examples.