Large Training Datasets Research Articles

Random Forest Spatial Interpolation Techniques for Crop Yield Estimation at District Level

General Crop Estimation Surveys (GCES) based on Crop Cutting Experiments (CCEs) are conducted for estimation of crop yield following random sampling approach for almost all major crops. About 13 lakh CCEs are conducted every year which has now increased rapidly due to the Pradhan Mantri Fasal Bima Yojana (PMFBY) which is yield based insurance scheme. As suggested by Ministry of Agriculture and Farmers’ Welfare (MoA&FW), this number needs to be reduced drastically by developing sampling procedures based on the use of advanced technologies and advanced survey techniques for crop yield estimation. In this study, an attempt has been made to develop crop yield estimation procedures using Random Forest Spatial Interpolation (RFSI) technique including the spatial variables like spatial distance and nearest neighbours as covariates. RFSI is one of the most adaptable and user-friendly interpolation techniques, as well as one of the fastest in large training datasets. Estimates of yield of wheat were obtained for all the six tehsils of Barabanki district using the estimator under stratified two stage sampling technique. The district level estimates were also obtained by pooling area under wheat crop in each tehsil along with the district level estimate of crop yield, estimate of variance, estimate of standard error (SE) and percentage SE (%SE) of these estimates were also computed in order to make comparison. The results of this study suggest that the estimates derived using RFSI are comparable to kriging and superior to inverse distance weighting (IDW) for the prediction of yield at unknown locations using distance and nearest neighbours.

Advancements in oligometastatic breast cancer: a comprehensive review of current strategies and the role of artificial intelligence.

In the dynamic landscape of Breast Cancer (BC), Oligo- Metastatic Breast Cancer (OMBC) presents unique challenges and opportunities. This comprehensive review delves into current strategies for addressing OMBC, covering locoregional and site-specific metastasis management, and addressing both surgical and minimally invasive therapies as essential components. Moreover, the transformative role of Artificial Intelligence (AI) is spotlighted. However, while the future looks promising, several limitations need addressing, including the need for further research, especially in diverse patient populations and resource-challenged settings. AI implementation may require overcoming the lack of Electronic Health Records acceptance in resource-challenged countries, which contributes to a scarcity of large datasets for AI training. As AI continues to evolve, validation and regulatory aspects must be continually addressed for seamless integration into clinical practice. In summary, this review outlines the evolving landscape of OMBC management, emphasizing the need for comprehensive research, global collaboration, and innovative AI solutions to enhance patient care and outcomes.

Evaluation of a Cascaded Deep Learning-based Algorithm for Prostate Lesion Detection at Biparametric MRI.

Background Multiparametric MRI (mpMRI) improves prostate cancer (PCa) detection compared with systematic biopsy, but its interpretation is prone to interreader variation, which results in performance inconsistency. Artificial intelligence (AI) models can assist in mpMRI interpretation, but large training data sets and extensive model testing are required. Purpose To evaluate a biparametric MRI AI algorithm for intraprostatic lesion detection and segmentation and to compare its performance with radiologist readings and biopsy results. Materials and Methods This secondary analysis of a prospective registry included consecutive patients with suspected or known PCa who underwent mpMRI, US-guided systematic biopsy, or combined systematic and MRI/US fusion-guided biopsy between April 2019 and September 2022. All lesions were prospectively evaluated using Prostate Imaging Reporting and Data System version 2.1. The lesion- and participant-level performance of a previously developed cascaded deep learning algorithm was compared with histopathologic outcomes and radiologist readings using sensitivity, positive predictive value (PPV), and Dice similarity coefficient (DSC). Results A total of 658 male participants (median age, 67 years [IQR, 61-71 years]) with 1029 MRI-visible lesions were included. At histopathologic analysis, 45% (294 of 658) of participants had lesions of International Society of Urological Pathology (ISUP) grade group (GG) 2 or higher. The algorithm identified 96% (282 of 294; 95% CI: 94%, 98%) of all participants with clinically significant PCa, whereas the radiologist identified 98% (287 of 294; 95% CI: 96%, 99%; P = .23). The algorithm identified 84% (103 of 122), 96% (152 of 159), 96% (47 of 49), 95% (38 of 40), and 98% (45 of 46) of participants with ISUP GG 1, 2, 3, 4, and 5 lesions, respectively. In the lesion-level analysis using radiologist ground truth, the detection sensitivity was 55% (569 of 1029; 95% CI: 52%, 58%), and the PPV was 57% (535 of 934; 95% CI: 54%, 61%). The mean number of false-positive lesions per participant was 0.61 (range, 0-3). The lesion segmentation DSC was 0.29. Conclusion The AI algorithm detected cancer-suspicious lesions on biparametric MRI scans with a performance comparable to that of an experienced radiologist. Moreover, the algorithm reliably predicted clinically significant lesions at histopathologic examination. ClinicalTrials.gov Identifier: NCT03354416 © RSNA, 2024 Supplemental material is available for this article.

An Automated Video Analysis System for Retrospective Assessment and Real-Time Monitoring of Endoscopic Procedures (with Video)

Background and Aims: Accurate recognition of endoscopic instruments facilitates quantitative evaluation and quality control of endoscopic procedures. However, no relevant research has been reported. In this study, we aimed to develop a computer-assisted system, EndoAdd, for automated endoscopic surgical video analysis based on our dataset of endoscopic instrument images. Methods: Large training and validation datasets containing 45,143 images of 10 different endoscopic instruments and a test dataset of 18,375 images collected from several medical centers were used in this research. Annotated image frames were used to train the state-of-the-art object detection model, YOLO-v5, to identify the instruments. Based on the frame-level prediction results, we further developed a hidden Markov model to perform video analysis and generate heatmaps to summarize the videos. Results: EndoAdd achieved high accuracy (>97%) on the test dataset for all 10 endoscopic instrument types. The mean average accuracy, precision, recall, and F1-score were 99.1%, 92.0%, 88.8%, and 89.3%, respectively. The area under the curve values exceeded 0.94 for all instrument types. Heatmaps of endoscopic procedures were generated for both retrospective and real-time analyses. Conclusions: We successfully developed an automated endoscopic video analysis system, EndoAdd, which supports retrospective assessment and real-time monitoring. It can be used for data analysis and quality control of endoscopic procedures in clinical practice.

Segmentation of liver CT images based on weighted medical transformer model.

Deep convolutional neural networks have made significant strides in the field of medical image segmentation. Although existing convolutional structures enhance performance by leveraging local image information, they often lose the interdependence information between contexts. Therefore, the article utilizes the multi-attention mechanism of the Transformer structure to more comprehensively express relationships between contexts and introduced the Transformer network architecture into the field of medical image segmentation. Most models based on this Transformer structure typically require large datasets for training. However, in the medical field, the limited size of datasets makes training models with the Transformer structure challenging. To address this, the article propose a Weighted Medical Transformer (WMT) model that imposes low requirements on dataset quantity. The weighting mechanism in the WMT model aims to improve the issue of inaccurate relative positional coding when dealing with small medical datasets. Additionally, a coarse-grained and fine-grained segmentation mechanism is introduced, focusing on both the detailed aspects within image blocks and the boundary information connecting blocks. Experimental results on a liver dataset demonstrate that the model achieves F1 and IoU scores of 88.48% and 79.41%, respectively. Results on the MoNuSeg dataset show comparable high F1 and IoU scores of 79.58% and 66.19%, respectively. The model's accuracy surpasses that of U-Net++ and U-Net models. Compared to other models, this approach is applicable to scenarios with limited datasets, exhibiting high execution efficiency and accuracy.

Quasi-Classical Trajectory Calculation of Rate Constants Using an Ab Initio Trained Machine Learning Model (aML-MD) with Multifidelity Data.

Machine learning (ML) provides a great opportunity for the construction of models with improved accuracy in classical molecular dynamics (MD). However, the accuracy of a ML trained model is limited by the quality and quantity of the training data. Generating large sets of accurate ab initio training data can require significant computational resources. Furthermore, inconsistent or incompatible data with different accuracies obtained using different methods may lead to biased or unreliable ML models that do not accurately represent the underlying physics. Recently, transfer learning showed its potential for avoiding these problems as well as for improving the accuracy, efficiency, and generalization of ML models using multifidelity data. In this work, ab initio trained ML-based MD (aML-MD) models are developed through transfer learning using DFT and multireference data from multiple sources with varying accuracy within the Deep Potential MD framework. The accuracy of the force field is demonstrated by calculating rate constants for the H + HO2 → H2 + 3O2 reaction using quasi-classical trajectories. We show that the aML-MD model with transfer learning can accurately predict the rate constants while reducing the computational cost by more than five times compared to the use of more expensive quantum chemistry training data sets. Hence, the aML-MD model with transfer learning shows great potential in using multifidelity data to reduce the computational cost involved in generating the training set for these potentials.

Fast-Track Development and Multi-institutional Clinical Validation of an Artificial Intelligence Algorithm for Detection of Lymph Node Metastasis in Colorectal Cancer

Lymph node metastasis (LNM) detection can be automated using artificial intelligence (AI)-based diagnostic tools. Only limited studies have addressed this task for colorectal cancer (CRC). This study aimed to develop of a clinical-grade digital pathology tool for LNM detection in CRC using the original fast-track framework.The training cohort included 432 slides from one department. A segmentation algorithm detecting 8 relevant tissue classes was trained. The test cohorts consisted of materials from 5 pathology departments digitized by 4 different scanning systems.A high-quality, large training data set was generated within 7 days and a minimal amount of annotation work using fast-track principles. The AI tool showed very high accuracy for LNM detection in all cohorts, with sensitivity, negative predictive value, and specificity ranges of 0.980 to 1.000, 0.997 to 1.000, and 0.913 to 0.990, correspondingly. Only 5 of 14,460 analyzed test slides with tumor cells over all cohorts were classified as false negative (3/5 representing clusters of tumor cells in lymphatic vessels).A clinical-grade tool was trained in a short time using fast-track development principles and validated using the largest international, multi-institutional, multiscanner cohort of cases to date, showing very high precision for LNM detection in CRC. We are releasing a part of the test data sets to facilitate academic research.

AI Driven Applications for Precision Weed Management in Agricultural Crops

Agriculture plays a most important role in our Indian economy and therefore lowering the cost of production and improving the quality of agricultural products is highly demanded. A weed is a plant which grows in wrong place at the wrong time and doing mostly harm than crops. Weed competes with the crops for water, light, nutrients and space, and therefore it prevents crop yields. This paper proposes a new method in a contrary way, which combines deep learning and image processing technology to prevent these weeds. Machine learning technologies, are becoming crucial in agriculture to increase productivity, where advanced automation and control have been required. Based on large training datasets and pre-trained models, (Deep Learning) DL-based Convolutional Neural Networks (CNN) methods have proven to be more accurate than previous traditional techniques. Recently, Deep Learning (DL) has gained much attention due to its advantages in object detection, classification, and feature extraction. The system implementation of image processing technique for weed detection, a trained image is taken as a sample in order to demonstrate the difference between weed and the crop. Yolo frame work is used for annotate boundary boxes to the image with datasets. The effectiveness of the (You Only Live Once) YOLO-WEED system for real-time Unmanned Aerial Vehicle (UAV) weed detection, given its high speed and high accuracy in detection. After certain steps, we get desired output, where the weeds are separated from the crop that has been taken in the sample image. Key Words: DeepLearning,CNN, Unmanned Aerial Vehicle (UAV), Precision Weed Detection.

Deep Learning Algorithm for Real-time Disease Detection and Classification in Radiological Images with Enhanced Diagnostic Accuracy

Radiological imaging plays a pivotal role in disease diagnosis, but the manual interpretation is prone to errors and time-consuming. This research explores the utilization of deep learning algorithms for real-time disease detection and classification in radiological images, aiming to enhance diagnostic accuracy and efficiency in healthcare settings. Deep learning algorithms offer a promising solution by automating the detection and classification process, potentially reducing diagnosis time and improving patient outcomes. However, deploying deep learning algorithms for real-time disease detection in radiological images presents several challenges. These challenges also consist of the need for diverse and large datasets for model training, addressing class imbalance and data variability, ensuring robustness to noise and artifacts, and interpreting model decisions for clinical validation. To overcome these challenges, this research proposes several methods, including data augmentation techniques for increasing dataset diversity, transfer learning from pre-trained models for leveraging existing knowledge, ensemble learning for combining multiple models for improved performance, and attention mechanisms to focus on relevant image regions. Additionally, techniques for uncertainty estimation and model interpretability are explored to enhance trust and acceptance of automated diagnosis systems in clinical practice. By addressing these challenges and implementing appropriate methods, deep learning algorithms show promise for real-time disease detection and classification in radiological images, offering a transformative approach to medical imaging analysis. Results show an accuracy of 97%, sensitivity and specificity at 97% and 95%, and an F1 score of 96%.

Deep Learning Algorithm for Real-time Disease Detection and Classification in Radiological Images with Enhanced Diagnostic Accuracy

Radiological imaging plays a pivotal role in disease diagnosis, but the manual interpretation is prone to errors and time-consuming. This research explores the utilization of deep learning algorithms for real-time disease detection and classification in radiological images, aiming to enhance diagnostic accuracy and efficiency in healthcare settings. Deep learning algorithms offer a promising solution by automating the detection and classification process, potentially reducing diagnosis time and improving patient outcomes. However, deploying deep learning algorithms for real-time disease detection in radiological images presents several challenges. These challenges also consist of the need for diverse and large datasets for model training, addressing class imbalance and data variability, ensuring robustness to noise and artifacts, and interpreting model decisions for clinical validation. To overcome these challenges, this research proposes several methods, including data augmentation techniques for increasing dataset diversity, transfer learning from pre-trained models for leveraging existing knowledge, ensemble learning for combining multiple models for improved performance, and attention mechanisms to focus on relevant image regions. Additionally, techniques for uncertainty estimation and model interpretability are explored to enhance trust and acceptance of automated diagnosis systems in clinical practice. By addressing these challenges and implementing appropriate methods, deep learning algorithms show promise for real-time disease detection and classification in radiological images, offering a transformative approach to medical imaging analysis. Results show an accuracy of 97%, sensitivity and specificity at 97% and 95%, and an F1 score of 96%.

A Generic Machine Learning Model for Spatial Query Optimization based on Spatial Embeddings

Machine learning (ML) and deep learning (DL) techniques are increasingly applied to produce efficient query optimizers, in particular in regards to big data systems. The optimization of spatial operations is even more challenging due to the inherent complexity of such kind of operations, like spatial join or range query, and the peculiarities of spatial data. Although a few ML-based spatial query optimizers have been proposed in literature, their design limits their use, since each one is tailored for a specific collection of datasets, a specific operation, or a specific hardware setting. Changes to any of these will require building and training a completely new model which entails collecting a new very large training dataset to obtain a good model. This paper proposes a different approach which exploits the use of the novel notion of spatial embedding to overcome these limitations. In particular, a preliminary model is defined which captures the relevant features of spatial datasets, independently from the operation to be optimized and in an unsupervised manner. This model is trained with a large amount of both synthetic and real-world data, with the aim to produce meaningful spatial embeddings. The construction of an embedding model could be intended as a preliminary step for the optimization of many different spatial operations, so the cost of its building can be compensated during the subsequent construction of specific models. Indeed, for each considered spatial operation, a specific tailored model will be trained but by using spatial embeddings as input, so a very little amount of training data points is required for them. Three peculiar operations are considered as proof of concept in this paper: range query, self-join, and binary spatial join. Finally, a comparison with an alternative technique, known as transfer learning, is provided and the advantages of the proposed technique over it are highlighted.

Modification of a convolutional neural network for the weave pattern classification

The fabric quality in textile industry is characterized by the texture (weave pattern) as it plays a vital role for the production and design of best quality fabric. The earlier proposed automated weave identification methods based on image processing techniques are highly dependent on the lighting conditions. The machine learning methods have been reported to show better accuracy. However, they require very large training datasets, very high processing power and computation time. This study proposes improved accuracy with smaller dataset and reduced computation time by proposing a modification of VGG16 model by adding two additional pooling layers. Using evaluation metrics of both models, the modified model results were analysed according to accuracy, balanced accuracy, and F1-score. On the basis of investigational outcomes, a comparison has been performed with earlier work. The results show that the proposed VGG-16 model is capable to achieve state-of-the-art accuracy and avoid unnecessary activation features by freezing the main convolutional base layers. Ultimately, as evidenced by the performance of the modified VGG-16 deep learning model, the proposed method demonstrated improved accuracy. The study results show that the proposed modified VGG16 algorithm is able to recognize the features of provided database with 90% accuracy and F1-Score ranging from 0.8 to1.

Synthetic Iris Images: A Comparative Analysis between Cartesian and Polar Representation.

In recent years, the advancement of generative techniques, particularly generative adversarial networks (GANs), has opened new possibilities for generating synthetic biometric data from different modalities, including-among others-images of irises, fingerprints, or faces in different representations. This study presents the process of generating synthetic images of human irises, using the recent StyleGAN3 model. The novelty presented in this work consists in producing generated content in both Cartesian and polar coordinate representations, typically used in iris recognition pipelines, such as the foundational work proposed by John Daugman, but hitherto not used in generative AI experiments. The main objective of this study was to conduct a qualitative analysis of the synthetic samples and evaluate the iris texture density and suitability for meaningful feature extraction. During this study, a total of 1327 unique irises were generated, and experimental results carried out using the well-known OSIRIS open-source iris recognition software and the equivalent software, wordlcoin-openiris, newly published at the end of 2023 to prove that (1) no "identity leak" from the training set was observed, and (2) the generated irises had enough unique textural information to be successfully differentiated between both themselves and between them and real, authentic iris samples. The results of our research demonstrate the promising potential of synthetic iris data generation as a valuable tool for augmenting training datasets and improving the overall performance of iris recognition systems. By exploring the synthetic data in both Cartesian and polar representations, we aim to understand the benefits and limitations of each approach and their implications for biometric applications. The findings suggest that synthetic iris data can significantly contribute to the advancement of iris recognition technology, enhancing its accuracy and robustness in real-world scenarios by greatly augmenting the possibilities to gather large and diversified training datasets.

Monitoring manufacturing systems using AI: A method based on a digital factory twin to train CNNs on synthetic data

Modern cyber–physical production systems provide advanced solutions to enhance factory throughput and efficiency. However, monitoring its behaviour and performance becomes challenging as the complexity of a manufacturing system increases. Artificial Intelligence (AI) provides techniques to manage not only decision-making tasks but also to support monitoring. The integration of AI into a factory can be facilitated by a reliable Digital Twin (DT) that enables knowledge-based and data-driven approaches. While computer vision and convolutional neural networks (CNNs) are crucial for monitoring production systems, the need for extensive training data hinders their adoption in real factories. The proposed methodology leverages the Digital Twin of a factory to generate labelled synthetic data for training CNN-based object detection models. Regarding their position and state, the focus is on monitoring entities in manufacturing systems, such as parts, components, fixtures, and tools. This approach reduces the need for large training datasets and enables training when the actual system is unavailable. The trained CNN model is evaluated in various scenarios, with a real case study involving an industrial pilot plant for repairing and recycling Printed Circuit Boards (PCBs).

TinyMSI: A cost-effective handheld device for non-contact diabetic wound monitoring

Devices characterizing diabetic foot ulcers and other wounds currently fall into two categories. Expensive clinically-oriented devices that use mature technologies such as X-ray CT and hyperspectral imaging or low-cost solutions that leverage deep learning to infer wound characterization from conventional smartphone camera images or simple surrogate markers. Mature medical-grade devices are too expensive for primary care and assisted living facilities. Low-cost solutions rely too much on indirect statistical inference to be clinically suitable. Therefore, we propose a device that leverages mature, clinically suitable optical technologies to provide a solution for these facilities. Recognizing that individual combinations of 1–2 bands of active illumination are used individually to capture pulsation, vascular, and oxygenation images. We combine all these bands into a single multispectral lighting source to create a multi-functional, reliable device for wound assessment. We selected these bands to leverage CMOS cameras near orthogonality between the RGB channels and leverage that CMOS cameras can also sense near IR light if a filter is not present, reducing overall system complexity and needed bands. For each function, the necessary lights are turned on, and the captured raw video is then fed to the corresponding sequence of image processing steps. No deep learning models are used, so large training datasets are not required. Our device is also small, lightweight, and handheld.

Structurally constrained and pathology-aware convolutional transformer generative adversarial network for virtual histology staining of human coronary optical coherence tomography images.

There is a significant need for the generation of virtual histological information from coronary optical coherence tomography (OCT) images to better guide the treatment of coronary artery disease (CAD). However, existing methods either require a large pixel-wise paired training dataset or have limited capability to map pathological regions. The aim of this work is to generate virtual histological information from coronary OCT images, without a pixel-wise paired training dataset while capable of providing pathological patterns. We design a structurally constrained, pathology-aware, transformer generative adversarial network, namely structurally constrained pathology-aware convolutional transformer generative adversarial network (SCPAT-GAN), to generate virtual stained H&E histology from OCT images. We quantitatively evaluate the quality of virtual stained histology images by measuring the Fréchet inception distance (FID) and perceptual hash value (PHV). Moreover, we invite experienced pathologists to evaluate the virtual stained images. Furthermore, we visually inspect the virtual stained image generated by SCPAT-GAN. Also, we perform an ablation study to validate the design of the proposed SCPAT-GAN. Finally, we demonstrate 3D virtual stained histology images. Compared to previous research, the proposed SCPAT-GAN achieves better FID and PHV scores. The visual inspection suggests that the virtual histology images generated by SCPAT-GAN resemble both normal and pathological features without artifacts. As confirmed by the pathologists, the virtual stained images have good quality compared to real histology images. The ablation study confirms the effectiveness of the combination of proposed pathological awareness and structural constraining modules. The proposed SCPAT-GAN is the first to demonstrate the feasibility of generating both normal and pathological patterns without pixel-wisely supervised training. We expect the SCPAT-GAN to assist in the clinical evaluation of treating the CAD by providing 2D and 3D histopathological visualizations.

DDViT: Double-Level Fusion Domain Adapter Vision Transformer (Student Abstract)

With the help of Vision transformers (ViTs), medical image segmentation was able to achieve outstanding performance. In particular, they overcome the limitation of convolutional neural networks (CNNs) which rely on local receptive fields. ViTs use self-attention mechanisms to consider relationships between all image pixels or patches simultaneously. However, they require large datasets for training and did not perform well on capturing low-level features. To that end, we propose DDViT, a novel ViT model that unites a CNN to alleviate data-hunger for medical image segmentation with two multi-scale feature representations. Significantly, our approach incorporates a ViT with a plug-in domain adapter (DA) with Double-Level Fusion (DLF) technique, complemented by a mutual knowledge distillation paradigm, facilitating the seamless exchange of knowledge between a universal network and specialized domain-specific network branches. The DLF framework plays a pivotal role in our encoder-decoder architecture, combining the innovation of the TransFuse module with a robust CNN-based encoder. Extensive experimentation across diverse medical image segmentation datasets underscores the remarkable efficacy of DDViT when compared to alternative approaches based on CNNs and Transformer-based models.

Few-Shot Learning via Repurposing Ensemble of Black-Box Models

This paper investigates the problem of exploiting existing solution models of previous tasks to address a related target task with limited training data. Existing approaches addressing this problem often require access to the internal parameterization of the existing solution models and possibly their training data, which is not possible in many practical settings. To relax this requirement, We approach this problem from a new perspective of black-box re-purposing, which augments the target inputs and leverages their corresponding outputs generated by existing black-box APIs into a feature ensemble. We hypothesize that such feature ensemble can be learned to incorporate and encode relevant black-box knowledge into the feature representation of target data, which will compensate for their scarcity. This hypothesis is confirmed via the reported successes of our proposed black-box ensemble in solving multiple few-shot learning tasks derived from various benchmark datasets. All reported results show consistently that the set of heterogeneous black-box solutions of previous tasks can indeed be reused and combined effectively to solve a reasonably related target task without requiring access to a large training dataset. This is the first step towards enabling new possibilities to further supplement existing techniques in transfer or meta learning with black-box knowledge.

Deep learning-based automatic pipeline for 3D needle localization on intra-procedural 3D MRI.

Accurate and rapid needle localization on 3D magnetic resonance imaging (MRI) is critical for MRI-guided percutaneous interventions. The current workflow requires manual needle localization on 3D MRI, which is time-consuming and cumbersome. Automatic methods using 2D deep learning networks for needle segmentation require manual image plane localization, while 3D networks are challenged by the need for sufficient training datasets. This work aimed to develop an automatic deep learning-based pipeline for accurate and rapid 3D needle localization on in vivo intra-procedural 3D MRI using a limited training dataset. The proposed automatic pipeline adopted Shifted Window (Swin) Transformers and employed a coarse-to-fine segmentation strategy: (1) initial 3D needle feature segmentation with 3D Swin UNEt TRansfomer (UNETR); (2) generation of a 2D reformatted image containing the needle feature; (3) fine 2D needle feature segmentation with 2D Swin Transformer and calculation of 3D needle tip position and axis orientation. Pre-training and data augmentation were performed to improve network training. The pipeline was evaluated via cross-validation with 49 in vivo intra-procedural 3D MR images from preclinical pig experiments. The needle tip and axis localization errors were compared with human intra-reader variation using the Wilcoxon signed rank test, with p < 0.05 considered significant. The average end-to-end computational time for the pipeline was 6s per 3D volume. The median Dice scores of the 3D Swin UNETR and 2D Swin Transformer in the pipeline were 0.80 and 0.93, respectively. The median 3D needle tip and axis localization errors were 1.48mm (1.09 pixels) and 0.98°, respectively. Needle tip localization errors were significantly smaller than human intra-reader variation (median 1.70mm; p < 0.01). The proposed automatic pipeline achieved rapid pixel-level 3D needle localization on intra-procedural 3D MRI without requiring a large 3D training dataset and has the potential to assist MRI-guided percutaneous interventions.

Comparing machine learning potentials for water: Kernel-based regression and Behler-Parrinello neural networks.

In this paper, we investigate the performance of different machine learning potentials (MLPs) in predicting key thermodynamic properties of water using RPBE + D3. Specifically, we scrutinize kernel-based regression and high-dimensional neural networks trained on a highly accurate dataset consisting of about 1500 structures, as well as a smaller dataset, about half the size, obtained using only on-the-fly learning. This study reveals that despite minor differences between the MLPs, their agreement on observables such as the diffusion constant and pair-correlation functions is excellent, especially for the large training dataset. Variations in the predicted density isobars, albeit somewhat larger, are also acceptable, particularly given the errors inherent to approximate density functional theory. Overall, this study emphasizes the relevance of the database over the fitting method. Finally, this study underscores the limitations of root mean square errors and the need for comprehensive testing, advocating the use of multiple MLPs for enhanced certainty, particularly when simulating complex thermodynamic properties that may not be fully captured by simpler tests.