OBJECTIVE: Convolutional Neural Networks (CNNs) have emerged as transformative tools in the field of radiation oncology, significantly advancing the precision of contouring practices. However, the adaptability of these algorithms across diverse scanners, institutions, and imaging protocols remains a considerable obstacle. This study aims to investigate the effects of incorporating institution-specific datasets into the training regimen of CNNs to assess their generalization ability in real-world clinical environments. Focusing on a data-centric analysis, the influence of varying multi- and single center training approaches on algorithm performance is conducted.
METHODS: nnU-Net is trained using a dataset comprising 161 18F-PSMA-1007 PET images collected from four distinct institutions (Freiburg: n = 96, Munich: n = 19, Cyprus: n = 32, Dresden: n = 14). The dataset is partitioned such that data from each center are systematically excluded from training and used solely for testing to assess the model\'s generalizability and adaptability to data from unfamiliar sources. Performance is compared through a 5-Fold Cross-Validation, providing a detailed comparison between models trained on datasets from single centers to those trained on aggregated multi-center datasets. Dice Similarity Score, Hausdorff distance and volumetric analysis are used as primary evaluation metrics.
RESULTS: The mixed training approach yielded a median DSC of 0.76 (IQR: 0.64-0.84) in a five-fold cross-validation, showing no significant differences (p = 0.18) compared to models trained with data exclusion from each center, which performed with a median DSC of 0.74 (IQR: 0.56-0.86). Significant performance improvements regarding multi-center training were observed for the Dresden cohort (multi-center median DSC 0.71, IQR: 0.58-0.80 vs. single-center 0.68, IQR: 0.50-0.80, p < 0.001) and Cyprus cohort (multi-center 0.74, IQR: 0.62-0.83 vs. single-center 0.72, IQR: 0.54-0.82, p < 0.01). While Munich and Freiburg also showed performance improvements with multi-center training, results showed no statistical significance (Munich: multi-center DSC 0.74, IQR: 0.60-0.80 vs. single-center 0.72, IQR: 0.59-0.82, p > 0.05; Freiburg: multi-center 0.78, IQR: 0.53-0.87 vs. single-center 0.71, IQR: 0.53-0.83, p = 0.23).
CONCLUSIONS: CNNs trained for auto contouring intraprostatic GTV in 18F-PSMA-1007 PET on a diverse dataset from multiple centers mostly generalize well to unseen data from other centers. Training on a multicentric dataset can improve performance compared to training exclusively with a single-center dataset regarding intraprostatic 18F-PSMA-1007 PET GTV segmentation. The segmentation performance of the same CNN can vary depending on the dataset employed for training and testing.

Advancements in optical coherence control1-5 have unlocked many cutting-edge applications, including long-haul communication, light detection and ranging (LiDAR) and optical coherence tomography6-8. Prevailing wisdom suggests that using more coherent light sources leads to enhanced system performance and device functionalities9-11. Our study introduces a photonic convolutional processing system that takes advantage of partially coherent light to boost computing parallelism without substantially sacrificing accuracy, potentially enabling larger-size photonic tensor cores. The reduction of the degree of coherence optimizes bandwidth use in the photonic convolutional processing system. This breakthrough challenges the traditional belief that coherence is essential or even advantageous in integrated photonic accelerators, thereby enabling the use of light sources with less rigorous feedback control and thermal-management requirements for high-throughput photonic computing. Here we demonstrate such a system in two photonic platforms for computing applications: a photonic tensor core using phase-change-material photonic memories that delivers parallel convolution operations to classify the gaits of ten patients with Parkinson\'s disease with 92.2% accuracy (92.7% theoretically) and a silicon photonic tensor core with embedded electro-absorption modulators (EAMs) to facilitate 0.108 tera operations per second (TOPS) convolutional processing for classifying the Modified National Institute of Standards and Technology (MNIST) handwritten digits dataset with 92.4% accuracy (95.0% theoretically).

When a disaster occurs, the authority must prioritise two things. First, the search and rescue of lives, and second, the identification and management of deceased individuals. However, with thousands of dead bodies to be individually identified in mass disasters, forensic teams face challenges such as long working hours resulting in a delayed identification process and a public health concern caused by the decomposition of the body. Using dental panoramic imaging, teeth have been used in forensics as a physical marker to estimate the age of an individual. Traditionally, dental age estimation has been performed manually by experts. Although the procedure is fairly simple, the large number of victims and the limited amount of time available to complete the assessment during large-scale disasters make forensic work even more challenging. The emergence of artificial intelligence (AI) in the fields of medicine and dentistry has led to the suggestion of automating the current process as an alternative to the conventional method. This study aims to test the accuracy and performance of the developed deep convolutional neural network system for age estimation in large, out-of-sample Malaysian children dataset using digital dental panoramic imaging. Forensic Dental Estimation Lab (F-DentEst Lab) is a computer application developed to perform the dental age estimation digitally. The introduction of this system is to improve the conventional method of age estimation that significantly increase the efficiency of the age estimation process based on the AI approach. A total number of one-thousand-eight-hundred-and-ninety-two digital dental panoramic images were retrospectively collected to test the F-DentEst Lab. Data training, validation, and testing have been conducted in the early stage of the development of F-DentEst Lab, where the allocation involved 80 % training and the remaining 20 % for testing. The methodology was comprised of four major steps: image preprocessing, which adheres to the inclusion criteria for panoramic dental imaging, segmentation, and classification of mandibular premolars using the Dynamic Programming-Active Contour (DP-AC) method and Deep Convolutional Neural Network (DCNN), respectively, and statistical analysis. The suggested DCNN approach underestimated chronological age with a small ME of 0.03 and 0.05 for females and males, respectively.

暂无摘要。

Meiotic errors of relatively small chromosomes in oocytes result in egg aneuploidies that cause miscarriages and congenital diseases. Unlike somatic cells, which preferentially mis-segregate larger chromosomes, aged oocytes preferentially mis-segregate smaller chromosomes through unclear processes. Here, we provide a comprehensive three-dimensional chromosome identifying-and-tracking dataset throughout meiosis I in live mouse oocytes. This analysis reveals a prometaphase pathway that actively moves smaller chromosomes to the inner region of the metaphase plate. In the inner region, chromosomes are pulled by stronger bipolar microtubule forces, which facilitates premature chromosome separation, a major cause of segregation errors in aged oocytes. This study reveals a spatial pathway that facilitates aneuploidy of small chromosomes preferentially in aged eggs and implicates the role of the M phase in creating a chromosome size-based spatial arrangement.

暂无摘要。

Genetic variation that influences gene expression and splicing is a key source of phenotypic diversity1-5. Although invaluable, studies investigating these links in humans have been strongly biased towards participants of European ancestries, which constrains generalizability and hinders evolutionary research. Here to address these limitations, we developed MAGE, an open-access RNA sequencing dataset of lymphoblastoid cell lines from 731 individuals from the 1000 Genomes Project6, spread across 5 continental groups and 26 populations. Most variation in gene expression (92%) and splicing (95%) was distributed within versus between populations, which mirrored the variation in DNA sequence. We mapped associations between genetic variants and expression and splicing of nearby genes (cis-expression quantitative trait loci (eQTLs) and cis-splicing QTLs (sQTLs), respectively). We identified more than 15,000 putatively causal eQTLs and more than 16,000 putatively causal sQTLs that are enriched for relevant epigenomic signatures. These include 1,310 eQTLs and 1,657 sQTLs that are largely private to underrepresented populations. Our data further indicate that the magnitude and direction of causal eQTL effects are highly consistent across populations. Moreover, the apparent \'population-specific\' effects observed in previous studies were largely driven by low resolution or additional independent eQTLs of the same genes that were not detected. Together, our study expands our understanding of human gene expression diversity and provides an inclusive resource for studying the evolution and function of human genomes.

Gene expression in Arabidopsis is regulated by more than 1,900 transcription factors (TFs), which have been identified genome-wide by the presence of well-conserved DNA-binding domains. Activator TFs contain activation domains (ADs) that recruit coactivator complexes; however, for nearly all Arabidopsis TFs, we lack knowledge about the presence, location and transcriptional strength of their ADs1. To address this gap, here we use a yeast library approach to experimentally identify Arabidopsis ADs on a proteome-wide scale, and find that more than half of the Arabidopsis TFs contain an AD. We annotate 1,553 ADs, the vast majority of which are, to our knowledge, previously unknown. Using the dataset generated, we develop a neural network to accurately predict ADs and to identify sequence features that are necessary to recruit coactivator complexes. We uncover six distinct combinations of sequence features that result in activation activity, providing a framework to interrogate the subfunctionalization of ADs. Furthermore, we identify ADs in the ancient AUXIN RESPONSE FACTOR family of TFs, revealing that AD positioning is conserved in distinct clades. Our findings provide a deep resource for understanding transcriptional activation, a framework for examining function in intrinsically disordered regions and a predictive model of ADs.

BACKGROUND: Global longitudinal strain (GLS) is reported to be more reproducible and prognostic than ejection fraction. Automated, transparent methods may increase trust and uptake.
OBJECTIVE: The authors developed open machine-learning-based GLS methodology and validate it using multiexpert consensus from the Unity UK Echocardiography AI Collaborative.
METHODS: We trained a multi-image neural network (Unity-GLS) to identify annulus, apex, and endocardial curve on 6,819 apical 4-, 2-, and 3-chamber images. The external validation dataset comprised those 3 views from 100 echocardiograms. End-systolic and -diastolic frames were each labelled by 11 experts to form consensus tracings and points. They also ordered the echocardiograms by visual grading of longitudinal function. One expert calculated global strain using 2 proprietary packages.
RESULTS: The median GLS, averaged across the 11 individual experts, was -16.1 (IQR: -19.3 to -12.5). Using each case\'s expert consensus measurement as the reference standard, individual expert measurements had a median absolute error of 2.00 GLS units. In comparison, the errors of the machine methods were: Unity-GLS 1.3, proprietary A 2.5, proprietary B 2.2. The correlations with the expert consensus values were for individual experts 0.85, Unity-GLS 0.91, proprietary A 0.73, proprietary B 0.79. Using the multiexpert visual ranking as the reference, individual expert strain measurements found a median rank correlation of 0.72, Unity-GLS 0.77, proprietary A 0.70, and proprietary B 0.74.
CONCLUSIONS: Our open-source approach to calculating GLS agrees with experts\' consensus as strongly as the individual expert measurements and proprietary machine solutions. The training data, code, and trained networks are freely available online.

暂无摘要。