Ki-67 evaluation using deep-learning model-assisted digital image analysis in breast cancer.

AIMS: To test the efficacy of artificial intelligence (AI)-assisted Ki-67 digital image analysis in invasive breast carcinoma (IBC) with quantitative assessment of AI model performance. METHODS AND RESULTS: This study used 494 cases of Ki-67 slide images of IBC core needle biopsies. The methods were divided into two steps: (i) construction of a deep-learning model (DL); and (ii) DL implementation for Ki-67 analysis. First, a DL tissue classifier model (DL-TC) and a DL nuclear detection model (DL-ND) were constructed using HALO AI DenseNet V2 algorithm with 31,924 annotations in 300 Ki-67 digital slide images. Whether the class predicted by DL-TC in the test set was correct compared with the annotation of ground truth at the pixel level was evaluated. Second, DL-TC- and DL-ND-assisted digital image analysis (DL-DIA) was performed in the other 194 luminal-type cases and correlations with manual counting and clinical outcome were investigated to confirm the accuracy and prognostic potential of DL-DIA. The performance of DL-TC was excellent and invasive carcinoma nests were well segmented from other elements (average precision: 0.851; recall: 0.878; F1-score: 0.858). Ki-67 index data and the number of nuclei from DL-DIA were positively correlated with data from manual counting (ρ = 0.961, and 0.928, respectively). High Ki-67 index (cutoff 20%) cases showed significantly worse recurrence-free survival and breast cancer-specific survival (P = 0.024, and 0.032, respectively). CONCLUSION: The performances of DL-TC and DL-ND were excellent. DL-DIA demonstrated a high degree of concordance with manual counting of Ki-67 and the results of this approach have prognostic potential.

Automatic feature selection for performing Unit 2 of vault in wheel gymnastics.

We propose a framework to analyze the relationship between the movement features of a wheel gymnast around the mounting phase of Unit 2 of the vault event and execution (E-score) deductions from a machine-learning perspective. We first developed an automation system from a video of a wheel gymnast performing a tuck-front somersault to extract the four frames highlighting its Unit 2 performance of the vault event, such as take-off, pike-mount, the starting point of time on the wheel, and final position before the thrust. We implemented this automation using recurrent all-pairs field transforms (RAFT) and XMem, i.e., deep network architectures respectively for optical flow estimation and video object segmentation. We then used a markerless pose-estimation system called OpenPose to acquire the coordinates of the gymnast's body joints, such as shoulders, hips, and knees then calculate the joint angles at the extracted video frames. Finally, we constructed a regression model to estimate the E-score deductions during Unit 2 on the basis of the joint angles using an ensemble learning algorithm called Random Forests, with which we could automatically select a small number of features with the nonzero values of feature importances. By applying our framework of markerless motion analysis to videos of male wheel gymnasts performing the vault, we achieved precise estimation of the E-score deductions during Unit 2 with a determination coefficient of 0.79. We found the two movement features of particular importance for them to avoid significant deductions: time on the wheel and angles of knees at the pike-mount position. The selected features well reflected the maturity of the gymnast's skills related to the motions of riding the wheel, easily noticeable to the judges, and their branching conditions were almost consistent with the general vault regulations.

PCA-based unsupervised feature extraction for gene expression analysis of COVID-19 patients.

Coronavirus disease 2019 (COVID-19) is raging worldwide. This potentially fatal infectious disease is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). However, the complete mechanism of COVID-19 is not well understood. Therefore, we analyzed gene expression profiles of COVID-19 patients to identify disease-related genes through an innovative machine learning method that enables a data-driven strategy for gene selection from a data set with a small number of samples and many candidates. Principal-component-analysis-based unsupervised feature extraction (PCAUFE) was applied to the RNA expression profiles of 16 COVID-19 patients and 18 healthy control subjects. The results identified 123 genes as critical for COVID-19 progression from 60,683 candidate probes, including immune-related genes. The 123 genes were enriched in binding sites for transcription factors NFKB1 and RELA, which are involved in various biological phenomena such as immune response and cell survival: the primary mediator of canonical nuclear factor-kappa B (NF-κB) activity is the heterodimer RelA-p50. The genes were also enriched in histone modification H3K36me3, and they largely overlapped the target genes of NFKB1 and RELA. We found that the overlapping genes were downregulated in COVID-19 patients. These results suggest that canonical NF-κB activity was suppressed by H3K36me3 in COVID-19 patient blood.

Domain knowledge integration into deep learning for typhoon intensity classification.

In this report, we propose a deep learning technique for high-accuracy estimation of the intensity class of a typhoon from a single satellite image, by incorporating meteorological domain knowledge. By using the Visual Geometric Group's model, VGG-16, with images preprocessed with fisheye distortion, which enhances a typhoon's eye, eyewall, and cloud distribution, we achieved much higher classification accuracy than that of a previous study, even with sequential-split validation. Through comparison of t-distributed stochastic neighbor embedding (t-SNE) plots for the feature maps of VGG with the original satellite images, we also verified that the fisheye preprocessing facilitated cluster formation, suggesting that our model could successfully extract image features related to the typhoon intensity class. Moreover, gradient-weighted class activation mapping (Grad-CAM) was applied to highlight the eye and the cloud distributions surrounding the eye, which are important regions for intensity classification; the results suggest that our model qualitatively gained a viewpoint similar to that of domain experts. A series of analyses revealed that the data-driven approach using only deep learning has limitations, and the integration of domain knowledge could bring new breakthroughs.

Influences of Phenyl Rings on NHC Ligands with Bicyclic Architectures.

In addition to phosphanes, olefins, amines, and amides, over the past two decades N-heterocyclic carbene (NHC) has emerged as a useful alternative ligand. Based on a number of derivatization studies on NHC ligands, imidazol-2-ylidene and imidazolin-2-ylidene became the standard heterocyclic form, and bulky substituents have commonly been introduced on the nitrogen(s) adjacent to carbenic carbons. Our group previously developed NHCs equipped with noncarbenic carbons with a bicyclic architecture that gives them unique steric properties that make them bulky but accessible. In this study, we synthesized a novel type of NHC ligand that possesses a bicyclo[2.2.1]heptane architecture, and we compared five derivatives using copper-catalyzed allylic arylations with aryl Grignard reagents. The regioselectivity of the substitution obviously indicates that a phenyl ring over an active site has a characteristic effect on the resultant copper catalysts when γ-substitution is the major pathway.

Change in hippocampal theta oscillation associated with multiple lever presses in a bimanual two-lever choice task for robot control in rats.

Hippocampal theta oscillations have been implicated in working memory and attentional process, which might be useful for the brain-machine interface (BMI). To further elucidate the properties of the hippocampal theta oscillations that can be used in BMI, we investigated hippocampal theta oscillations during a two-lever choice task. During the task body-restrained rats were trained with a food reward to move an e-puck robot towards them by pressing the correct lever, ipsilateral to the robot several times, using the ipsilateral forelimb. The robot carried food and moved along a semicircle track set in front of the rat. We demonstrated that the power of hippocampal theta oscillations gradually increased during a 6-s preparatory period before the start of multiple lever pressing, irrespective of whether the correct lever choice or forelimb side were used. In addition, there was a significant difference in the theta power after the first choice, between correct and incorrect trials. During the correct trials the theta power was highest during the first lever-releasing period, whereas in the incorrect trials it occurred during the second correct lever-pressing period. We also analyzed the hippocampal theta oscillations at the termination of multiple lever pressing during the correct trials. Irrespective of whether the correct forelimb side was used, the power of hippocampal theta oscillations gradually decreased with the termination of multiple lever pressing. The frequency of theta oscillation also demonstrated an increase and decrease, before and after multiple lever pressing, respectively. There was a transient increase in frequency after the first lever press during the incorrect trials, while no such increase was observed during the correct trials. These results suggested that hippocampal theta oscillations reflect some aspects of preparatory and cognitive neural activities during the robot controlling task, which could be used for BMI.

Optimal design for hetero-associative memory: hippocampal CA1 phase response curve and spike-timing-dependent plasticity.

Recently reported experimental findings suggest that the hippocampal CA1 network stores spatio-temporal spike patterns and retrieves temporally reversed and spread-out patterns. In this paper, we explore the idea that the properties of the neural interactions and the synaptic plasticity rule in the CA1 network enable it to function as a hetero-associative memory recalling such reversed and spread-out spike patterns. In line with Lengyel's speculation (Lengyel et al., 2005), we firstly derive optimally designed spike-timing-dependent plasticity (STDP) rules that are matched to neural interactions formalized in terms of phase response curves (PRCs) for performing the hetero-associative memory function. By maximizing object functions formulated in terms of mutual information for evaluating memory retrieval performance, we search for STDP window functions that are optimal for retrieval of normal and doubly spread-out patterns under the constraint that the PRCs are those of CA1 pyramidal neurons. The system, which can retrieve normal and doubly spread-out patterns, can also retrieve reversed patterns with the same quality. Finally, we demonstrate that purposely designed STDP window functions qualitatively conform to typical ones found in CA1 pyramidal neurons.