Activation Mapping Research Articles

CT-Net: an interpretable CNN-Transformer fusion network for fNIRS classification.

Functional near-infrared spectroscopy (fNIRS), an optical neuroimaging technique, has been widely used in the field of brain activity recognition and brain-computer interface. Existing works have proposed deep learning-based algorithms for the fNIRS classification problem. In this paper, a novel approach based on convolutional neural network and Transformer, named CT-Net, is established to guide the deep modeling for the classification of mental arithmetic (MA) tasks. We explore the effect of data representations, and design a temporal-level combination of two raw chromophore signals to improve the data utilization and enrich the feature learning of the model. We evaluate our model on two open-access datasets and achieve the classification accuracy of 98.05% and 77.61%, respectively. Moreover, we explain our model by the gradient-weighted class activation mapping, which presents a high consistent between the contributing value of features learned by the model and the mapping of brain activity in the MA task. The results suggest the feasibility and interpretability of CT-Net for decoding MA tasks.

Targeting attack activity-driven networks.

Real-world complex systems demonstrated temporal features, i.e., the network topology varies with time and should be described as temporal networks since the traditional static networks cannot accurately characterize. To describe the deliberate attack events in the temporal networks, we propose an activity-based targeted attack on the activity-driven network to investigate temporal networks' temporal percolation properties and resilience. Based on the node activity and network mapping framework, the giant component and temporal percolation threshold are solved according to percolation theory and generating function. The theoretical results coincide with the simulation results near the thresholds. We find that targeted attacks can affect the temporal network, while random attacks cannot. As the probability of a highly active node being deleted increases, the temporal percolation threshold increases, and the giant component increases, thus enhancing robustness. When the network's activity distribution is extremely heterogeneous, network robustness decreases consequently. These findings help us to analyze and understand real-world temporal networks.

Class activation map-based slicing-concatenation and contrastive learning: A novel strategy for record-level atrial fibrillation detection

BackgroundDeep learning-based models for atrial fibrillation (AF) detection require extensive training data, which often necessitates labor-intensive professional annotation. While data augmentation techniques have been employed to mitigate the scarcity of annotated electrocardiogram (ECG) data, specific augmentation methods tailored for recording-level ECG annotations are lacking. This gap hampers the development of robust deep learning models for AF detection. MethodsWe propose a novel strategy, a combination of Class Activation Map-based Slicing-Concatenation (CAM-SC) data augmentation and contrastive learning, to address the current challenges. Initially, a baseline model incorporating a global average pooling layer is trained for classification and to generate class activation maps (CAMs), which highlight indicative ECG segments. After that, in each recording, indicative and non-indicative segments are sliced. These segments are subsequently concatenated randomly based on starting and ending Q points of QRS complexes, with indicative segments preserved to maintain label correctness. Finally, the augmented dataset undergoes contrastive learning to learn general representations, thereby enhancing AF detection performance. ResultsUsing ResNet-101 as the baseline model, training with the augmented data yielded the highest F1-score of 0.861 on the Computing in Cardiology (CinC) Challenge 2017 dataset, a typical AF dataset with recording-level annotations. The metrics outperform most previous studies. ConclusionsThis study introduces an innovative data augmentation method specifically designed for recording-level ECG annotations, significantly enhancing AF detection using deep learning models. This approach has substantial implications for future AF detection research.

Early-stage cardiomegaly detection and classification from X-ray images using convolutional neural networks and transfer learning

Cardiomyopathy is a serious condition that can result in heart failure, sudden cardiac death, malignant arrhythmias, and thromboembolism. It is a significant contributor to morbidity and mortality globally. The initial finding of cardiomegaly on radiological imaging may signal a deterioration of a known heart condition, an unknown heart disease, or a heart complication related to another illness. Further cardiological evaluation is needed to confirm the diagnosis and determine appropriate treatment. A chest radiograph (X-ray) is the main imaging method used to identify cardiomegaly when the heart is enlarged. A prompt and accurate diagnosis is essential to help healthcare providers determine the most appropriate treatment options before the condition worsens. This study aims to utilize convolutional neural networks and transfer learning techniques, specifically Inception, DenseNet-169, and ResNet-50, to classify cardiomegaly from chest X-ray images automatically. The utilization of block-matching and 3D filtering (BM3D) techniques aimed at enhancing image edge retention, decreasing noise, and utilizing contrast limited adaptive histogram equalization (CLAHE) to enhance contrast in low-intensity images. Gradient-weighted Class Activation Mapping (GradCAM) was used to visualize the significant activation regions contributing to the model's decision. After evaluating all the models, the ResNet-50 model showed outstanding performance. It achieved perfect accuracy of 100 % in both training, and validation, and an impressive 99.8 % accuracy in testing. Additionally, it displayed complete 100 % precision, recall, and F1-score. These findings demonstrate that ResNet-50 surpassed all other models in the study. As a result, the impressive performance of the ResNet-50 model suggests that it could be a valuable tool in improving the efficiency and accuracy of cardiomyopathy diagnosis, ultimately leading to better patient outcomes.

Enhanced cancer classification and critical feature visualization using Raman spectroscopy and convolutional neural networks

Cell misuse and cross-contamination pose a significant threat to the accuracy of cell research outcomes, often leading to the wasteful expenditure of time, manpower, and material resources. Consequently, the accurate identification of cell lines is paramount. However, traditional identification methods, which often involve staining and culturing procedures, are not only time-consuming but also laborious. This underscores the need for a novel approach that enables rapid and automated cell line identification, thereby enhancing research efficiency and accuracy. Raman spectroscopy, renowned for its label-free, rapid, and noninvasive nature, offers invaluable molecular insights into samples, making it a widely utilized technique in the biological field. In this study, the identification of one normal and five cancer cell lines was achieved using a sparrow search algorithm-convolutional neural networks (SSA-CNN), considering both the full spectra and fingerprint region perspectives. The SSA-CNN model demonstrated exceptional performance, not just in binary classification, but also in accurately distinguishing among six cell lines. It achieved the highest accuracy (around 95 %), and the lowest standard error (≤3%), for both the full spectra and fingerprint region. Based on the highly accurate SSA-CNN model, proposed the application of gradient-weighted class activation mapping (Grad-CAM) to visualize the Raman feature peaks. Upon comparing the visualized Raman features with reported biomarkers, found that not only were common biomolecules such as glucose, proteins, and liquids visualized, but specific feature peaks also aligned with reported biomarkers. The aforementioned results clearly demonstrated that the proposed strategy not only classifies cancer cell lines with remarkable accuracy but also served as a valuable tool for the discovery of novel biomarkers.

The application of explainable artificial intelligence methods to models for automatic creativity assessment.

The study is devoted to comparing various models based on Artificial Intelligence to determine the level of creativity based on drawings performed using the Urban test, as well as analyzing the results of applying explainable artificial intelligence methods to a trained model to identify the most relevant features in drawings that influence the model's prediction. The dataset is represented by a set of 1,823 scanned forms of drawings of participants performed according to the Urban test. The test results of each participant were assessed by an expert. Preprocessed images were used for fine-tuning pre-trained models such as MobileNet, ResNet18, AlexNet, DenseNet, ResNext, EfficientNet, ViT with additional linear layers to predict the participant's score. Visualization of the areas that are of greatest importance from the point of view of the model was carried out using the Gradient-weighted Class Activation Mapping (Grad-CAM) method. Trained models based on MobileNet showed the highest prediction accuracy rate of 76%. The results of the application of explainable artificial intelligence demonstrated areas of interest that correlated with the criteria for expert assessment according to the Urban test. Analysis of erroneous predictions of the model in terms of interpretation of areas of interest made it possible to clarify the features of the drawing on which the model relies, contrary to the expert. The study demonstrated the possibility of using neural network methods for automated diagnosis of the level of creativity according to the Urban test based on the respondents' drawings. The application of explainable artificial intelligence methods to the trained model demonstrated the compliance of the identified activation zones with the rules of expert assessment according to the Urban test.

Clinical knowledge-guided hybrid classification network for automatic periodontal disease diagnosis in X-ray image

Accurate classification of periodontal disease through panoramic X-ray images carries immense clinical importance for effective diagnosis and treatment. Recent methodologies attempt to classify periodontal diseases from X-ray images by estimating bone loss within these images, supervised by manual radiographic annotations for segmentation or keypoint detection. However, these annotations often lack consistency with the clinical gold standard of probing measurements, potentially causing measurement inaccuracy and leading to unstable classifications. Additionally, the diagnosis of periodontal disease necessitates exceptional sensitivity. To address these challenges, we introduce HC-Net, an innovative hybrid classification framework devised for accurately classifying periodontal disease from X-ray images. This framework comprises three main components: tooth-level classification, patient-level classification, and a learnable adaptive noisy-OR gate. In the tooth-level classification, we initially employ instance segmentation to individually identify each tooth, followed by tooth-level periodontal disease classification. For patient-level classification, we utilize a multi-task strategy to concurrently learn patient-level classification and a Classification Activation Map (CAM) that signifies the confidence of local lesion areas within the panoramic X-ray image. Eventually, our adaptive noisy-OR gate acquires a hybrid classification by amalgamating predictions from both levels. In particular, we incorporate clinical knowledge into the workflows used by professional dentists, targeting the enhanced handling of sensitivity of periodontal disease diagnosis. Extensive empirical testing on a dataset amassed from real-world clinics demonstrates that our proposed HC-Net achieves unparalleled performance in periodontal disease classification, exhibiting substantial potential for practical application.

Utilizing customized CNN for brain tumor prediction with explainable AI

Timely diagnosis of brain tumors using MRI and its potential impact on patient survival are critical issues addressed in this study. Traditional DL models often lack transparency, leading to skepticism among medical experts owing to their "black box" nature. This study addresses this gap by presenting an innovative approach for brain tumor detection. It utilizes a customized Convolutional Neural Network (CNN) model empowered by three advanced explainable artificial intelligence (XAI) techniques: Shapley Additive Explana-tions (SHAP), Local Interpretable Model-agnostic Explanations (LIME), and Gradient-weighted Class Activation Mapping (Grad-CAM). The study utilized the BR35H dataset, which includes 3060 brain MRI images encompassing both tumorous and non-tumorous cases. The proposed model achieved a remarkable training accuracy of 100 % and validation accuracy of 98.67 %. Precision, recall, and F1 score metrics demonstrated exceptional performance at 98.50 %, confirming the accuracy of the model in tumor detection. Detailed result analysis, including a confusion matrix, comparison with existing models, and generalizability tests on other datasets, establishes the superiority of the proposed approach and sets a new benchmark for accuracy. By integrating a customized CNN model with XAI techniques, this research enhances trust in AI-driven medical diagnostics and offers a promising pathway for early tumor detection and potentially life-saving interventions.

Time-guided convolutional neural networks for spatiotemporal urban flood modelling

Urban flood modelling is key to understand flood risks and develop effective interventions in flood management. Deep learning (DL), known for its robust and automatic feature extraction capabilities, has been applied for urban flood predictions. However, the hybrid spatiotemporal structure of conventional DL-enabled urban flood models is limited in terms of accuracy and efficiency. To address this gap, this study develops a new DL model guided by time information. This model uses a classic CNN (Convolution Neural Network) architecture, Unet, as its backbone. Time information is integrated into inputs via an extra channel to specify the desired prediction time, facilitating the simulation of the spatiotemporal flood process. Additionally, a modified loss function is formulated to tackle the sample imbalance problem between flooded and non-flooded sites. The model performance is assessed in an urban area in Dalian, China with a total of 18 rainfall events of varying return periods. The model attains average precision and recall values of 0.90 and 0.81, respectively, across different time steps during various events. Furthermore, the model exhibits transferability in ungauged regions where a high influence of surrounding environments on local flood processes is identified by Grad-CAM (Gradient-weighted Class Activation Mapping) analysis. The results show that the new Unet model has great promise in efficiently providing accurate spatiotemporal flood simulations. The time-guided Unet model can serve as practical tools for rapid flood simulation in urban areas.

Probabilistic and explainable modeling of Phase–Phase Cross-Frequency Coupling patterns in EEG. Application to dyslexia diagnosis

This work explores the intricate neural dynamics associated with dyslexia through the lens of Cross-Frequency Coupling (CFC) analysis applied to electroencephalography (EEG) signals evaluated from 48 seven-year-old Spanish readers from the LEEDUCA research platform. The analysis focuses on CFS (Cross-Frequency phase Synchronization) maps, capturing the interaction between different frequency bands during low-level auditory processing stimuli. Then, making use of Gaussian Mixture Models (GMMs), CFS activations are quantified and classified, offering a compressed representation of EEG activation maps. The study unveils promising results specially at the Theta-Gamma coupling (Area Under the Curve = 0.821), demonstrating the method’s sensitivity to dyslexia-related neural patterns and highlighting potential applications in the early identification of dyslexic individuals.

Multielectrode catheter-induced ectopy mapping: a novel technique for ablation of infrequent premature ventricular contractions

BackgroundAblation of infrequent premature ventricular complexes (PVC) is challenging. ObjectivesA novel mapping strategy for patients with infrequent PVCs, called “multielectrode catheter-induced ectopy mapping” (MECIE-mapping) is described, aiming at performing a hybrid activation/template matching map by taking advantage of multielectrode catheter-induced arrhythmogenicity. MethodsPatients referred to three tertiary centers for PVC ablation were prospectively enrolled if they had infrequent PVCs (less than 1 PVC per minute) at onset of procedure, preventing the realization of an accurate activation map. A detailed MECIE-map was created using the arrhythmogenic property of multielectrode catheters, corresponding to a local activation time (LAT) map generated by annotating LAT from mechanical PVCs. Selecting mechanical PVCs with ≥ 99% concordance with the clinical PVC spotted the site of origin where ablation was delivered. The primary endpoint was long-term success, defined as an > 80% reduction in PVC burden during follow-up. Results29 patients were included, with 25 [IQR 7-30] PVCs in the initial 30 minutes of procedure. During MECIE-mapping, 67 [IQR 1-332] points with ≥ 99% concordance were acquired. The best LAT was 34.0±9.5 ms before QRS onset. Pace-mapping was 97.4±3.1% compared to the clinical PVC. Ablation was locally performed. After 13.2±5.1 months of follow-up, 27 patients (93.1 %) had 80% reduction in PVC burden, and only two patients had a symptomatic recurrence. ConclusionsA detailed MECIE-map taking advantage of multielectrode catheter arrhythmogenicity may be generated to spot the origin of PVCs, a strategy resulting in a good procedural success rate.

A deep learning framework for automated anomaly detection and localization in fused filament fabrication

Under-extrusion poses a common challenge in filament 3D printing, necessitating precise adjustments to printing parameters for optimal resolution. Swift identification of this flaw is crucial for implementing timely corrective measures. In response, we present a novel machine learning approach designed to detect anomalies in 3D printing, specifically in the fused filament fabrication process (FFF). Our framework utilizes “You Only Look Once” (YOLO), a real-time object detection system, and “Visual Geometry Group-16” (VGG-16), a Convolutional Neural Network (CNN) model for image recognition, to accurately identify and localize under-extrusion events. Initially, models such as VGG-16, VGG-19, and ResNet-50 were trained without the inclusion of YOLO to establish baseline accuracies. Subsequently, an automated image pre-processing phase employs YOLO to discern the nozzle head, facilitating subsequent cropping around the region of interest for retraining the models. This inclusion of the nozzle head detection approach significantly improved the accuracy of all models, with the combined application of YOLO and VGG-16 demonstrating the most substantial enhancement, boosting detection accuracy to 97%. Moreover, Gradient-weighted Class Activation Mapping (Grad-CAM) technique, another CNN-based method, is employed to effectively highlight predicted areas in under-extrusion scenarios. While our method significantly advances the automatic detection of printing anomalies, it primarily serves as a diagnostic tool, signaling the need for intervention. Our rigorous testing on images of varying complexities confirms the model’s robustness, evaluated using comprehensive metrics such as precision, recall, and F1 score.

3D-targeted, electrocardiographic imaging-aided stereotactic radioablation for ventricular tachycardia storm: a case report

Abstract Background Stereotactic arrhythmia radioablation (STAR) is a promising non-invasive therapy for patients with ventricular tachycardia (VT). Accurate identification of the arrhythmogenic volume, or clinical target volume (CTV), on the radiotherapy (RT) 4D planning computed tomography (CT) scan is key for STAR efficacy and safety. This case report illustrates our workflow of electro-structural image integration for CTV delineation. Case summary A 72-year-old man with ischaemic cardiomyopathy and VT storm, despite two (endocardial and epicardial) catheter-based ablations, was consented for STAR. A 3D electro-structural arrhythmia model was generated from co-registered electroanatomical voltage and activation maps, electrocardiographic (ECG) imaging, and the cardiac CT angiography scan (in ADAS 3D), pinpointing the VT isthmus and inferoapical VT exit. At this location, an area with short recovery times was found with ECG imaging. A multidisciplinary team delineated the CTV on the transmural ventricular myocardium, which was fused with the 4D planning CT scan using a digital images and communication in medicine (DICOM) radiotherapy file. The CTV was 63% smaller compared with using the conventional American Heart Association 17-segment approach (11 vs. 24 cm3). A single fraction of 25 Gy was delivered to the internal target volume. After an 8-week blanking period, no VT recurrences or radiation-related side-effects were noted. Eight months later, the patient died from end-stage heart failure. Discussion We report a novel workflow for 3D-targeted and ECG imaging-aided CTV delineation for STAR, resulting in a smaller irradiated volume compared with segmental approaches. Acute and intermediate outcome and safety were favourable. Non-invasive ECG imaging at baseline and during induced VT holds promise for STAR guidance.

Classification of psychosis spectrum disorders using graph convolutional networks with structurally constrained functional connectomes

This article considers the problem of classifying individuals in a dataset of diverse psychosis spectrum conditions, including persons with subsyndromal psychotic-like experiences (PLEs) and healthy controls. This task is more challenging than the traditional problem of distinguishing patients with a diagnosed disorder from controls using brain network features, since the neurobiological differences between PLE individuals and healthy persons are less pronounced. Further, examining a transdiagnostic sample compared to controls is concordant with contemporary approaches to understanding the full spectrum of neurobiology of psychoses. We consider both support vector machines (SVMs) and graph convolutional networks (GCNs) for classification, with a variety of edge selection methods for processing the inputs. We also employ the MultiVERSE algorithm to generate network embeddings of the functional and structural networks for each subject, which are used as inputs for the SVMs. The best models among SVMs and GCNs yielded accuracies >63%. Investigation of network connectivity between persons with PLE and controls identified a region within the right inferior parietal cortex, called the PGi, as a central region for communication among modules (network hub). Class activation mapping revealed that the PLE group had salient regions in the dorsolateral prefrontal, orbital and polar frontal cortices, and the lateral temporal cortex, whereas the controls did not. Our study demonstrates the potential usefulness of deep learning methods to distinguish persons with subclinical psychosis and diagnosable disorders from controls. In the long term, this could help improve accuracy and reliability of clinical diagnoses, provide neurobiological bases for making diagnoses, and initiate early intervention strategies.

Detailed analysis of electrogram peak frequency to guide ventricular tachycardia substrate mapping.

Differentiating near-field (NF) and far-field (FF) electrograms (EGMs) is crucial in identifying critical arrhythmogenic substrate during ventricular tachycardia (VT) ablation. A novel algorithm annotates NF-fractionated signals enabling EGM peak frequency (PF) determination using wavelet transformation. This study evaluated the algorithms' effectiveness in identifying critical components of the VT circuit during substrate mapping. A multicentre, international cohort undergoing VT ablation was investigated. VT activation maps were used to demarcate the isthmus zone (IZ). Offline analysis was performed to evaluate the diagnostic performance of low-voltage area (LVA) PF substrate mapping. A total of 30 patients encompassing 198 935 EGMs were included. The IZ PF was significantly higher in sinus rhythm (SR) compared to right ventricular paced (RVp) substrate maps (234 Hz (195-294) vs. 197 Hz (166-220); P = 0.010). Compared to LVA PF, the IZ PF was significantly higher in both SR and RVp substrate maps (area under curve, AUC: 0.74 and 0.70, respectively). The LVA PF threshold of ≥200 Hz was optimal in SR maps (sensitivity 69%; specificity 64%) and RVp maps (sensitivity 60%; specificity 64%) in identifying the VT isthmus. In amiodarone-treated patients (n = 20), the SR substrate map IZ PF was significantly lower (222 Hz (186-257) vs. 303 Hz (244-375), P = 0.009) compared to amiodarone-naïve patients (n = 10). The ≥200 Hz LVA PF threshold resulted in an 80% freedom from VT with a trend towards reduced ablation lesions and radiofrequency times. LVA PF substrate mapping identifies critical components of the VT circuit with an optimal threshold of ≥200 Hz. Isthmus PF is influenced by chronic amiodarone therapy with lower values observed during RV pacing.

Optimized periphery-core interface increases fitness of the Bacillus subtilisglmS ribozyme.

Like other functional RNAs, ribozymes encode a conserved catalytic center supported by peripheral domains that vary among ribozyme sub-families. To understand how core-periphery interactions contribute to ribozyme fitness, we compared the cleavage kinetics of all single base substitutions at 152 sites across the Bacillus subtilis glmS ribozyme by high-throughput sequencing (k-seq). The in vitro activity map mirrored phylogenetic sequence conservation in glmS ribozymes, indicating that biological fitness reports all biochemically important positions. The k-seq results and folding assays showed that most deleterious mutations lower activity by impairing ribozyme self-assembly. All-atom molecular dynamics simulations of the complete ribozyme revealed how individual mutations in the core or the IL4 peripheral loop introduce a non-native tertiary interface that rewires the catalytic center, eliminating activity. We conclude that the need to avoid non-native helix packing powerfully constrains the evolution of tertiary structure motifs in RNA.

The correlation between the active/buried faults and aeromagnetic data in inner and inner-east Anatolia, Turkey

The paper focuses on investigating the active/buried faults in inner and inner-east Anatolia, Turkey based on aeromagnetic data and comparison with the active fault map. This region includes the major tectonic units of Turkey and the largest basins of Turkey (Tuzg.lü Basin (TB) and Sivas Basin (SB)). Thus, the region deserves a detail investigation to delineate the lineaments of the active/buried fault system or the causative sources located in the upper crust. To this end, the region has been divided into two parts as region-1 and 2 including TB and SB, respectively. In this context, the geophysical data processing techniques including power spectrum analysis, high‑pass filter, and second vertical derivative method (SVD) have been applied to the aeromagnetic data of the both regions. Firstly, by utilizing power spectrum analysis, the average depths of the deep and shallow causative sources in the region-1 have been calculated as 11.37 km and 5.01 km whereas those depths are 8.5 km and 3.24 km for region-2. Afterwards, to enhance the responses of the residual anomalies, the high‑pass filter and the SVD method have been applied to the data. Hereby, these potential lineaments produced from the SVD map have been compared with the active fault map to delineate the uncovering buried faults and their lineaments in the region. Contrary to the discussion in the literature, this study revealed for the first time that the Tuzgölü fault zone located in the eastern part of lake does not merge with the Ecemis fault (EF) zone. Furthermore, the study has shown that this fault zone extended from Kayseri city to Sivas city does not junction with North Anatolian Fault Zone in the north, as well.

Rice Disease Image Classification using MobileNetV2 Pretrained Model with AcivationAttention Visualization using Gradient-weighted Class Activation Mapping (Grad-CAM)

Rice diseases is one of the recurring factors of failed harvest and reduced rice production output and it is one of the biggest problems for Indonesia especially because of Indonesia’s population dependent of rice as it’s staple food. Pest control using fast dan early detection is one of the solutions to reduce the potential of failed harvest. Using a pretrained model of MobileNetV2, a model has been successfully built to detect for different types of rice diseases using convolutional neural network architecture and transfer learning method. The modelling process uses 7.077 image sample of rice infected by bacterial blight, blast, brown spot, and tungro. The model then can be accessed through website and has 99% of classification accuracy. Furthermore, using Gradient-Weighted Class Activation Mapping (Grad-CAM) rice disease can then be marked on top of the original image as a heat map, providing the users easier way to localize dan verify the classification result faster.

Tailoring a CRISPR/Cas-based Epigenome Editor for Programmable Chromatin Acylation and Decreased Cytotoxicity.

Engineering histone acylation states can inform mechanistic epigenetics and catalyze therapeutic epigenome editing opportunities. Here, we developed engineered lysine acyltransferases that enable the programmable deposition of acetylation and longer-chain acylations. We show that targeting an engineered lysine crotonyltransferase results in weak levels of endogenous enhancer activation yet retains potency when targeted to promoters. We further identify a single mutation within the catalytic core of human p300 that preserves enzymatic activity while substantially reducing cytotoxicity, enabling improved viral delivery. We leveraged these capabilities to perform single-cell CRISPR activation screening and map enhancers to the genes they regulate in situ. We also discover acylation-specific interactions and find that recruitment of p300, regardless of catalytic activity, to prime editing sites can improve editing efficiency. These new programmable epigenome editing tools and insights expand our ability to understand the mechanistic role of lysine acylation in epigenetic and cellular processes and perform functional genomic screens.

Deep Learning-Driven Prediction of Mechanical Properties of 316L Stainless Steel Metallographic by Laser Powder Bed Fusion.

This study developed a new metallography-property relationship neural network (MPR-Net) to predict the relationship between the microstructure and mechanical properties of 316L stainless steel built by laser powder bed fusion (LPBF). The accuracy R2 of MPR-Net was 0.96 and 0.91 for tensile strength and Vickers hardness predictions, respectively, based on optical metallurgy images. Feature visualisation methods, such as gradient-weighted class activation mapping (Grad-CAM) and clustering, were employed to interpret the abstract features within the MPR-Net, providing insights into the molten pool morphology and grain formation mechanisms during the LPBF process. Experimental results showed that the optimal process parameters-190 W laser power and 700 mm/s scanning speed-yielded a maximum tensile strength of 762.83 MPa and a Vickers hardness of 253.07 HV0.2 with nearly full densification (99.97%). The study marks the first application of a convolutional neural network (MPR-Net) to predict the mechanical properties of 316L stainless steel samples manufactured through laser powder bed fusion (LPBF) based on metallography. It innovatively employs techniques such as gradient-weighted class activation mapping (Grad-CAM), spatial coherence testing, and clustering to provide deeper insights into the workings of the machine learning model, enhancing the interpretability of complex neural network decisions in material science.