Enhanced choroid plexus segmentation with 3D UX-Net and its association with disease progression in multiple sclerosis.

BACKGROUND: The choroid plexus (CP) is suggested to be closely associated with the neuroinflammation of multiple sclerosis (MS). Segmentation based on deep learning (DL) could facilitate rapid and reproducible volume assessment of the CP, which is crucial for elucidating its role in MS. PURPOSE: To develop a reliable DL model for the automatic segmentation of CP, and further validate its clinical significance in MS. METHODS: The 3D UX-Net model (3D U-Net used for comparison) was trained and validated on T1-weighted MRI from a cohort of 216 relapsing-remitting MS (RRMS) patients and 75 healthy subjects. Among these, 53 RRMS with baseline and 2-year follow-up scans formed an internal test set (dataset1b). Another 58 RRMS from multi-center data served as an external test set (dataset2). Dice coefficient was computed to assess segmentation performance. Compare the correlation of CP volume obtained through automatic and manual segmentation with clinical outcomes in MS. Disability and cognitive function of patients were assessed using the Expanded Disability Status Scale (EDSS) and Symbol Digit Modalities Test (SDMT). RESULTS: The 3D UX-Net model achieved Dice coefficients of 0.875 ± 0.030 and 0.870 ± 0.044 for CP segmentation on dataset1b and dataset2, respectively, outperforming 3D U-Net's scores of 0.809 ± 0.098 and 0.601 ± 0.226. Furthermore, CP volumes segmented by the 3D UX-Net model aligned consistently with clinical outcomes compared to manual segmentation. In dataset1b, both manual and automatic segmentation revealed a significant positive correlation between normalized CP volume (nCPV) and EDSS scores at baseline (manual: r = 0.285, p = 0.045; automatic: r = 0.287, p = 0.044) and a negative correlation with SDMT scores (manual: r = -0.331, p = 0.020; automatic: r = -0.329, p = 0.021). In dataset2, similar correlations were found with EDSS scores (manual: r = 0.337, p = 0.021; automatic: r = 0.346, p = 0.017). Meanwhile, in dataset1b, both manual and automatic segmentation revealed a significant increase in nCPV from baseline to follow-up (p < 0.05). The increase of nCPV was more pronounced in patients with disability worsened than stable patients (manual: p = 0.023; automatic: p = 0.018). Patients receiving disease-modifying therapy (DMT) exhibited a significantly lower nCPV increase than untreated patients (manual: p = 0.004; automatic: p = 0.004). CONCLUSION: The 3D UX-Net model demonstrated strong segmentation performance for the CP, and the automatic segmented CP can be directly used in MS clinical practice. CP volume can serve as a surrogate imaging biomarker for monitoring disease progression and DMT response in MS patients.

Infinite-dimensional reservoir computing.

Reservoir computing approximation and generalization bounds are proved for a new concept class of input/output systems that extends the so-called generalized Barron functionals to a dynamic context. This new class is characterized by the readouts with a certain integral representation built on infinite-dimensional state-space systems. It is shown that this class is very rich and possesses useful features and universal approximation properties. The reservoir architectures used for the approximation and estimation of elements in the new class are randomly generated echo state networks with either linear or ReLU activation functions. Their readouts are built using randomly generated neural networks in which only the output layer is trained (extreme learning machines or random feature neural networks). The results in the paper yield a recurrent neural network-based learning algorithm with provable convergence guarantees that do not suffer from the curse of dimensionality when learning input/output systems in the class of generalized Barron functionals and measuring the error in a mean-squared sense.

Multi-level sequence denoising with cross-signal contrastive learning for sequential recommendation.

Sequential recommender systems (SRSs) aim to suggest next item for a user based on her historical interaction sequences. Recently, many research efforts have been devoted to attenuate the influence of noisy items in sequences by either assigning them with lower attention weights or discarding them directly. The major limitation of these methods is that the former would still prone to overfit noisy items while the latter may overlook informative items. To the end, in this paper, we propose a novel model named Multi-level Sequence Denoising with Cross-signal Contrastive Learning (MSDCCL) for sequential recommendation. To be specific, we first introduce a target-aware user interest extractor to simultaneously capture users' long and short term interest with the guidance of target items. Then, we develop a multi-level sequence denoising module to alleviate the impact of noisy items by employing both soft and hard signal denoising strategies. Additionally, we extend existing curriculum learning by simulating the learning pattern of human beings. It is worth noting that our proposed model can be seamlessly integrated with a majority of existing recommendation models and significantly boost their effectiveness. Experimental studies on five public datasets are conducted and the results demonstrate that the proposed MSDCCL is superior to the state-of-the-art baselines. The source code is publicly available at https://github.com/lalunex/MSDCCL/tree/main.

Class-incremental learning with Balanced Embedding Discrimination Maximization.

Class incremental learning is committed to solving representation learning and classification assignments while avoiding catastrophic forgetting in scenarios where categories are increasing. In this work, a unified method named Balanced Embedding Discrimination Maximization (BEDM) is developed to make the intermediate embedding more distinctive. Specifically, we utilize an orthogonality constraint based on doubly-blocked Toeplitz matrix to minimize the correlation of convolution kernels, and an algorithm for similarity visualization is introduced. Furthermore, uneven samples and distribution shift among old and new tasks eventuate strongly biased classifiers. To mitigate the imbalance, we propose an adaptive balance weighting in softmax to compensate insufficient categories dynamically. In addition, hybrid embedding learning is introduced to preserve knowledge from old models, which involves less hyper-parameters than conventional knowledge distillation. Our proposed method outperforms the existing approaches on three mainstream benchmark datasets. Moreover, we technically visualize that our method can produce a more uniform similarity histogram and more stable spectrum. Grad-CAM and t-SNE visualizations further confirm its effectiveness. Code is available at https://github.com/wqzh/BEDM.

Unsupervised and semi-supervised domain adaptation networks considering both global knowledge and prototype-based local class information for Motor Imagery Classification.

The non-stationarity of EEG signals results in variability across sessions, impeding model building and data sharing. In this paper, we propose a domain adaptation method called GPL, which simultaneously considers global knowledge and prototype-based local class information to enhance the classification accuracy of motor imagery signals. Depending on the amount of labeled data available in the target domain, the method is implemented in both unsupervised and semi-supervised versions. Specifically, at the global level, we employ the maximum mean difference (MMD) loss to globally constrain the feature space, achieving comprehensive alignment. In the context of class-level operations, we propose two memory banks designed to accommodate class prototypes in each domain and constrain feature embeddings by applying two prototype-based contrastive losses. The source contrastive loss is used to organize source features spatially based on categories, thereby reconciling inter-class and intra-class relationships, while the interactive contrastive loss is employed to facilitate cross-domain information interaction. Simultaneously, in unsupervised scenarios, to mitigate the adverse effects of excessive pseudo-labels, we introduce an entropy-aware strategy that dynamically evaluates the confidence level of target data and personalized constraints on the participation of interactive contrastive loss. To validate our approach, extensive experiments were conducted on a highly regarded public EEG dataset, namely Dataset IIa of the BCI Competition IV, as well as a large-scale EEG dataset called GigaDB. The experiments yielded average classification accuracies of 86.03% and 84.22% respectively. These results demonstrate that our method is an effective EEG decoding model, conducive to advancing the development of motor imagery brain-computer interfaces. The architecture proposed in this study and the code for data partitioning can be found at https://github.com/zhangdx21/GPL.

Continual pre-training mitigates forgetting in language and vision.

Pre-trained models are commonly used in Continual Learning to initialize the model before training on the stream of non-stationary data. However, pre-training is rarely applied during Continual Learning. We investigate the characteristics of the Continual Pre-Training scenario, where a model is continually pre-trained on a stream of incoming data and only later fine-tuned to different downstream tasks. We introduce an evaluation protocol for Continual Pre-Training which monitors forgetting against a Forgetting Control dataset not present in the continual stream. We disentangle the impact on forgetting of 3 main factors: the input modality (NLP, Vision), the architecture type (Transformer, ResNet) and the pre-training protocol (supervised, self-supervised). Moreover, we propose a Sample-Efficient Pre-training method (SEP) that speeds up the pre-training phase. We show that the pre-training protocol is the most important factor accounting for forgetting. Surprisingly, we discovered that self-supervised continual pre-training in both NLP and Vision is sufficient to mitigate forgetting without the use of any Continual Learning strategy. Other factors, like model depth, input modality and architecture type are not as crucial.

Progressive Neighbor-masked Contrastive Learning for Fusion-style Deep Multi-view Clustering.

Fusion-style Deep Multi-view Clustering (FDMC) can efficiently integrate comprehensive feature information from latent embeddings of multiple views and has drawn much attention recently. However, existing FDMC methods suffer from the interference of view-specific information for fusion representation, affecting the learning of discriminative cluster structure. In this paper, we propose a new framework of Progressive Neighbor-masked Contrastive Learning for FDMC (PNCL-FDMC) to tackle the aforementioned issues. Specifically, by using neighbor-masked contrastive learning, PNCL-FDMC can explicitly maintain the local structure during the embedding aggregation, which is beneficial to the common semantics enhancement on the fusion view. Based on the consistent aggregation, the fusion view is further enhanced by diversity-aware cluster structure enhancement. In this process, the enhanced cluster assignments and cluster discrepancies are employed to guide the weighted neighbor-masked contrastive alignment of semantic structure between individual views and the fusion view. Extensive experiments validate the effectiveness of the proposed framework, revealing its ability in discriminative representation learning and improving clustering performance.

Optimized efficient attention-based network for facial expressions analysis in neurological health care.

Facial Expression Analysis (FEA) plays a vital role in diagnosing and treating early-stage neurological disorders (NDs) like Alzheimer's and Parkinson's. Manual FEA is hindered by expertise, time, and training requirements, while automatic methods confront difficulties with real patient data unavailability, high computations, and irrelevant feature extraction. To address these challenges, this paper proposes a novel approach: an efficient, lightweight convolutional block attention module (CBAM) based deep learning network (DLN) to aid doctors in diagnosing ND patients. The method comprises two stages: data collection of real ND patients, and pre-processing, involving face detection and an attention-enhanced DLN for feature extraction and refinement. Extensive experiments with validation on real patient data showcase compelling performance, achieving an accuracy of up to 73.2%. Despite its efficacy, the proposed model is lightweight, occupying only 3MB, making it suitable for deployment on resource-constrained mobile healthcare devices. Moreover, the method exhibits significant advancements over existing FEA approaches, holding tremendous promise in effectively diagnosing and treating ND patients. By accurately recognizing emotions and extracting relevant features, this approach empowers medical professionals in early ND detection and management, overcoming the challenges of manual analysis and heavy models. In conclusion, this research presents a significant leap in FEA, promising to enhance ND diagnosis and care.The code and data used in this work are available at: https://github.com/munsif200/Neurological-Health-Care.

Unveiling the evolution of policies for enhancing protein structure predictions: A comprehensive analysis.

Predicting protein structure is both fascinating and formidable, playing a crucial role in structure-based drug discovery and unraveling diseases with elusive origins. The Critical Assessment of Protein Structure Prediction (CASP) serves as a biannual battleground where global scientists converge to untangle the intricate relationships within amino acid chains. Two primary methods, Template-Based Modeling (TBM) and Template-Free (TF) strategies, dominate protein structure prediction. The trend has shifted towards Template-Free predictions due to their broader sequence coverage with fewer templates. The predictive process can be broadly classified into contact map, binned-distance, and real-valued distance predictions, each with distinctive strengths and limitations manifested through tailored loss functions. We have also introduced revolutionary end-to-end, and all-atom diffusion-based techniques that have transformed protein structure predictions. Recent advancements in deep learning techniques have significantly improved prediction accuracy, although the effectiveness is contingent upon the quality of input features derived from natural bio-physiochemical attributes and Multiple Sequence Alignments (MSA). Hence, the generation of high-quality MSA data holds paramount importance in harnessing informative input features for enhanced prediction outcomes. Remarkable successes have been achieved in protein structure prediction accuracy, however not enough for what structural knowledge was intended to, which implies need for development in some other aspects of the predictions. In this regard, scientists have opened other frontiers for protein structural prediction. The utilization of subsampling in multiple sequence alignment (MSA) and protein language modeling appears to be particularly promising in enhancing the accuracy and efficiency of predictions, ultimately aiding in drug discovery efforts. The exploration of predicting protein complex structure also opens up exciting opportunities to deepen our knowledge of molecular interactions and design therapeutics that are more effective. In this article, we have discussed the vicissitudes that the scientists have gone through to improve prediction accuracy, and examined the effective policies in predicting from different aspects, including the construction of high quality MSA, providing informative input features, and progresses in deep learning approaches. We have also briefly touched upon transitioning from predicting single-chain protein structures to predicting protein complex structures. Our findings point towards promoting open research environments to support the objectives of protein structure prediction.

Deep Bayesian active learning-to-rank with relative annotation for estimation of ulcerative colitis severity.

Automatic image-based severity estimation is an important task in computer-aided diagnosis. Severity estimation by deep learning requires a large amount of training data to achieve a high performance. In general, severity estimation uses training data annotated with discrete (i.e., quantized) severity labels. Annotating discrete labels is often difficult in images with ambiguous severity, and the annotation cost is high. In contrast, relative annotation, in which the severity between a pair of images is compared, can avoid quantizing severity and thus makes it easier. We can estimate relative disease severity using a learning-to-rank framework with relative annotations, but relative annotation has the problem of the enormous number of pairs that can be annotated. Therefore, the selection of appropriate pairs is essential for relative annotation. In this paper, we propose a deep Bayesian active learning-to-rank that automatically selects appropriate pairs for relative annotation. Our method preferentially annotates unlabeled pairs with high learning efficiency from the model uncertainty of the samples. We prove the theoretical basis for adapting Bayesian neural networks to pairwise learning-to-rank and demonstrate the efficiency of our method through experiments on endoscopic images of ulcerative colitis on both private and public datasets. We also show that our method achieves a high performance under conditions of significant class imbalance because it automatically selects samples from the minority classes.

Application of triple-branch artificial neural network system for catalytic pellets combustion.

On the international level, it is common to act on reducing emissions from energy systems. However, in addition to industrial emissions, low-stack emissions also make a significant contribution. A good step in reducing its environmental impact, is to move to biofuels, including biomass. This paper examines the impact of placing a catalytic system in a retort boiler to minimize emissions of greenhouse gases, dust and other pollutants when burning pellets. The effect of platinum, and oxides of selected metals placed on the deflector as a solid catalyst was studied. Based on the experimental data, a branched artificial neural network was constructed and trained. The routing of three parallel topologies made it possible to achieve high accuracy while keeping the input data relatively simple. The system showed an average error of 3.54% against arbitrary test data. On the basis of experimental data as well as predictions returned by the artificial neural network, recommendations were shown for the catalysts used and their amounts. Depending on the biomass from which the pellet was produced, the experiment suggested the use of titanium or copper oxides. In the case of the neural network, it was able to select a better system, based on platinum, improving emission reductions by up to more than 19%, depending on the type of pellet used.

Investigating the features of risky driving behaviors on expressway diverge area based on conflict and modeling analysis.

Driving behaviors are important cause of expressway crash. In this study, method based on modified time-to-collision (MTTC) to identify risky driving behaviors on an expressway diverge area is proposed, thus investigating the impact of velocity and acceleration features of risky driving behavior. Firstly, MTTC is applied to judge whether the behavior is risky. Then, the relationships between velocity, acceleration and different driving behavior on the expressway diverge area were fit by binary logistic regression models (BLR) with L2 regularization and random forests (RF) models, and the models were interpreted by feature importance plots and partial dependency plots. The results show that the AUC metric of 4 RF models for 4 types of driving behaviors, namely, left lane change, right lane change, acceleration and deceleration, are 0.932, 0.845, 0.846 and 0.860 separately. The interpretation of models demonstrates that velocity and absolute value of acceleration greatly affect the risk of the driving behaviors. Different driving behaviors with a certain acceleration have a range of safety speed range. The range will get narrower with the growth of maximum absolute value of acceleration rate, and will be nearly non-exist when the acceleration is over 5 m/s2. In conclusion, this study provided a methodology to measure the risk of driving behaviors and establish a model to recognize of risky driving behaviors. The results can lay the foundation for making countermeasures to prevent risky driving behaviors by managing the vehicle speed.

Neural activity shaping in visual prostheses with deep learning.

OBJECTIVE: The visual perception provided by retinal prostheses is limited by the overlapping current spread of adjacent electrodes. This reduces the spatial resolution attainable with unipolar stimulation. Conversely, simultaneous multipolar stimulation guided by the measured neural responses - Neural Activity Shaping (NAS) - can attenuate excessive spread of excitation allowing for more precise control over the pattern of neural activation. However, predicting the results of a multipolar stimulus pattern is a challenging task. Previous attempts focused on analytical solutions based on an assumed linear nonlinear model of retinal response; an Analytical Model Inversion (AMI) approach. Here, we propose a model-free solution for NAS, using Artificial Neural Networks (ANNs) that could be trained with data acquired from the implant. APPROACH: Our method consists of two ANNs trained sequentially. The Measurement Predictor Network (MPN) is trained on data from the implant and is used to predict how the retina responds to multipolar stimulation. The Stimulus Generator Network (STG) is trained on a large dataset of natural images and uses the trained MPN to determine efficient multipolar stimulus patterns by learning its inverse model. We validate our method in silico using a realistic model of retinal response to multipolar stimulation. Main Results: We show that our ANN-based NAS approach produces sharper retinal activations than the conventional unipolar stimulation strategy. As a theoretical bench-mark of optimal NAS results, we implemented AMI stimulation by inverting the model used to simulate the retina. Our ANN strategy produced equivalent results to AMI, while not being restricted to any specific type of retina model and being three orders of magnitude more computationally efficient. SIGNIFICANCE: Our novel protocol provides a method for efficient and personalized retinal stimulation, which may improve the visual experience and quality of life of retinal prosthesis users. .

Permutation conditional mutual information to quantify TMS-evoked cortical connectivity in disorders of consciousness.

OBJECTIVE: To improve the understanding and diagnostic accuracy of disorders of consciousness (DOC) by quantifying transcranial magnetic stimulation evoked electroencephalography (TMS-EEG) connectivity using Permutation Conditional Mutual Information (PCMI). Approach: PCMI can characterize the functional connectivity between different brain regions. This study employed PCMI to analyze TMS-evoked cortical connectivity (TEC) in 154 DOC patients and 16 normal controls, focusing on optimizing parameter selection for PCMI (Data length, Order length, Time delay). We compared short-range and long-range PCMI values across different consciousness states-unresponsive wakefulness syndrome (UWS), minimally conscious state (MCS), and normal (NOR)-and assessed various feature selection and classification techniques to distinguish these states. Main Results: 1) PCMI can quantify TEC. We found optimal parameters to be Data length: 500ms; Order: 3; Time delay: 6ms. 2) TMS evoked potentials (TEP) for NOR showed a rich response, while MCS patients showed only a few components, and UWS patients had almost no significant components. The values of PCMI connectivity metrics demonstrated its usefulness for measuring cortical connectivity evoked by TMS. From NOR to MCS to UWS, the number and strength of TEC decreased. Quantitative analysis revealed significant differences in the strength and number of TEC in the entire brain, local regions and inter-regions among different consciousness states. 3) A decision tree with feature selection by mutual information performed the best (balanced accuracy: 87.0% and accuracy: 83.5%). This model could accurately identify NOR (100.0%), but had lower identification accuracy for UWS (86.5%) and MCS (74.1%). Significance: The application of PCMI in measuring TMS-evoked connectivity provides a robust metric that enhances our ability to differentiate between various states of consciousness in DOC patients. This approach not only aids in clinical diagnosis but also contributes to the broader understanding of cortical connectivity and consciousness. .

DP-MP: a novel cross-subject fatigue detection framework with DANN-based prototypical representation and mix-up pairwise learning.

OBJECTIVE: Electroencephalography (EEG) is widely recognized as an effective method for detecting fatigue. However, practical applications of EEG for fatigue detection in real-world scenarios are often challenging, particularly in cases involving subjects not included in the training datasets, owing to bio-individual differences and noisy labels. This study aims to develop an effective framework for cross-subject fatigue detection by addressing these challenges. APPROACH: In this study, we propose a novel framework, termed DP-MP, for cross-subject fatigue detection, which utilizes a Domain-Adversarial Neural Network (DANN)-based prototypical representation in conjunction with Mix-up pairwise learning. Our proposed DP-MP framework aims to mitigate the impact of bio-individual differences by encoding fatigue-related semantic structures within EEG signals and exploring shared fatigue prototype features across individuals. Notably, to the best of our knowledge, this work is the first to conceptualize fatigue detection as a pairwise learning task, thereby effectively reducing the interference from noisy labels. Furthermore, we propose the Mix-up pairwise learning (MixPa) approach in the field of fatigue detection, which broadens the advantages of pairwise learning by introducing more diverse and informative relationships among samples. RESULTS: Cross-subject experiments were conducted on two benchmark databases, SEED-VIG and FTEF, achieving state-of-the-art performance with average accuracies of 88.14% and 97.41%, respectively. These promising results demonstrate our model's effectiveness and excellent generalization capability. SIGNIFICANCE: This is the first time EEG-based fatigue detection has been conceptualized as a pairwise learning task, offering a novel perspective to this field. Moreover, our proposed DP-MP framework effectively tackles the challenges of bio-individual differences and noisy labels in the fatigue detection field and demonstrates superior performance. Our work provides valuable insights for future research, promoting the application of brain-computer interfaces for fatigue detection in real-world scenarios. .

DiffuseRT: predicting likely anatomical deformations of patients undergoing radiotherapy.

Predicting potential deformations of patients can improve radiotherapy treatment planning. Here, we introduce new deep-learning models that predict likely anatomical changes during radiotherapy for head and neck cancer patients. Denoising diffusion probabilistic models (DDPMs) were developed to generate fraction-specific anatomical changes based on a reference cone-beam CT (CBCT), the fraction number and treatment dose. Three distinct DDPMs were developed: (1) the image model was trained to directly generate likely future CBCTs, (2) the deformable vector field (DVF) model was trained to generate DVFs that deform a reference CBCT and (3) the hybrid model was trained similarly to the DVF model, but without relying on an external deformable registration algorithm. The models were trained on 9 patients with longitudinal CBCT images (224 CBCTs) and evaluated on 5 patients (152 CBCTs). The generated images mainly exhibited random positioning shifts and small anatomical changes for early fractions. For later fractions, all models predicted weight losses in accordance with the training data. The distributions of volume and position changes of the body, esophagus, and parotids generated with the image and hybrid models were more similar to the ground truth distribution than the DVF model, evident from the lower Wasserstein distance achieved with the image (0.26) and hybrid model (0.25) compared to the DVF model (0.36). Generating several images for the same fraction did not yield the expected variability since the ground truth anatomical changes were only in 70% of the fractions within the 95% bounds predicted with the best model. Using the generated images for robust optimization of simplified proton therapy plans improved the worst-case clinical target volume V95 with 7% compared to optimizing with 3 mm set-up robustness while maintaining a similar integral dose. In conclusion, the newly developed DDPMs generate distributions similar to the real anatomical changes and have the potential to be used for robust anatomical optimization.

A deep learning-based method for the detection and segmentation of breast masses in ultrasound images.

OBJECTIVE: Automated detection and segmentation of breast masses in ultrasound images are critical for breast cancer diagnosis, but remain challenging due to limited image quality and complex breast tissues. This study aims to develop a deep learning-based method that enables accurate breast mass detection and segmentation in ultrasound images. Approach. A novel convolutional neural network-based framework that combines the You Only Look Once (YOLO) v5 network and the Global-Local (GOLO) strategy was developed. First, YOLOv5 was applied to locate the mass regions of interest (ROIs). Second, a Global Local-Connected Multi-Scale Selection (GOLO-CMSS) network was developed to segment the masses. The GOLO-CMSS operated on both the entire images globally and mass ROIs locally, and then integrated the two branches for a final segmentation output. Particularly, in global branch, CMSS applied Multi-Scale Selection (MSS) modules to automatically adjust the receptive fields, and Multi-Input (MLI) modules to enable fusion of shallow and deep features at different resolutions. The USTC dataset containing 28,477 breast ultrasound images was collected for training and test. The proposed method was also tested on three public datasets, UDIAT, BUSI and TUH. The segmentation performance of GOLO-CMSS was compared with others networks and three experienced radiologists. Main results. YOLOv5 outperformed other detection models with average precisions of 99.41%, 95.15%, 93.69% and 96.42% on the USTC, UDIAT, BUSI and TUH datasets, respectively. The proposed GOLO-CMSS showed superior segmentation performance over other state-of-the-art networks, with Dice similarity coefficients (DSCs) of 93.19%, 88.56%, 87.58% and 90.37% on the USTC, UDIAT, BUSI and TUH datasets, respectively. The mean DSC between GOLO-CMSS and each radiologist was significantly better than that between radiologists (p < 0.001). Significance. Our proposed method can accurately detect and segment breast masses with a decent performance comparable to radiologists, highlighting its great potential for clinical implementation in breast ultrasound examination.

The role of national nutrition programs on stunting reduction in Rwanda using machine learning classifiers: a retrospective study.

BACKGROUND: In Rwanda, the prevalence of childhood stunting has slightly decreased over the past five years, from 38% in 2015 to about 33% in 2020. It is evident whether Rwanda's multi-sectorial approach to reducing child stunting is consistent with the available scientific knowledge. The study was to examine the benefits of national nutrition programs on stunting reduction under two years in Rwanda using machine learning classifiers. METHODS: Data from the Rwanda DHS 2015-2020, MEIS and LODA household survey were used. By evaluating the best method for predicting the stunting reduction status of children under two years old, the five machine learning algorithms were modelled: Support Vector Machine, Logistic Regression, K-Near Neighbor, Random Forest, and Decision Tree. The study estimated the hazard ratio for the Cox Proportional Hazard Model and drew the Kaplan-Meier curve to compare the survivor risk of being stunted between program beneficiaries and non-beneficiaries. Logistic regression was used to identify the nutrition programs related to stunting reduction. Precision, recall, F1 score, accuracy, and Area under the Curve (AUC) are the metrics that were used to evaluate each classifier's performance to find the best one. RESULTS: Based on the provided data, the study revealed that the early childhood development (ECD) program (p-value = 0.041), nutrition sensitive direct support (NSDS) program (p-value = 0.03), ubudehe category (p-value = 0.000), toilet facility (p-value = 0.000), antenatal care (ANC) 4 visits (p-value = 0.002), fortified blended food (FBF) program (p-value = 0.038) and vaccination (p-value = 0.04) were found to be significant predictors of stunting reduction among under two children in Rwanda. Additionally, beneficiaries of early childhood development (p < .0001), nutrition sensitive direct support (p = 0.0055), antenatal care (p = 0.0343), Fortified Blended Food (p = 0.0136) and vaccination (p = 0.0355) had a lower risk of stunting than non-beneficiaries. Finally, Random Forest performed better than other classifiers, with precision scores of 83.7%, recall scores of 90.7%, F1 scores of 87.1%, accuracy scores of 83.9%, and AUC scores of 82.4%. CONCLUSION: The early childhood development (ECD) program, receiving the nutrition sensitive direct support (NSDS) program, focusing on households with the lowest wealth quintile (ubudehe category), sanitation facilities, visiting health care providers four times, receiving fortified blended food (FBF), and receiving all necessary vaccines are what determine the stunting reduction under two among the 17 districts of Rwanda. Finally, when compared to other models, Random Forest was shown to be the best machine learning (ML) classifier. Random forest is the best classifier for predicting the stunting reduction status of children under two years old.

Metabolism score and machine learning models for the prediction of esophageal squamous cell carcinoma progression.

The incomplete prediction of prognosis in esophageal squamous cell carcinoma (ESCC) patients is attributed to various therapeutic interventions and complex prognostic factors. Consequently, there is a pressing demand for enhanced predictive biomarkers that can facilitate clinical management and treatment decisions. This study recruited 491 ESCC patients who underwent surgical treatment at Huashan Hospital, Fudan University. We incorporated 14 blood metabolic indicators and identified independent prognostic indicators for overall survival through univariate and multivariate analyses. Subsequently, a metabolism score formula was established based on the biochemical markers. We constructed a nomogram and machine learning models utilizing the metabolism score and clinically significant prognostic features, followed by an evaluation of their predictive accuracy and performance. We identified alkaline phosphatase, free fatty acids, homocysteine, lactate dehydrogenase, and triglycerides as independent prognostic indicators for ESCC. Subsequently, based on these five indicators, we established a metabolism score that serves as an independent prognostic factor in ESCC patients. By utilizing this metabolism score in conjunction with clinical features, a nomogram can precisely predict the prognosis of ESCC patients, achieving an area under the curve (AUC) of 0.89. The random forest (RF) model showed superior predictive ability (AUC = 0.90, accuracy = 86%, Matthews correlation coefficient = 0.55). Finally, we used an RF model with optimal performance to establish an online predictive tool. The metabolism score developed in this study serves as an independent prognostic indicator for ESCC patients.

Aetiology and predictors of outcome in non-shockable in-hospital cardiac arrest: A retrospective cohort study from the Swedish Registry for Cardiopulmonary Resuscitation.

BACKGROUND: Non-shockable in-hospital cardiac arrest (IHCA) is a condition with diverse aetiology, predictive factors, and outcome. This study aimed to compare IHCA with initial asystole or pulseless electrical activity (PEA), focusing specifically on their aetiologies and the significance of predictive factors. METHODS: Using the Swedish Registry of Cardiopulmonary Resuscitation, adult non-shockable IHCA cases from 2018 to 2022 (n = 5788) were analysed. Exposure was initial rhythm, while survival to hospital discharge was the primary outcome. A random forest model with 28 variables was used to generate permutation-based variable importance for outcome prediction. RESULTS: Overall, 60% of patients (n = 3486) were male and the median age was 75 years (IQR 67-81). The most frequent arrest location (46%) was on general wards. Comorbidities were present in 79% of cases and the most prevalent comorbidity was heart failure (33%). Initial rhythm was PEA in 47% (n = 2702) of patients, and asystole in 53% (n = 3086). The most frequent aetiologies in both PEA and asystole were cardiac ischemia (24% vs. 19%, absolute difference [AD]: 5.4%; 95% confidence interval [CI] 3.0% to 7.7%), and respiratory failure (14% vs. 13%, no significant difference). Survival was higher in asystole (24%) than in PEA (17%) (AD: 7.3%; 95% CI 5.2% to 9.4%). Cardiopulmonary resuscitation (CPR) durations were longer in PEA, 18 vs 15 min (AD 4.9 min, 95% CI 4.0-5.9 min). The duration of CPR was the single most important predictor of survival across all subgroup and sensitivity analyses. Aetiology ranked as the second most important predictor in most analyses, except in the asystole subgroup where responsiveness at cardiac arrest team arrival took precedence. CONCLUSIONS: In this nationwide registry study of non-shockable IHCA comparing asystole to PEA, cardiac ischemia and respiratory failure were the predominant aetiologies. Duration of CPR was the most important predictor of survival, followed by aetiology. Asystole was associated with higher survival compared to PEA, possibly due to shorter CPR durations and a larger proportion of reversible aetiologies.