BACKGROUND: The occurrence, development, and outbreak of tea diseases and pests pose a significant challenge to the quality and yield of tea, necessitating prompt identification and control measures. Given the vast array of tea diseases and pests, coupled with the intricacies of the tea planting environment, accurate and rapid diagnosis remains elusive. In addressing this issue, the present study investigates the utilization of transfer learning convolution neural networks for the identification of tea diseases and pests. Our objective is to facilitate the accurate and expeditious detection of diseases and pests affecting the Yunnan Big leaf kind of tea within its complex ecological niche.
RESULTS: Initially, we gathered 1878 image data encompassing 10 prevalent types of tea diseases and pests from complex environments within tea plantations, compiling a comprehensive dataset. Additionally, we employed data augmentation techniques to enrich the sample diversity. Leveraging the ImageNet pre-trained model, we conducted a comprehensive evaluation and identified the Xception architecture as the most effective model. Notably, the integration of an attention mechanism within the Xeption model did not yield improvements in recognition performance. Subsequently, through transfer learning and the freezing core strategy, we achieved a test accuracy rate of 98.58% and a verification accuracy rate of 98.2310%.
CONCLUSIONS: These outcomes signify a significant stride towards accurate and timely detection, holding promise for enhancing the sustainability and productivity of Yunnan tea. Our findings provide a theoretical foundation and technical guidance for the development of online detection technologies for tea diseases and pests in Yunnan.

$Objective.$ Domain adaptation has been recognized as a potent solution to the challenge of limited training data for electroencephalography (EEG) classification tasks. Existing studies primarily focus on homogeneous environments, however, the heterogeneous properties of EEG data arising from device diversity cannot be overlooked. This motivates the development of heterogeneous domain adaptation methods that can fully exploit the knowledge from an auxiliary heterogeneous domain for EEG classification. $Approach.$ In this article, we propose a novel model named Informative Representation Fusion (IRF) to tackle the problem of unsupervised heterogeneous domain adaptation in the context of EEG data. In IRF, we consider different perspectives of data, i.e., independent identically distributed (iid) and non-iid, to learn different representations. Specifically, from the non-iid perspective, IRF models high-order correlations among data by hypergraphs and develops hypergraph encoders to obtain data representations of each domain. From the non-iid perspective, by applying multi-layer perceptron networks to the source and target domain data, we achieve another type of representation for both domains. Subsequently, an attention mechanism is used to fuse these two types of representations to yield informative features. To learn transferable representations, the Maximum Mean Discrepancy is utilized to align the distributions of the source and target domains based on the fused features. $Main~results.$ Experimental results on several real-world datasets demonstrate the effectiveness of the proposed model. $Significance.$ This article handles an EEG classification situation where the source and target EEG data lie in different spaces, and what\'s more, under an unsupervised learning setting. This situation is practical in the real world but barely studied in the literature. The proposed model achieves high classification accuracy, and this study is important for the commercial applications of EEG-based BCIs.

UNASSIGNED: Cotton yield estimation is crucial in the agricultural process, where the accuracy of boll detection during the flocculation period significantly influences yield estimations in cotton fields. Unmanned Aerial Vehicles (UAVs) are frequently employed for plant detection and counting due to their cost-effectiveness and adaptability.
UNASSIGNED: Addressing the challenges of small target cotton bolls and low resolution of UAVs, this paper introduces a method based on the YOLO v8 framework for transfer learning, named YOLO small-scale pyramid depth-aware detection (SSPD). The method combines space-to-depth and non-strided convolution (SPD-Conv) and a small target detector head, and also integrates a simple, parameter-free attentional mechanism (SimAM) that significantly improves target boll detection accuracy.
UNASSIGNED: The YOLO SSPD achieved a boll detection accuracy of 0.874 on UAV-scale imagery. It also recorded a coefficient of determination (R2) of 0.86, with a root mean square error (RMSE) of 12.38 and a relative root mean square error (RRMSE) of 11.19% for boll counts.
UNASSIGNED: The findings indicate that YOLO SSPD can significantly improve the accuracy of cotton boll detection on UAV imagery, thereby supporting the cotton production process. This method offers a robust solution for high-precision cotton monitoring, enhancing the reliability of cotton yield estimates.

OBJECTIVE: Cardiac perfusion MRI is vital for disease diagnosis, treatment planning, and risk stratification, with anomalies serving as markers of underlying ischemic pathologies. AI-assisted methods and tools enable accurate and efficient left ventricular (LV) myocardium segmentation on all DCE-MRI timeframes, offering a solution to the challenges posed by the multidimensional nature of the data. This study aims to develop and assess an automated method for LV myocardial segmentation on DCE-MRI data of a local hospital.
METHODS: The study consists of retrospective DCE-MRI data from 55 subjects acquired at the local hospital using a 1.5 T MRI scanner. The dataset included subjects with and without cardiac abnormalities. The timepoint for the reference frame (post-contrast LV myocardium) was identified using standard deviation across the temporal sequences. Iterative image registration of other temporal images with respect to this reference image was performed using Maxwell\'s demons algorithm. The registered stack was fed to the model built using the U-Net framework for predicting the LV myocardium at all timeframes of DCE-MRI.
RESULTS: The mean and standard deviation of the dice similarity coefficient (DSC) for myocardial segmentation using pre-trained network Net_cine is 0.78 ± 0.04, and for the fine-tuned network Net_dyn which predicts mask on all timeframes individually, it is 0.78 ± 0.03. The DSC for Net_dyn ranged from 0.71 to 0.93. The average DSC achieved for the reference frame is 0.82 ± 0.06.
CONCLUSIONS: The study proposed a fast and fully automated AI-assisted method to segment LV myocardium on all timeframes of DCE-MRI data. The method is robust, and its performance is independent of the intra-temporal sequence registration and can easily accommodate timeframes with potential registration errors.

Artificial intelligence (AI) has played a vital role in computer-aided drug design (CADD). This development has been further accelerated with the increasing use of machine learning (ML), mainly deep learning (DL), and computing hardware and software advancements. As a result, initial doubts about the application of AI in drug discovery have been dispelled, leading to significant benefits in medicinal chemistry. At the same time, it is crucial to recognize that AI is still in its infancy and faces a few limitations that need to be addressed to harness its full potential in drug discovery. Some notable limitations are insufficient, unlabeled, and non-uniform data, the resemblance of some AI-generated molecules with existing molecules, unavailability of inadequate benchmarks, intellectual property rights (IPRs) related hurdles in data sharing, poor understanding of biology, focus on proxy data and ligands, lack of holistic methods to represent input (molecular structures) to prevent pre-processing of input molecules (feature engineering), etc. The major component in AI infrastructure is input data, as most of the successes of AI-driven efforts to improve drug discovery depend on the quality and quantity of data, used to train and test AI algorithms, besides a few other factors. Additionally, data-gulping DL approaches, without sufficient data, may collapse to live up to their promise. Current literature suggests a few methods, to certain extent, effectively handle low data for better output from the AI models in the context of drug discovery. These are transferring learning (TL), active learning (AL), single or one-shot learning (OSL), multi-task learning (MTL), data augmentation (DA), data synthesis (DS), etc. One different method, which enables sharing of proprietary data on a common platform (without compromising data privacy) to train ML model, is federated learning (FL). In this review, we compare and discuss these methods, their recent applications, and limitations while modeling small molecule data to get the improved output of AI methods in drug discovery. Article also sums up some other novel methods to handle inadequate data.

BACKGROUND: Virtual and augmented reality surgical simulators, integrated with machine learning, are becoming essential for training psychomotor skills, and analyzing surgical performance. Despite the promise of methods like the Connection Weights Algorithm, the small sample sizes (small number of participants (N)) typical of these trials challenge the generalizability and robustness of models. Approaches like data augmentation and transfer learning from models trained on similar surgical tasks address these limitations.
OBJECTIVE: To demonstrate the efficacy of artificial neural network and transfer learning algorithms in evaluating virtual surgical performances, applied to a simulated oblique lateral lumbar interbody fusion technique in an augmented and virtual reality simulator.
METHODS: The study developed and integrated artificial neural network algorithms within a novel simulator platform, using data from the simulated tasks to generate 276 performance metrics across motion, safety, and efficiency. Innovatively, it applies transfer learning from a pre-trained ANN model developed for a similar spinal simulator, enhancing the training process, and addressing the challenge of small datasets.
METHODS: Musculoskeletal Biomechanics Research Lab; Neurosurgical Simulation and Artificial Intelligence Learning Centre, McGill University, Montreal, Canada.
METHODS: Twenty-seven participants divided into 3 groups: 9 post-residents, 6 senior and 12 junior residents.
RESULTS: Two models, a stand-alone model trained from scratch and another leveraging transfer learning, were trained on nine selected surgical metrics achieving 75 % and 87.5 % testing accuracy respectively.
CONCLUSIONS: This study presents a novel blueprint for addressing limited datasets in surgical simulations through the strategic use of transfer learning and data augmentation. It also evaluates and reinforces the application of the Connection Weights Algorithm from our previous publication. Together, these methodologies not only enhance the precision of performance classification but also advance the validation of surgical training platforms.

OBJECTIVE: Recent advancements in brain-computer interface (BCI) technology have seen a significant shift towards incorporating complex decoding models such as deep neural networks (DNNs) to enhance performance. These models are particularly crucial for sophisticated tasks such as regression for decoding arbitrary movements. However, these BCI models trained and tested on individual data often face challenges with limited performance and generalizability across different subjects. This limitation is primarily due to a tremendous number of parameters of DNN models. Training complex models demands extensive datasets. Nevertheless, group data from many subjects may not produce sufficient decoding performance because of inherent variability in neural signals both across individuals and over time METHODS: To address these challenges, this study proposed a transfer learning approach that could effectively adapt to subject-specific variability in cortical regions. Our method involved training two separate movement decoding models: one on individual data and another on pooled group data. We then created a salience map for each cortical region from the individual model, which helped us identify the input\'s contribution variance across subjects. Based on the contribution variance, we combined individual and group models using a modified knowledge distillation framework. This approach allowed the group model to be universally applicable by assigning greater weights to input data, while the individual model was fine-tuned to focus on areas with significant individual variance RESULTS: Our combined model effectively encapsulated individual variability. We validated this approach with nine subjects performing arm-reaching tasks, with our method outperforming (mean correlation coefficient, r = 0.75) both individual (r = 0.70) and group models (r = 0.40) in decoding performance. In particular, there were notable improvements in cases where individual models showed low performances (e.g., r = 0.50 in the individual decoder to r = 0.61 in the proposed decoder) CONCLUSIONS: These results not only demonstrate the potential of our method for robust BCI, but also underscore its ability to generalize individual data for broader applicability.

Thyroid nodule evaluation using ultrasound is dependent on radiologist experience, but deep learning (DL) models can improve intra-reader agreements. DL model development for medical imaging with small datasets can be challenging. Transfer learning is a technique used in the development of DL models to improve model performance in data-limited scenarios. Here, we investigate the impact of transfer learning with domain-specific RadImageNet dataset and non-medical ImageNet on the robustness of classifying thyroid nodules into benign and malignant. We retrospectively collected 822 ultrasound images of thyroid nodules of patients who underwent fine needle aspiration in our institute. We split our data and used 101 cases in a test set and 721 cases for cross-validation. A Resnet-18 model was trained to classify thyroid nodules into benign and malignant. Then, we trained the same model architecture with transferred weights from ImageNet and RadImageNet. The model without transfer learning for thyroid nodule classification achieved an AUROC of 0.69. The AUROC of our model after transfer learning with ImageNet pre-trained weights was 0.79. Our model achieved an AUROC of 0.83 from transfer learning of the RadImageNet pre-trained weights. The AUROC from the classification model without transfer learning significantly improved after transfer learning with ImageNet (p-value = 0.03) and RadImageNet transfer learning (p-value <0.01). There was a statistically significant distinction in performance between the model utilizing RadImageNet transfer learning and that employing ImageNet transfer learning (p-value <0.01). We demonstrate the potential of RadImageNet as a domain-specific source for transfer learning in thyroid nodule classification.

To understand neural basis of animal behavior, it is necessary to monitor neural activity and behavior in freely moving animal before building relationship between them. Here we use light sheet fluorescence microscope (LSFM) combined with microfluidic chip to simultaneously capture neural activity and body movement in small freely behaving Drosophila larva. We develop a transfer learning based method to simultaneously track the continuously changing body posture and activity of neurons that move together using a sub-region tracking network with a precise landmark estimation network for the inference of target landmark trajectory. Based on the tracking of each labelled neuron, the activity of the neuron indicated by fluorescent intensity is calculated. For each video, annotation of only 20 frames in a video is sufficient to yield human-level accuracy for all other frames. The validity of this method is further confirmed by reproducing the activity pattern of PMSIs (period-positive median segmental interneurons) and larval movement as previously reported. Using this method, we disclosed the correlation between larval movement and left-right asymmetry in activity of a group of unidentified neurons labelled by R52H01-Gal4 and further confirmed the roles of these neurons in bilateral balance of body contraction during larval crawling by genetic inhibition of these neurons. Our method provides a new tool for accurate extraction of neural activities and movement of freely behaving small-size transparent animals.

The combination of upconverting nanoparticles (UCNPs) and immunochromatography has become a widely used and promising new detection technique for point-of-care testing (POCT). However, their low luminescence efficiency, non-specific adsorption, and image noise have always limited their progress toward practical applications. Recently, artificial intelligence (AI) has demonstrated powerful representational learning and generalization capabilities in computer vision. We report for the first time a combination of AI and upconversion nanoparticle-based lateral flow assays (UCNP-LFAs) for the quantitative detection of commercial internet of things (IoT) devices. This universal UCNPs quantitative detection strategy combines high accuracy, sensitivity, and applicability in the field detection environment. By using transfer learning to train AI models in a small self-built database, we not only significantly improved the accuracy and robustness of quantitative detection, but also efficiently solved the actual problems of data scarcity and low computing power of POCT equipment. Then, the trained AI model was deployed in IoT devices, whereby the detection process does not require detailed data preprocessing to achieve real-time inference of quantitative results. We validated the quantitative detection of two detectors using eight transfer learning models on a small dataset. The AI quickly provided ultra-high accuracy prediction results (some models could reach 100% accuracy) even when strong noise was added. Simultaneously, the high flexibility of this strategy promises to be a general quantitative detection method for optical biosensors. We believe that this strategy and device have a scientific significance in revolutionizing the existing POCT technology landscape and providing excellent commercial value in the in vitro diagnostics (IVD) industry.