Automatic Target Detection from Satellite Imagery Using Machine Learning

Intraoperative AI-assisted early prediction of parathyroid and ischemia alert in endoscopic thyroid surgery

BACKGROUND

The preservation of parathyroid glands is crucial in endoscopic thyroid surgery to prevent hypocalcemia and related complications. However, current methods for identifying and protecting these glands have limitations. We propose a novel technique that has the potential to improve the safety and efficacy of endoscopic thyroid surgery.

PURPOSE

Our study aims to develop a deep learning model called PTAIR 2.0 (Parathyroid gland Artificial Intelligence Recognition) to enhance parathyroid gland recognition during endoscopic thyroidectomy. We compare its performance against traditional surgeon-based identification methods.

MATERIALS AND METHODS

Parathyroid tissues were annotated in 32 428 images extracted from 838 endoscopic thyroidectomy videos, forming the internal training cohort. An external validation cohort comprised 54 full-length videos. Six candidate algorithms were evaluated to select the optimal one. We assessed the model's performance in terms of initial recognition time, identification duration, and recognition rate and compared it with the performance of surgeons.

RESULTS

Utilizing the YOLOX algorithm, we developed PTAIR 2.0, which demonstrated superior performance with an AP50 score of 92.1%. The YOLOX algorithm achieved a frame rate of 25.14 Hz, meeting real-time requirements. In the internal training cohort, PTAIR 2.0 achieved AP50 values of 94.1%, 98.9%, and 92.1% for parathyroid gland early prediction, identification, and ischemia alert, respectively. Additionally, in the external validation cohort, PTAIR outperformed both junior and senior surgeons in identifying and tracking parathyroid glands (p < 0.001).

CONCLUSION

The AI-driven PTAIR 2.0 model significantly outperforms both senior and junior surgeons in parathyroid gland identification and ischemia alert during endoscopic thyroid surgery, offering potential for enhanced surgical precision and patient outcomes.

DEKR-SPrior: An Efficient Bottom-Up Keypoint Detection Model for Accurate Pod Phenotyping in Soybean

The pod and seed counts are important yield-related traits in soybean. High-precision soybean breeders face the major challenge of accurately phenotyping the number of pods and seeds in a high-throughput manner. Recent advances in artificial intelligence, especially deep learning (DL) models, have provided new avenues for high-throughput phenotyping of crop traits with increased precision. However, the available DL models are less effective for phenotyping pods that are densely packed and overlap in in situ soybean plants; thus, accurate phenotyping of the number of pods and seeds in soybean plant is an important challenge. To address this challenge, the present study proposed a bottom-up model, DEKR-SPrior (disentangled keypoint regression with structural prior), for in situ soybean pod phenotyping, which considers soybean pods and seeds analogous to human people and joints, respectively. In particular, we designed a novel structural prior (SPrior) module that utilizes cosine similarity to improve feature discrimination, which is important for differentiating closely located seeds from highly similar seeds. To further enhance the accuracy of pod location, we cropped full-sized images into smaller and high-resolution subimages for analysis. The results on our image datasets revealed that DEKR-SPrior outperformed multiple bottom-up models, viz., Lightweight-OpenPose, OpenPose, HigherHRNet, and DEKR, reducing the mean absolute error from 25.81 (in the original DEKR) to 21.11 (in the DEKR-SPrior) in pod phenotyping. This paper demonstrated the great potential of DEKR-SPrior for plant phenotyping, and we hope that DEKR-SPrior will help future plant phenotyping.

DeepHLAPred: a deep learning-based method for non-classical HLA binder prediction

Human leukocyte antigen (HLA) is closely involved in regulating the human immune system. Despite great advance in detecting classical HLA Class I binders, there are few methods or toolkits for recognizing non-classical HLA Class I binders. To fill in this gap, we have developed a deep learning-based tool called DeepHLAPred. The DeepHLAPred used electron-ion interaction pseudo potential, integer numerical mapping and accumulated amino acid frequency as initial representation of non-classical HLA binder sequence. The deep learning module was used to further refine high-level representations. The deep learning module comprised two parallel convolutional neural networks, each followed by maximum pooling layer, dropout layer, and bi-directional long short-term memory network. The experimental results showed that the DeepHLAPred reached the state-of-the-art performanceson the cross-validation test and the independent test. The extensive test demonstrated the rationality of the DeepHLAPred. We further analyzed sequence pattern of non-classical HLA class I binders by information entropy. The information entropy of non-classical HLA binder sequence implied sequence pattern to a certain extent. In addition, we have developed a user-friendly webserver for convenient use, which is available at http://www.biolscience.cn/DeepHLApred/ . The tool and the analysis is helpful to detect non-classical HLA Class I binder. The source code and data is available at https://github.com/tangxingyu0/DeepHLApred .

Crack Detection in Concrete Structures Using Deep Learning

Infrastructure, such as buildings, bridges, pavement, etc., needs to be examined periodically to maintain its reliability and structural health. Visual signs of cracks and depressions indicate stress and wear and tear over time, leading to failure/collapse if these cracks are located at critical locations, such as in load-bearing joints. Manual inspection is carried out by experienced inspectors who require long inspection times and rely on their empirical and subjective knowledge. This lengthy process results in delays that further compromise the infrastructure’s structural integrity. To address this limitation, this study proposes a deep learning (DL)-based autonomous crack detection method using the convolutional neural network (CNN) technique. To improve the CNN classification performance for enhanced pixel segmentation, 40,000 RGB images were processed before training a pretrained VGG16 architecture to create different CNN models. The chosen methods (grayscale, thresholding, and edge detection) have been used in image processing (IP) for crack detection, but not in DL. The study found that the grayscale models (F1 score for 10 epochs: 99.331%, 20 epochs: 99.549%) had a similar performance to the RGB models (F1 score for 10 epochs: 99.432%, 20 epochs: 99.533%), with the performance increasing at a greater rate with more training (grayscale: +2 TP, +11 TN images; RGB: +2 TP, +4 TN images). The thresholding and edge-detection models had reduced performance compared to the RGB models (20-epoch F1 score to RGB: thresholding −0.723%, edge detection −0.402%). This suggests that DL crack detection does not rely on colour. Hence, the model has implications for the automated crack detection of concrete infrastructures and the enhanced reliability of the gathered information.

Predicting and Improving the Waterlogging Resilience of Urban Communities in China—A Case Study of Nanjing

In recent years, urban communities in China have been continuously affected by extreme weather and emergencies, among which the rainstorm and waterlogging disasters pose a great threat to infrastructure and personnel safety. Chinese governments issue a series of waterlogging prevention and control policies, but the waterlogging prevention and mitigation of urban communities still needs to be optimized. The concept of “resilience” has unique advantages in the field of community disaster management, and building resilient communities can effectively make up for the limitations of the traditional top-down disaster management. Therefore, this paper focuses on the pre-disaster prevention and control of waterlogging in urban communities of China, following the idea of “concept analysis–influencing factor identification–evaluation indicators selection–impact mechanism analysis–resilience simulation prediction–empirical research–disaster adaptation strategy formulation”. The structural equation model and BP neural network are used by investigating the existing anti-waterlogging capitals of the target community to predict the future waterlogging resilience. Based on this simulation prediction model, and combined with the incentive and restraint mechanisms, suggestions on corrective measures can be put forward before the occurrence of waterlogging.

Big Data Management in Drug–Drug Interaction: A Modern Deep Learning Approach for Smart Healthcare

The detection and classification of drug–drug interactions (DDI) from existing data are of high importance because recent reports show that DDIs are among the major causes of hospital-acquired conditions and readmissions and are also necessary for smart healthcare. Therefore, to avoid adverse drug interactions, it is necessary to have an up-to-date knowledge of DDIs. This knowledge could be extracted by applying text-processing techniques to the medical literature published in the form of ‘Big Data’ because, whenever a drug interaction is investigated, it is typically reported and published in healthcare and clinical pharmacology journals. However, it is crucial to automate the extraction of the interactions taking place between drugs because the medical literature is being published in immense volumes, and it is impossible for healthcare professionals to read and collect all of the investigated DDI reports from these Big Data. To avoid this time-consuming procedure, the Information Extraction (IE) and Relationship Extraction (RE) techniques that have been studied in depth in Natural Language Processing (NLP) could be very promising. Since 2011, a lot of research has been reported in this particular area, and there are many approaches that have been implemented that can also be applied to biomedical texts to extract DDI-related information. A benchmark corpus is also publicly available for the advancement of DDI extraction tasks. The current state-of-the-art implementations for extracting DDIs from biomedical texts has employed Support Vector Machines (SVM) or other machine learning methods that work on manually defined features and that might be the cause of the low precision and recall that have been achieved in this domain so far. Modern deep learning techniques have also been applied for the automatic extraction of DDIs from the scientific literature and have proven to be very promising for the advancement of DDI extraction tasks. As such, it is pertinent to investigate deep learning techniques for the extraction and classification of DDIs in order for them to be used in the smart healthcare domain. We proposed a deep neural network-based method (SEV-DDI: Severity-Drug–Drug Interaction) with some further-integrated units/layers to achieve higher precision and accuracy. After successfully outperforming other methods in the DDI classification task, we moved a step further and utilized the methods in a sentiment analysis task to investigate the severity of an interaction. The ability to determine the severity of a DDI will be very helpful for clinical decision support systems in making more accurate and informed decisions, ensuring the safety of the patients.

Fast identification and quantification of c-Fos protein using you-only-look-once-v5

In neuroscience, protein activity characterizes neuronal excitability in response to a diverse array of external stimuli and represents the cell state throughout the development of brain diseases. Importantly, it is necessary to characterize the proteins involved in disease progression, nuclear function determination, stimulation method effect, and other aspects. Therefore, the quantification of protein activity is indispensable in neuroscience. Currently, ImageJ software and manual counting are two of the most commonly used methods to quantify proteins. To improve the efficiency of quantitative protein statistics, the you-only-look-once-v5 (YOLOv5) model was proposed. In this study, c-Fos immunofluorescence images data set as an example to verify the efficacy of the system using protein quantitative statistics. The results indicate that YOLOv5 was less time-consuming or obtained higher accuracy than other methods (time: ImageJ software: 80.12 ± 1.67 s, manual counting: 3.41 ± 0.25 s, YOLOv5: 0.0251 ± 0.0003 s, p < 0.0001, n = 83; simple linear regression equation: ImageJ software: Y = 1.013 × X + 0.776, R ² = 0.837; manual counting: Y = 1.0*X + 0, R ² = 1; YOLOv5: Y = 0.9730*X + 0.3821, R ² = 0.933, n = 130). The findings suggest that the YOLOv5 algorithm provides feasible methods for quantitative statistical analysis of proteins and has good potential for application in detecting target proteins in neuroscience.