Heliyon. 2024 Jul 17;10(14):e34726. doi: 10.1016/j.heliyon.2024.e34726. eCollection 2024 Jul 30.

ABSTRACT

Cataracts are a leading cause of blindness worldwide, making accurate diagnosis and effective surgical planning critical. However, grading the severity of the lens nucleus is challenging because deep learning (DL) models pretrained using ImageNet perform poorly when applied directly to medical data due to the limited availability of labeled medical images and high interclass similarity. Self-supervised pretraining offers a solution by circumventing the need for cost-intensive data annotations and bridging domain disparities. In this study, to address the challenges of intelligent grading, we proposed a hybrid model called nuclear cataract mask encoder network (NCME-Net), which utilizes self-supervised pretraining for the four-class analysis of nuclear cataract severity. A total of 792 images of nuclear cataracts were categorized into the training set (533 images), the validation set (139 images), and the test set (100 images). NCME-Net achieved a diagnostic accuracy of 91.0 % on the test set, a 5.0 % improvement over the best-performing DL model (ResNet50). Experimental results demonstrate NCME-Net's ability to distinguish between cataract severities, particularly in scenarios with limited samples, making it a valuable tool for intelligently diagnosing cataracts. In addition, the effect of different self-supervised tasks on the model's ability to capture the intrinsic structure of the data was studied. Findings indicate that image restoration tasks significantly enhance semantic information extraction.

PMID:39149020 | PMC:PMC11324988 | DOI:10.1016/j.heliyon.2024.e34726

Food Res Int. 2024 Sep;192:114836. doi: 10.1016/j.foodres.2024.114836. Epub 2024 Jul 25.

ABSTRACT

The classification of carambola, also known as starfruit, according to quality parameters is usually conducted by trained human evaluators through visual inspections. This is a costly and subjective method that can generate high variability in results. As an alternative, computer vision systems (CVS) combined with deep learning (DCVS) techniques have been introduced in the industry as a powerful and an innovative tool for the rapid and non-invasive classification of fruits. However, validating the learning capability and trustworthiness of a DL model, aka black box, to obtain insights can be challenging. To reduce this gap, we propose an integrated eXplainable Artificial Intelligence (XAI) method for the classification of carambolas at different maturity stages. We compared two Residual Neural Networks (ResNet) and Visual Transformers (ViT) to identify the image regions that are enhanced by a Random Forest (RF) model, with the aim of providing more detailed information at the feature level for classifying the maturity stage. Changes in fruit colour and physicochemical data throughout the maturity stages were analysed, and the influence of these parameters on the maturity stages was evaluated using the Gradient-weighted Class Activation Mapping (Grad-CAM), the Attention Maps using RF importance. The proposed approach provides a visualization and description of the most important regions that led to the model decision, in wide visualization follows the models an importance features from RF. Our approach has promising potential for standardized and rapid carambolas classification, achieving 91 % accuracy with ResNet and 95 % with ViT, with potential application for other fruits.

PMID:39147524 | DOI:10.1016/j.foodres.2024.114836

Int J Radiat Oncol Biol Phys. 2024 Aug 13:S0360-3016(24)03172-9. doi: 10.1016/j.ijrobp.2024.07.2334. Online ahead of print.

ABSTRACT

BACKGROUND AND PURPOSE: Conventional normal tissue complication probability (NTCP) models for head and neck cancer (HNC) patients are typically based on single-value variables, which for radiation-induced xerostomia are baseline xerostomia and mean salivary gland doses. This study aims to improve the prediction of late xerostomia by utilizing 3D information from radiation dose distributions, CT imaging, organ-at-risk segmentations, and clinical variables with deep learning (DL).

MATERIALS AND METHODS: An international cohort of 1208 HNC patients from two institutes was used to train and twice validate DL models (DCNN, EfficientNet-v2, and ResNet) with 3D dose distribution, CT scan, organ-at-risk segmentations, baseline xerostomia score, sex, and age as input. The NTCP endpoint was moderate-to-severe xerostomia 12 months post-radiotherapy. The DL models' prediction performance was compared to a reference model: a recently published xerostomia NTCP model that used baseline xerostomia score and mean salivary gland doses as input. Attention maps were created to visualize the focus regions of the DL predictions. Transfer learning was conducted to improve the DL model performance on the external validation set.

RESULTS: All DL-based NTCP models showed better performance (AUCtest=0.78 - 0.79) than the reference NTCP model (AUCtest=0.74) in the independent test. Attention maps showed that the DL model focused on the major salivary glands, particularly the stem cell-rich region of the parotid glands. DL models obtained lower external validation performance (AUCexternal=0.63) than the reference model (AUCexternal=0.66). After transfer learning on a small external subset, the DL model (AUCtl, external=0.66) performed better than the reference model (AUCtl, external=0.64).

CONCLUSION: DL-based NTCP models performed better than the reference model when validated in data from the same institute. Improved performance in the external dataset was achieved with transfer learning, demonstrating the need for multicenter training data to realize generalizable DL-based NTCP models.

PMID:39147208 | DOI:10.1016/j.ijrobp.2024.07.2334

Mod Pathol. 2024 Aug 13:100591. doi: 10.1016/j.modpat.2024.100591. Online ahead of print.

ABSTRACT

Despite recent advances, the adoption of computer vision methods into clinical and commercial applications has been hampered by the limited availability of accurate ground truth tissue annotations required to train robust supervised models. Generating such ground truth can be accelerated by annotating tissue molecularly using immunofluorescence staining (IF) and mapping these annotations to a post-IF H&E (terminal H&E). Mapping the annotations between the IF and the terminal H&E increases both the scale and accuracy by which ground truth could be generated. However, discrepancies between terminal H&E and conventional H&E caused by IF tissue processing have limited this implementation. We sought to overcome this challenge and achieve compatibility between these parallel modalities using synthetic image generation, in which a cycle-consistent generative adversarial network (CycleGAN) was applied to transfer the appearance of conventional H&E such that it emulates the terminal H&E. These synthetic emulations allowed us to train a deep learning (DL) model for the segmentation of epithelium in the terminal H&E that could be validated against the IF staining of epithelial-based cytokeratins. The combination of this segmentation model with the CycleGAN stain transfer model enabled performative epithelium segmentation in conventional H&E images. The approach demonstrates that the training of accurate segmentation models for the breadth of conventional H&E data can be executed free of human-expert annotations by leveraging molecular annotation strategies such as IF, so long as the tissue impacts of the molecular annotation protocol are captured by generative models that can be deployed prior to the segmentation process.

PMID:39147031 | DOI:10.1016/j.modpat.2024.100591

Trends Cardiovasc Med. 2024 Aug 13:S1050-1738(24)00075-6. doi: 10.1016/j.tcm.2024.08.002. Online ahead of print.

ABSTRACT

Hypertrophic Cardiomyopathy (HCM) presents a complex diagnostic and prognostic challenge due to its heterogeneous phenotype and clinical course. Artificial Intelligence (AI) and Machine Learning (ML) techniques hold promise in transforming the role of Electrocardiography (ECG) in HCM diagnosis, prognosis, and management. AI, including Deep Learning (DL), enables computers to learn patterns from data, allowing for the development of models capable of analyzing ECG signals. DL models, such as convolutional neural networks, have shown promise in accurately identifying HCM-related abnormalities in ECGs, surpassing traditional diagnostic methods. In diagnosing HCM, ML models have demonstrated high accuracy in distinguishing between HCM and other cardiac conditions, even in cases with normal ECG findings. Additionally, AI models have enhanced risk assessment by predicting arrhythmic events leading to sudden cardiac death and identifying patients at risk for atrial fibrillation and heart failure. These models incorporate clinical and imaging data, offering a comprehensive evaluation of patient risk profiles. Challenges remain, including the need for larger and more diverse datasets to improve model generalizability and address imbalances inherent in rare event prediction. Nevertheless, AI-driven approaches have the potential to revolutionize HCM management by providing timely and accurate diagnoses, prognoses, and personalized treatment strategies based on individual patient risk profiles. This review explores the current landscape of AI applications in ECG analysis for HCM, focusing on advancements in AI methodologies and their specific implementation in HCM care.

PMID:39147002 | DOI:10.1016/j.tcm.2024.08.002

Arq Neuropsiquiatr. 2024 Aug;82(8):1-10. doi: 10.1055/s-0044-1788657. Epub 2024 Aug 15.

ABSTRACT

BACKGROUND: The early diagnosis of Alzheimer's disease (AD) and mild cognitive impairment (MCI) remains a significant challenge in neurology, with conventional methods often limited by subjectivity and variability in interpretation. Integrating deep learning with artificial intelligence (AI) in magnetic resonance imaging (MRI) analysis emerges as a transformative approach, offering the potential for unbiased, highly accurate diagnostic insights.

OBJECTIVE: A meta-analysis was designed to analyze the diagnostic accuracy of deep learning of MRI images on AD and MCI models.

METHODS: A meta-analysis was performed across PubMed, Embase, and Cochrane library databases following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, focusing on the diagnostic accuracy of deep learning. Subsequently, methodological quality was assessed using the QUADAS-2 checklist. Diagnostic measures, including sensitivity, specificity, likelihood ratios, diagnostic odds ratio, and area under the receiver operating characteristic curve (AUROC) were analyzed, alongside subgroup analyses for T1-weighted and non-T1-weighted MRI.

RESULTS: A total of 18 eligible studies were identified. The Spearman correlation coefficient was -0.6506. Meta-analysis showed that the combined sensitivity and specificity, positive likelihood ratio, negative likelihood ratio, and diagnostic odds ratio were 0.84, 0.86, 6.0, 0.19, and 32, respectively. The AUROC was 0.92. The quiescent point of hierarchical summary of receiver operating characteristic (HSROC) was 3.463. Notably, the images of 12 studies were acquired by T1-weighted MRI alone, and those of the other 6 were gathered by non-T1-weighted MRI alone.

CONCLUSION: Overall, deep learning of MRI for the diagnosis of AD and MCI showed good sensitivity and specificity and contributed to improving diagnostic accuracy.

PMID:39146974 | DOI:10.1055/s-0044-1788657

Am J Hum Genet. 2024 Jul 31:S0002-9297(24)00255-6. doi: 10.1016/j.ajhg.2024.07.011. Online ahead of print.

ABSTRACT

Large language models (LLMs) are generating interest in medical settings. For example, LLMs can respond coherently to medical queries by providing plausible differential diagnoses based on clinical notes. However, there are many questions to explore, such as evaluating differences between open- and closed-source LLMs as well as LLM performance on queries from both medical and non-medical users. In this study, we assessed multiple LLMs, including Llama-2-chat, Vicuna, Medllama2, Bard/Gemini, Claude, ChatGPT3.5, and ChatGPT-4, as well as non-LLM approaches (Google search and Phenomizer) regarding their ability to identify genetic conditions from textbook-like clinician questions and their corresponding layperson translations related to 63 genetic conditions. For open-source LLMs, larger models were more accurate than smaller LLMs: 7b, 13b, and larger than 33b parameter models obtained accuracy ranges from 21%-49%, 41%-51%, and 54%-68%, respectively. Closed-source LLMs outperformed open-source LLMs, with ChatGPT-4 performing best (89%-90%). Three of 11 LLMs and Google search had significant performance gaps between clinician and layperson prompts. We also evaluated how in-context prompting and keyword removal affected open-source LLM performance. Models were provided with 2 types of in-context prompts: list-type prompts, which improved LLM performance, and definition-type prompts, which did not. We further analyzed removal of rare terms from descriptions, which decreased accuracy for 5 of 7 evaluated LLMs. Finally, we observed much lower performance with real individuals' descriptions; LLMs answered these questions with a maximum 21% accuracy.

PMID:39146935 | DOI:10.1016/j.ajhg.2024.07.011

Comput Biol Med. 2024 Aug 14;180:108981. doi: 10.1016/j.compbiomed.2024.108981. Online ahead of print.

ABSTRACT

Early detection of polyps is essential to decrease colorectal cancer(CRC) incidence. Therefore, developing an efficient and accurate polyp segmentation technique is crucial for clinical CRC prevention. In this paper, we propose an end-to-end training approach for polyp segmentation that employs diffusion model. The images are considered as priors, and the segmentation is formulated as a mask generation process. In the sampling process, multiple predictions are generated for each input image using the trained model, and significant performance enhancements are achieved through the use of majority vote strategy. Four public datasets and one in-house dataset are used to train and test the model performance. The proposed method achieves mDice scores of 0.934 and 0.967 for datasets Kvasir-SEG and CVC-ClinicDB respectively. Furthermore, one cross-validation is applied to test the generalization of the proposed model, and the proposed methods outperformed previous state-of-the-art(SOTA) models to the best of our knowledge. The proposed method also significantly improves the segmentation accuracy and has strong generalization capability.

PMID:39146839 | DOI:10.1016/j.compbiomed.2024.108981

Rofo. 2024 Sep;196(9):890. doi: 10.1055/a-2315-4527. Epub 2024 Aug 15.

NO ABSTRACT

PMID:39146725 | DOI:10.1055/a-2315-4527

Med Image Anal. 2024 Aug 8;97:103293. doi: 10.1016/j.media.2024.103293. Online ahead of print.

ABSTRACT

Congenital heart disease (CHD) encompasses a spectrum of cardiovascular structural abnormalities, often requiring customized treatment plans for individual patients. Computational modeling and analysis of these unique cardiac anatomies can improve diagnosis and treatment planning and may ultimately lead to improved outcomes. Deep learning (DL) methods have demonstrated the potential to enable efficient treatment planning by automating cardiac segmentation and mesh construction for patients with normal cardiac anatomies. However, CHDs are often rare, making it challenging to acquire sufficiently large patient cohorts for training such DL models. Generative modeling of cardiac anatomies has the potential to fill this gap via the generation of virtual cohorts; however, prior approaches were largely designed for normal anatomies and cannot readily capture the significant topological variations seen in CHD patients. Therefore, we propose a type- and shape-disentangled generative approach suitable to capture the wide spectrum of cardiac anatomies observed in different CHD types and synthesize differently shaped cardiac anatomies that preserve the unique topology for specific CHD types. Our DL approach represents generic whole heart anatomies with CHD type-specific abnormalities implicitly using signed distance fields (SDF) based on CHD type diagnosis. To capture the shape-specific variations, we then learn invertible deformations to morph the learned CHD type-specific anatomies and reconstruct patient-specific shapes. After training with a dataset containing the cardiac anatomies of 67 patients spanning 6 CHD types and 14 combinations of CHD types, our method successfully captures divergent anatomical variations across different types and the meaningful intermediate CHD states across the spectrum of related CHD diagnoses. Additionally, our method demonstrates superior performance in CHD anatomy generation in terms of CHD-type correctness and shape plausibility. It also exhibits comparable generalization performance when reconstructing unseen cardiac anatomies. Moreover, our approach shows potential in augmenting image-segmentation pairs for rarer CHD types to significantly enhance cardiac segmentation accuracy for CHDs. Furthermore, it enables the generation of CHD cardiac meshes for computational simulation, facilitating a systematic examination of the impact of CHDs on cardiac functions.

PMID:39146700 | DOI:10.1016/j.media.2024.103293

J Environ Manage. 2024 Aug 14;368:122135. doi: 10.1016/j.jenvman.2024.122135. Online ahead of print.

ABSTRACT

Monitoring chlorophyll-a concentrations (Chl-a, μg·L-1) in aquatic ecosystems has attracted much attention due to its direct link to harmful algal blooms. However, there has been a lack of a cost-effective method for measuring Chl-a in small waterbodies. Inspired by the increase of smartphone photography, a Smartphone-based convolutional neural networks (CNN) framework (SCCA) was developed to estimate Chl-a in Aquatic ecosystem. To evaluate the performance of SCCA, 238 paired records (a smartphone image with a 12-color background and a measured Chl-a value) were collected from diverse aquatic ecosystems (e.g., rivers, lakes and ponds) across China in 2023. Our performance-evaluation results revealed a NS and R2 value of 0.90 and 0.94 in Chl-a estimation, demonstrating a satisfactory (NS = 0.84, R2 = 0.86) model fit in lower Chl-a (<30 μg L-1) conditions. SCCA had involved a realtime-update method with hyperparameter optimization technology. In comparison with the existing methods of measuring Chl-a, SCCA provides a useful screening tool for cost-effective measurement of Chl-a and has the potential for being an algal bloom screening means in small waterbodies, using Huajin River as a case study, especially under limited resources for water measurement. Overall, we highlight that the SCCA can be potentially integrated into a smartphone application in the future to diverse waterbodies in environmental management.

PMID:39146650 | DOI:10.1016/j.jenvman.2024.122135

Ophthalmic Epidemiol. 2024 Aug 15:1-14. doi: 10.1080/09286586.2024.2373956. Online ahead of print.

ABSTRACT

PURPOSE: Artificial intelligence (AI) has gained significant attention in ophthalmology. This paper reviews, classifies, and summarizes the research literature in this field and aims to provide readers with a detailed understanding of the current status and future directions, laying a solid foundation for further research and decision-making.

METHODS: Literature was retrieved from the Web of Science database. Bibliometric analysis was performed using VOSviewer, CiteSpace, and the R package Bibliometrix.

RESULTS: The study included 3,377 publications from 4,035 institutions in 98 countries. China and the United States had the most publications. Sun Yat-sen University is a leading institution. Translational Vision Science & Technology"published the most articles, while "Ophthalmology" had the most co-citations. Among 13,145 researchers, Ting DSW had the most publications and citations. Keywords included "Deep learning," "Diabetic retinopathy," "Machine learning," and others.

CONCLUSION: The study highlights the promising prospects of AI in ophthalmology. Automated eye disease screening, particularly its core technology of retinal image segmentation and recognition, has become a research hotspot. AI is also expanding to complex areas like surgical assistance, predictive models. Multimodal AI, Generative Adversarial Networks, and ChatGPT have driven further technological innovation. However, implementing AI in ophthalmology also faces many challenges, including technical, regulatory, and ethical issues, and others. As these challenges are overcome, we anticipate more innovative applications, paving the way for more effective and safer eye disease treatments.

PMID:39146462 | DOI:10.1080/09286586.2024.2373956

PLoS One. 2024 Aug 15;19(8):e0307571. doi: 10.1371/journal.pone.0307571. eCollection 2024.

ABSTRACT

The study's primary objectives encompass the following: (i) To implement the object detection of ovarian follicles using you only look once (YOLO)v8 and subsequently segment the identified follicles using a hybrid fuzzy c-means-based active contour technique. (ii) To extract statistical features and evaluate the effectiveness of both machine learning (ML) and deep learning (DL) classifiers in detecting polycystic ovary syndrome (PCOS). The research involved a two different dataset in which dataset1 comprising both normal (N = 50) and PCOS (N = 50) subjects, dataset 2 consists of 100 normal and 100 PCOS affected subjects for classification. The YOLOv8 method was employed for follicle detection, whereas statistical features were derived using Gray-level co-occurrence matrices (GLCM). For PCOS classification, various ML models such as Random Forest (RF), k- star, and stochastic gradient descent (SGD) were employed. Additionally, pre-trained models such as MobileNet, ResNet152V2, and DenseNet121 and Vision transformer were applied for the categorization of PCOS and healthy controls. Furthermore, a custom model named Follicles Net (F-Net) was developed to enhance the performance and accuracy in PCOS classification. Remarkably, the F-Net model outperformed among all ML and DL classifiers, achieving an impressive classification accuracy of 95% for dataset1 and 97.5% for dataset2 respectively in detecting PCOS. Consequently, the custom F-Net model holds significant potential as an effective automated diagnostic tool for distinguishing between normal and PCOS.

PMID:39146307 | DOI:10.1371/journal.pone.0307571

Curr Opin Otolaryngol Head Neck Surg. 2024 Jul 25. doi: 10.1097/MOO.0000000000000999. Online ahead of print.

ABSTRACT

PURPOSE OF REVIEW: The purpose of this review is to summarize the existing literature on artificial intelligence technology utilization in laryngology, highlighting recent advances and current barriers to implementation.

RECENT FINDINGS: The volume of publications studying applications of artificial intelligence in laryngology has rapidly increased, demonstrating a strong interest in utilizing this technology. Vocal biomarkers for disease screening, deep learning analysis of videolaryngoscopy for lesion identification, and auto-segmentation of videofluoroscopy for detection of aspiration are a few of the new ways in which artificial intelligence is poised to transform clinical care in laryngology. Increasing collaboration is ongoing to establish guidelines and standards for the field to ensure generalizability.

SUMMARY: Artificial intelligence tools have the potential to greatly advance laryngology care by creating novel screening methods, improving how data-heavy diagnostics of laryngology are analyzed, and standardizing outcome measures. However, physician and patient trust in artificial intelligence must improve for the technology to be successfully implemented. Additionally, most existing studies lack large and diverse datasets, external validation, and consistent ground-truth references necessary to produce generalizable results. Collaborative, large-scale studies will fuel technological innovation and bring artificial intelligence to the forefront of patient care in laryngology.

PMID:39146248 | DOI:10.1097/MOO.0000000000000999

IEEE Trans Med Imaging. 2024 Aug 15;PP. doi: 10.1109/TMI.2024.3443292. Online ahead of print.

ABSTRACT

Accelerated MRI protocols routinely involve a predefined sampling pattern that undersamples the k-space. Finding an optimal pattern can enhance the reconstruction quality, however this optimization is a challenging task. To address this challenge, we introduce a novel deep learning framework, AutoSamp, based on variational information maximization that enables joint optimization of sampling pattern and reconstruction of MRI scans. We represent the encoder as a non-uniform Fast Fourier Transform that allows continuous optimization of k-space sample locations on a non-Cartesian plane, and the decoder as a deep reconstruction network. Experiments on public 3D acquired MRI datasets show improved reconstruction quality of the proposed AutoSamp method over the prevailing variable density and variable density Poisson disc sampling for both compressed sensing and deep learning reconstructions. We demonstrate that our data-driven sampling optimization method achieves 4.4dB, 2.0dB, 0.75dB, 0.7dB PSNR improvements over reconstruction with Poisson Disc masks for acceleration factors of R = 5, 10, 15, 25, respectively. Prospectively accelerated acquisitions with 3D FSE sequences using our optimized sampling patterns exhibit improved image quality and sharpness. Furthermore, we analyze the characteristics of the learned sampling patterns with respect to changes in acceleration factor, measurement noise, underlying anatomy, and coil sensitivities. We show that all these factors contribute to the optimization result by affecting the sampling density, k-space coverage and point spread functions of the learned sampling patterns.

PMID:39146168 | DOI:10.1109/TMI.2024.3443292

Opt Lett. 2024 Aug 15;49(16):4709-4712. doi: 10.1364/OL.528796.

ABSTRACT

Elliptical beams (EBs), an essential family of structured light, have been investigated theoretically due to their intriguing mathematical properties. However, their practical application has been significantly limited due to the inability to determine all their physical quantities, particularly the ellipticity factor, a unique parameter for EBs of different families. In this paper, to our knowledge, we proposed the first high-accuracy approach that can effectively distinguish EBs with an ellipticity factor difference of 0.01, equivalent to 99.9% field similarities. The method is based on a transformer deep learning (DL) network, and the accuracy has reached 99% for two distinct families of exemplified EBs. To prove that the high performance of this model can dramatically extend the practical aspect of EBs, we used EBs as information carriers in free-space optical communication for an image transmission task, and an error bit rate as low as 0.22% is achieved. Advancing the path of such a DL approach will facilitate the research of EBs for many practical applications such as optical imaging, optical sensing, and quantum-related systems.

PMID:39146140 | DOI:10.1364/OL.528796

Opt Lett. 2024 Aug 15;49(16):4634-4637. doi: 10.1364/OL.531552.

ABSTRACT

The accurate measurement of surface three-dimensional (3D) profile and roughness on the groove sidewalls of components is of great significance to diverse fields such as precision manufacturing, machining processes, energy transportation, medical equipment, and semiconductor industry. However, conventional optical measurement methods struggle to measure surface profiles on the sidewall of a small groove. Here, we present a deep-learning-assisted sidewall profiling white light interferometry system, which consists of a microprism-based interferometer, an optical path compensation device, and a convolutional neural network (CNN), for the accurate measurement of surface 3D profile and roughness on the sidewall of a small groove. We have demonstrated that the sidewall profiling white light interferometry system can achieve a measurement accuracy of 2.64 nm for the 3D profile on a groove sidewall. Moreover, we have demonstrated that the CNN-based single-image super-resolution (SISR) technique could improve the measurement accuracy of surface roughness by over 30%. Our system can be utilized in cases where the width of the groove is only 1 mm and beyond, limited only by the size of the microprism and the working distance of the objective used in our system.

PMID:39146122 | DOI:10.1364/OL.531552

Opt Lett. 2024 Aug 15;49(16):4481-4484. doi: 10.1364/OL.530281.

ABSTRACT

This study introduces an innovative optical coherence tomography (OCT) imaging system dedicated to high-throughput screening applications using ex vivo tissue culture. Leveraging OCT's non-invasive, high-resolution capabilities, the system is equipped with a custom-designed motorized platform and tissue detection ability for automated, successive imaging across samples. Transformer-based deep-learning segmentation algorithms further ensure robust, consistent, and efficient readouts meeting the standards for screening assays. Validated using retinal explant cultures from a mouse model of retinal degeneration, the system provides robust, rapid, reliable, unbiased, and comprehensive readouts of tissue response to treatments. This fully automated OCT-based system marks a significant advancement in tissue screening, promising to transform drug discovery, as well as other relevant research fields.

PMID:39146083 | DOI:10.1364/OL.530281

Environ Monit Assess. 2024 Aug 15;196(9):816. doi: 10.1007/s10661-024-12966-8.

ABSTRACT

Obtaining accurate cultivated land distribution data is crucial for sustainable agricultural development. The current cultivated land extraction studies mainly analyze crops on a regular shape and a small block scale. Aiming at the problem of fragmentation of plots in complexly shaped cultivated land leads to variable scales and blurred edges and the difficulty of extracting the context information by kernel convolution operation of the CNN-based model. We propose a complexly shaped farmland extraction network considering multi-scale features and edge priors (MFEPNet). Specifically, we design a context cross-attention fusion module to couple the local-global features extracted by the two-terminal path CNN-transformer network, which obtains more accurate cultivated land plot representations. This paper constructs the relation maps through a multi-scale feature reconstruction module to realize multi-scale information compensates by combining the gated weight parameter based on information entropy. Additionally, we design a texture-enhanced edge module, which uses the attention mechanism to fuse the edge information of texture feature extraction and the reconstructed feature map to enhance the edge features. In general, the network effectively reduces the influence of variable scale, blurred edges, and limited global field of view. The novel model proposed in this paper is compared with classical deep learning models such as UNet, DeeplabV3 +, DANet, PSPNet, RefineNet, SegNet, ACFNet, and OCRNet on the regular and irregular farmland datasets divided by IFLYTEK and Netherlands datasets. The experimental results show that MFEPNet achieves 92.40 % and 91.65 % MIoU on regular and irregular farmland datasets, which is better than the benchmark experimental model.

PMID:39145878 | DOI:10.1007/s10661-024-12966-8

Anal Methods. 2024 Aug 15. doi: 10.1039/d4ay01160k. Online ahead of print.

ABSTRACT

Polyimide polymer membranes have become critical materials in gas separation and storage applications due to their high selectivity and excellent permeability. However, with over 107 known types of polyimides, relying solely on experimental research means potential high-performance candidates are likely to be overlooked. This study employs a deep learning method optimized by negative correlation ensemble techniques to predict the gas permeability and selectivity of polyimide structures, enabling rapid and efficient material screening. We propose a deep neural network model based on negative correlation deep ensemble methods (DNN-NCL), using Morgan molecular fingerprints as input. The DNN-NCL model achieves an R2 value of approximately 0.95 on the test set, which is a 4% improvement over recent model performance, and effectively mitigates overfitting with a maximum discrepancy of less than 0.03 between the training and test sets. High-throughput screening of over 8 million hypothetical polymers identified hundreds of promising candidates for gas separation membranes, with 14 structures exceeding the Robeson upper bound for CO2/N2 separation. Visualization of high-throughput predictions shows that although the Robeson upper bound was never explicitly used as a model constraint, the majority of predictions are compressed below this limit, demonstrating the deep learning model's ability to reflect real-world physical conditions. Reverse analysis of model predictions using SHAP analysis achieved interpretability of the deep learning model's predictions and identified three key functional groups deemed important by the deep neural network for gas permeability: carbonyl, thiophene, and ester groups. This established a bridge between the structure and properties of polyimide materials. Additionally, we confirmed that two polyimide structures predicted by the model to have excellent CO2/N2 selectivity, namely 6-methylpyrimidin-5-amine and 1,4,5,6-tetrahydropyrimidin-2-amine, have been experimentally validated in previous studies. This research demonstrates the feasibility of using deep learning methods to explore the vast chemical space of polyimides, providing a powerful tool for discovering high-performance gas separation membranes.

PMID:39145470 | DOI:10.1039/d4ay01160k