Sci Rep. 2024 Jul 29;14(1):17447. doi: 10.1038/s41598-024-67826-9.

ABSTRACT

The bone marrow overproduces immature cells in the malignancy known as Acute Lymphoblastic Leukemia (ALL). In the United States, about 6500 occurrences of ALL are diagnosed each year in both children and adults, comprising nearly 25% of pediatric cancer cases. Recently, many computer-assisted diagnosis (CAD) systems have been proposed to aid hematologists in reducing workload, providing correct results, and managing enormous volumes of data. Traditional CAD systems rely on hematologists' expertise, specialized features, and subject knowledge. Utilizing early detection of ALL can aid radiologists and doctors in making medical decisions. In this study, Deep Dilated Residual Convolutional Neural Network (DDRNet) is presented for the classification of blood cell images, focusing on eosinophils, lymphocytes, monocytes, and neutrophils. To tackle challenges like vanishing gradients and enhance feature extraction, the model incorporates Deep Residual Dilated Blocks (DRDB) for faster convergence. Conventional residual blocks are strategically placed between layers to preserve original information and extract general feature maps. Global and Local Feature Enhancement Blocks (GLFEB) balance weak contributions from shallow layers for improved feature normalization. The global feature from the initial convolution layer, when combined with GLFEB-processed features, reinforces classification representations. The Tanh function introduces non-linearity. A Channel and Spatial Attention Block (CSAB) is integrated into the neural network to emphasize or minimize specific feature channels, while fully connected layers transform the data. The use of a sigmoid activation function concentrates on relevant features for multiclass lymphoblastic leukemia classification The model was analyzed with Kaggle dataset (16,249 images) categorized into four classes, with a training and testing ratio of 80:20. Experimental results showed that DRDB, GLFEB and CSAB blocks' feature discrimination ability boosted the DDRNet model F1 score to 0.96 with minimal computational complexity and optimum classification accuracy of 99.86% and 91.98% for training and testing data. The DDRNet model stands out from existing methods due to its high testing accuracy of 91.98%, F1 score of 0.96, minimal computational complexity, and enhanced feature discrimination ability. The strategic combination of these blocks (DRDB, GLFEB, and CSAB) are designed to address specific challenges in the classification process, leading to improved discrimination of features crucial for accurate multi-class blood cell image identification. Their effective integration within the model contributes to the superior performance of DDRNet.

PMID:39075091 | DOI:10.1038/s41598-024-67826-9

Nat Commun. 2024 Jul 29;15(1):6373. doi: 10.1038/s41467-024-50573-w.

ABSTRACT

Recent studies have highlighted the impact of both transcription and transcripts on 3D genome organization, particularly its dynamics. Here, we propose a deep learning framework, called AkitaR, that leverages both genome sequences and genome-wide RNA-DNA interactions to investigate the roles of chromatin-associated RNAs (caRNAs) on genome folding in HFFc6 cells. In order to disentangle the cis- and trans-regulatory roles of caRNAs, we have compared models with nascent transcripts, trans-located caRNAs, open chromatin data, or DNA sequence alone. Both nascent transcripts and trans-located caRNAs improve the models' predictions, especially at cell-type-specific genomic regions. Analyses of feature importance scores reveal the contribution of caRNAs at TAD boundaries, chromatin loops and nuclear sub-structures such as nuclear speckles and nucleoli to the models' predictions. Furthermore, we identify non-coding RNAs (ncRNAs) known to regulate chromatin structures, such as MALAT1 and NEAT1, as well as several new RNAs, RNY5, RPPH1, POLG-DT and THBS1-IT1, that might modulate chromatin architecture through trans-interactions in HFFc6. Our modeling also suggests that transcripts from Alus and other repetitive elements may facilitate chromatin interactions through trans R-loop formation. Our findings provide insights and generate testable hypotheses about the roles of caRNAs in shaping chromatin organization.

PMID:39075082 | DOI:10.1038/s41467-024-50573-w

J Magn Reson Imaging. 2024 Jul 29. doi: 10.1002/jmri.29543. Online ahead of print.

ABSTRACT

This comprehensive review explores the role of deep learning (DL) in glioma segmentation using multiparametric magnetic resonance imaging (MRI) data. The study surveys advanced techniques such as multiparametric MRI for capturing the complex nature of gliomas. It delves into the integration of DL with MRI, focusing on convolutional neural networks (CNNs) and their remarkable capabilities in tumor segmentation. Clinical applications of DL-based segmentation are highlighted, including treatment planning, monitoring treatment response, and distinguishing between tumor progression and pseudo-progression. Furthermore, the review examines the evolution of DL-based segmentation studies, from early CNN models to recent advancements such as attention mechanisms and transformer models. Challenges in data quality, gradient vanishing, and model interpretability are discussed. The review concludes with insights into future research directions, emphasizing the importance of addressing tumor heterogeneity, integrating genomic data, and ensuring responsible deployment of DL-driven healthcare technologies. EVIDENCE LEVEL: N/A TECHNICAL EFFICACY: Stage 2.

PMID:39074952 | DOI:10.1002/jmri.29543

Int J Biol Macromol. 2024 Jul 27:134215. doi: 10.1016/j.ijbiomac.2024.134215. Online ahead of print.

ABSTRACT

The utilization and economic benefits of biomass resources can be maximized through rational design and process optimization. In this study, an innovative approach for the simultaneous extraction of essential oil and lignin from Cinnamomum camphora leaves by deep eutectic solvent (DES) and optimization of the process parameters was achieved using deep learning tools. With the water content of 40 %, liquid-solid ratio of 9.00 mL/g, and distillation time of 51.81 min, the yields of the essential oil and lignin reached 3.15 ± 0.02 % and 9.75 ± 0.15 %, respectively. Notably, the efficiency of simultaneous extraction of essential oil improved by 23 % compared to that of traditional steam distillation. Moreover, the extraction mechanism of the process was clarified. The connection between lignin with cellulose and hemicellulose was disintegrated by the DES, resulting in lignin shedding and hence accelerating the dissolution of essential oil. Moreover, the compositions of lignin and essential oil were also identified.

PMID:39074705 | DOI:10.1016/j.ijbiomac.2024.134215

Am J Obstet Gynecol. 2024 Jul 27:S0002-9378(24)00796-8. doi: 10.1016/j.ajog.2024.07.037. Online ahead of print.

NO ABSTRACT

PMID:39074680 | DOI:10.1016/j.ajog.2024.07.037

Bone. 2024 Jul 27:117215. doi: 10.1016/j.bone.2024.117215. Online ahead of print.

ABSTRACT

Despite well-defined criteria for radiographic diagnosis of atypical femur fractures (AFFs), missed and delayed diagnosis is common. An AFF diagnostic software could provide timely AFF detection to prevent progression of incomplete or development of contralateral AFFs. In this study, we investigated the ability for an artificial intelligence (AI)-based application, using deep learning models (DLMs), particularly convolutional neural networks (CNNs), to detect AFFs from femoral radiographs. A labelled Australian dataset of pre-operative complete AFF (cAFF), incomplete AFF (iAFF), typical femoral shaft fracture (TFF), and non-fractured femoral (NFF) X-ray images in anterior-posterior view were used for training (N = 213, 49, 394, 1359, respectively). An AFFnet model was developed using a pretrained (ImageNet dataset) ResNet-50 backbone, and a novel Box Attention Guide (BAG) module to guide the model's scanning patterns to enhance its learning. All images were used to train and internally test the model using a 5-fold cross validation approach, and further validated by an external dataset. External validation of the model's performance was conducted on a Sweden dataset comprising 733 TFF and 290 AFF images. Precision, sensitivity, specificity, F1-score and AUC were measured and compared between AFFnet and a global approach with ResNet-50. Excellent diagnostic performance was recorded in both models (all AUC >0.97), however AFFnet recorded lower number of prediction errors, and improved sensitivity, F1-score and precision compared to ResNet-50 in both internal and external testing. Sensitivity in the detection of iAFF was higher for AFFnet than ResNet-50 (82 % vs 56 %). In conclusion, AFFnet achieved excellent diagnostic performance on internal and external validation, which was superior to a pre-existing model. Accurate AI-based AFF diagnostic software has the potential to improve AFF diagnosis, reduce radiologist error, and allow urgent intervention, thus improving patient outcomes.

PMID:39074569 | DOI:10.1016/j.bone.2024.117215

Expert Opin Drug Discov. 2024 Aug;19(8):933-948. doi: 10.1080/17460441.2024.2367014. Epub 2024 Jun 18.

ABSTRACT

INTRODUCTION: The transition from conventional cytotoxic chemotherapy to targeted cancer therapy with small-molecule anticancer drugs has enhanced treatment outcomes. This approach, which now dominates cancer treatment, has its advantages. Despite the regulatory approval of several targeted molecules for clinical use, challenges such as low response rates and drug resistance still persist. Conventional drug discovery methods are costly and time-consuming, necessitating more efficient approaches. The rise of artificial intelligence (AI) and access to large-scale datasets have revolutionized the field of small-molecule cancer drug discovery. Machine learning (ML), particularly deep learning (DL) techniques, enables the rapid identification and development of novel anticancer agents by analyzing vast amounts of genomic, proteomic, and imaging data to uncover hidden patterns and relationships.

AREA COVERED: In this review, the authors explore the important landmarks in the history of AI-driven drug discovery. They also highlight various applications in small-molecule cancer drug discovery, outline the challenges faced, and provide insights for future research.

EXPERT OPINION: The advent of big data has allowed AI to penetrate and enable innovations in almost every stage of medicine discovery, transforming the landscape of oncology research through the development of state-of-the-art algorithms and models. Despite challenges in data quality, model interpretability, and technical limitations, advancements promise breakthroughs in personalized and precision oncology, revolutionizing future cancer management.

PMID:39074493 | DOI:10.1080/17460441.2024.2367014

BMC Med Inform Decis Mak. 2024 Jul 29;24(1):213. doi: 10.1186/s12911-024-02614-z.

ABSTRACT

BACKGROUND: This study aims to predict the trend of procurement and storage of various blood products, as well as planning and monitoring the consumption of blood products in different centers across Iran based on artificial intelligence until the year 2027.

METHODS: This research constitutes a time-series investigation within the realm of longitudinal studies. In this study, information on the number of packed red blood cells (RBC), leukoreduced red blood cells (LR-RBC), and platelets (PLT), PLT-Apheresis, and fresh frozen plasma (FFP) was requested from all blood transfusion centers in the country and extracted using a unified protocol. After the initial examination of the information and addressing data issues and inconsistencies, the corrected data were analyzed. Both conventional and artificial intelligence approaches were used to predict each product in this study. The best model was selected based on goodness-of-fit indicators RMSE and MAPE.

RESULTS: Based on the obtained results, the FFP product will follow a relatively consistent process similar to previous years in the next five years. The PLT product is predicted to have a growing trend over the next 5 years, which applies to both the demand and supply of the product. The PLT-Apheresis product also shows a similar upward trend, albeit with a lower growth rate. The RBC product will have a constant trend over a 5-year period (long-term) according to both models, taking into account short-term changes. Similarly, there is a similar trend in LR-RBC, with the expectation that short-term pattern repetition will continue over a 5-year period (long-term). Comparing the goodness-of-fit results, the LSTM model proved to be better for predicting the dominant blood products.

CONCLUSIONS: The growth of the elderly population and diseases related to old age, and on the other hand, the trend of increasing the consumption of the product with a short lifespan (PLT) requires the activation of the management of the patient's blood, especially in relation to this product in medical centers. The trend for other products in the next five years is similar to previous years, and no growth in demand is observed. The LSTM method, considering periodic and cyclical events, has performed the prediction.

PMID:39075453 | DOI:10.1186/s12911-024-02614-z

BMC Cancer. 2024 Jul 29;24(1):910. doi: 10.1186/s12885-024-12619-6.

ABSTRACT

PURPOSE: A practical noninvasive method is needed to identify lymph node (LN) status in breast cancer patients diagnosed with a suspicious axillary lymph node (ALN) at ultrasound but a negative clinical physical examination. To predict ALN metastasis effectively and noninvasively, we developed an artificial intelligence-assisted ultrasound system and validated it in a retrospective study.

METHODS: A total of 266 patients treated with sentinel LN biopsy and ALN dissection at Peking Union Medical College & Hospital(PUMCH) between the year 2017 and 2019 were assigned to training, validation and test sets (8:1:1). A deep learning model architecture named DeepLabV3 + was used together with ResNet-101 as the backbone network to create an ultrasound image segmentation diagnosis model. Subsequently, the segmented images are classified by a Convolutional Neural Network to predict ALN metastasis.

RESULTS: The area under the receiver operating characteristic curve of the model for identifying metastasis was 0.799 (95% CI: 0.514-1.000), with good end-to-end classification accuracy of 0.889 (95% CI: 0.741-1.000). Moreover, the specificity and positive predictive value of this model was 100%, providing high accuracy for clinical diagnosis.

CONCLUSION: This model can be a direct and reliable tool for the evaluation of individual LN status. Our study focuses on predicting ALN metastasis by radiomic analysis, which can be used to guide further treatment planning in breast cancer.

PMID:39075447 | DOI:10.1186/s12885-024-12619-6

J Transl Med. 2024 Jul 29;22(1):695. doi: 10.1186/s12967-024-05421-2.

ABSTRACT

BACKGROUND: Although there has been some progress in the treatment of primary uveal melanoma (UVM), distant metastasis remains the leading cause of death in patients. Monitoring, staging, and treatment of metastatic disease have not yet reached consensus. Although more than half of metastatic tumors (62%) are diagnosed within five years after primary tumor treatment, the remainder are only detected in the following 25 years. The mechanisms of UVM metastasis and its impact on prognosis are not yet fully understood.

METHODS: scRNA-seq data of UVM samples were obtained and processed, followed by cell type identification and characterization of macrophage subpopulations. High-dimensional weighted gene co-expression network analysis (HdWGCNA) was performed to identify key gene modules associated with metastatic protective macrophages (MPMφ) in primary samples, and functional analyses were conducted. Non-negative matrix factorization (NMF) clustering and immune cell infiltration analyses were performed using the MPMφ gene signatures. Machine learning models were developed using the identified metastatic protective macrophages related genes (MPMRGs) to distinguish primary from metastatic patients. A deep learning convolutional neural network (CNN) model was constructed based on MPMRGs and cell type associations. Lastly, a prognostic model was established using the MPMRGs and validated in independent cohorts.

RESULTS: Single-cell RNA-seq analysis revealed a unique immune microenvironment landscape in primary samples compared to metastatic samples, with an enrichment of macrophage cells. Using HdWGCNA, MPMφ and marker genes were identified. Functional analysis showed an enrichment of genes related to antigen processing progress and immune response. Machine learning and deep learning models based on key genes showed significant effectiveness in distinguishing between primary and metastatic patients. The prognostic model based on key genes demonstrated substantial predictive value for the survival of UVM patients.

CONCLUSION: Our study identified key macrophage subpopulations related to metastatic samples, which have a profound impact on shaping the tumor immune microenvironment. A prognostic model based on macrophage cell genes can be used to predict the prognosis of UVM patients.

PMID:39075441 | DOI:10.1186/s12967-024-05421-2

BMC Med Inform Decis Mak. 2024 Jul 29;24(1):214. doi: 10.1186/s12911-024-02600-5.

ABSTRACT

Deep neural networks (DNN) have fundamentally revolutionized the artificial intelligence (AI) field. The transformer model is a type of DNN that was originally used for the natural language processing tasks and has since gained more and more attention for processing various kinds of sequential data, including biological sequences and structured electronic health records. Along with this development, transformer-based models such as BioBERT, MedBERT, and MassGenie have been trained and deployed by researchers to answer various scientific questions originating in the biomedical domain. In this paper, we review the development and application of transformer models for analyzing various biomedical-related datasets such as biomedical textual data, protein sequences, medical structured-longitudinal data, and biomedical images as well as graphs. Also, we look at explainable AI strategies that help to comprehend the predictions of transformer-based models. Finally, we discuss the limitations and challenges of current models, and point out emerging novel research directions.

PMID:39075407 | DOI:10.1186/s12911-024-02600-5

J Biophotonics. 2024 Jul 29:e202400151. doi: 10.1002/jbio.202400151. Online ahead of print.

ABSTRACT

Early diagnosis is crucial for effective treatment of socially significant diseases, such as type 1 diabetes mellitus (T1DM), pneumonia, and asthma. This study employs a diagnostic method based on infrared laser spectroscopy of human exhaled breath. The experimental setup comprises a quantum cascade laser, which emits in a pulsed mode with a peak power of up to 150 mW in the spectral range of 5.3-12.8 μm (780-1890 cm-1), and a Herriott multipass gas cell with a specific optical path length of 76 m. Using this setup, spectra of exhaled breath in the mid-infrared range were obtained from 165 volunteers, including healthy individuals, patients with T1DM, asthma, and pneumonia. The study proposes a hybrid approach for classifying these spectra, utilizing a variational autoencoder for dimensionality reduction and a support vector machine method for classification. The results demonstrate that the proposed hybrid approach outperforms other machine learning method combinations.

PMID:39075328 | DOI:10.1002/jbio.202400151

Invest Radiol. 2024 Jul 29. doi: 10.1097/RLI.0000000000001102. Online ahead of print.

ABSTRACT

OBJECTIVES: Deep learning (DL) studies for the detection of clinically significant prostate cancer (csPCa) on magnetic resonance imaging (MRI) often overlook potentially relevant clinical parameters such as prostate-specific antigen, prostate volume, and age. This study explored the integration of clinical parameters and MRI-based DL to enhance diagnostic accuracy for csPCa on MRI.

MATERIALS AND METHODS: We retrospectively analyzed 932 biparametric prostate MRI examinations performed for suspected csPCa (ISUP ≥2) at 2 institutions. Each MRI scan was automatically analyzed by a previously developed DL model to detect and segment csPCa lesions. Three sets of features were extracted: DL lesion suspicion levels, clinical parameters (prostate-specific antigen, prostate volume, age), and MRI-based lesion volumes for all DL-detected lesions. Six multimodal artificial intelligence (AI) classifiers were trained for each combination of feature sets, employing both early (feature-level) and late (decision-level) information fusion methods. The diagnostic performance of each model was tested internally on 20% of center 1 data and externally on center 2 data (n = 529). Receiver operating characteristic comparisons determined the optimal feature combination and information fusion method and assessed the benefit of multimodal versus unimodal analysis. The optimal model performance was compared with a radiologist using PI-RADS.

RESULTS: Internally, the multimodal AI integrating DL suspicion levels with clinical features via early fusion achieved the highest performance. Externally, it surpassed baselines using clinical parameters (0.77 vs 0.67 area under the curve [AUC], P < 0.001) and DL suspicion levels alone (AUC: 0.77 vs 0.70, P = 0.006). Early fusion outperformed late fusion in external data (0.77 vs 0.73 AUC, P = 0.005). No significant performance gaps were observed between multimodal AI and radiologist assessments (internal: 0.87 vs 0.88 AUC; external: 0.77 vs 0.75 AUC, both P > 0.05).

CONCLUSIONS: Multimodal AI (combining DL suspicion levels and clinical parameters) outperforms clinical and MRI-only AI for csPCa detection. Early information fusion enhanced AI robustness in our multicenter setting. Incorporating lesion volumes did not enhance diagnostic efficacy.

PMID:39074400 | DOI:10.1097/RLI.0000000000001102

Inorg Chem. 2024 Jul 29. doi: 10.1021/acs.inorgchem.4c02232. Online ahead of print.

ABSTRACT

Gossypol (Gsp) and antibiotics present in water bodies become organic pollutants that are harmful to human health and the ecological environment. Accurate and effective detection of these pollutants has far-reaching significance in many fields. A new three-dimensional metal-organic framework (MOF), {[Eu3(L)2(HCOO-)(H2O)3]·2H2O·2DMF}n (Eu-MOF), was synthesized from 3,5-bis(2,4-dicarboxylphenyl)nitrobenzene (H4L) ligand and Eu3+ via the solvothermal method in this paper. The Eu-MOF demonstrates strong red fluorescence and can remain stable in different pH solutions. The MOF fluorescence probe could detect organic pollutants through the "shut-off" effect, with a fast response speed and a low detection limit [Gsp, nitrofurantoin (NFT), and nitrofurazone (NFZ) for 0.43, 0.38, and 0.41 μM, respectively]. During the testing process, Eu-MOF exhibited good selectivity and recoverability. Furthermore, the mechanism of fluorescence quenching was investigated, and the recoveries were also good in real samples. This paper introduced a deep learning model to recognize the fluorescence images, a portable intelligent logic detector designed for real-time detection of Gsp by logic gate strategy, and an anticounterfeiting mark prepared based on inkjet printing. Importantly, this work provides a new way of thinking for the detection of organic pollutants in water with high sensitivity and practicality by combining the fluorescence probe with machine learning and logical judgment.

PMID:39074382 | DOI:10.1021/acs.inorgchem.4c02232

J Am Soc Mass Spectrom. 2024 Jul 29. doi: 10.1021/jasms.4c00180. Online ahead of print.

ABSTRACT

In computational proteomics, machine learning (ML) has emerged as a vital tool for enhancing data analysis. Despite significant advancements, the diversity of ML model architectures and the complexity of proteomics data present substantial challenges in the effective development and evaluation of these tools. Here, we highlight the necessity for high-quality, comprehensive data sets to train ML models and advocate for the standardization of data to support robust model development. We emphasize the instrumental role of key data sets like ProteomeTools and MassIVE-KB in advancing ML applications in proteomics and discuss the implications of data set size on model performance, highlighting that larger data sets typically yield more accurate models. To address data scarcity, we explore algorithmic strategies such as self-supervised pretraining and multitask learning. Ultimately, we hope that this discussion can serve as a call to action for the proteomics community to collaborate on data standardization and collection efforts, which are crucial for the sustainable advancement and refinement of ML methodologies in the field.

PMID:39074335 | DOI:10.1021/jasms.4c00180

Invest Radiol. 2024 Jul 30. doi: 10.1097/RLI.0000000000001107. Online ahead of print.

ABSTRACT

OBJECTIVES: Reducing gadolinium-based contrast agents to lower costs, the environmental impact of gadolinium-containing wastewater, and patient exposure is still an unresolved issue. Published methods have never been compared. The purpose of this study was to compare the performance of 2 reimplemented state-of-the-art deep learning methods (settings A and B) and a proposed method for contrast signal extraction (setting C) to synthesize artificial T1-weighted full-dose images from corresponding noncontrast and low-dose images.

MATERIALS AND METHODS: In this prospective study, 213 participants received magnetic resonance imaging of the brain between August and October 2021 including low-dose (0.02 mmol/kg) and full-dose images (0.1 mmol/kg). Fifty participants were randomly set aside as test set before training (mean age ± SD, 52.6 ± 15.3 years; 30 men). Artificial and true full-dose images were compared using a reader-based study. Two readers noted all false-positive lesions and scored the overall interchangeability in regard to the clinical conclusion. Using a 5-point Likert scale (0 being the worst), they scored the contrast enhancement of each lesion and its conformity to the respective reference in the true image.

RESULTS: The average counts of false-positives per participant were 0.33 ± 0.93, 0.07 ± 0.33, and 0.05 ± 0.22 for settings A-C, respectively. Setting C showed a significantly higher proportion of scans scored as fully or mostly interchangeable (70/100) than settings A (40/100, P < 0.001) and B (57/100, P < 0.001), and generated the smallest mean enhancement reduction of scored lesions (-0.50 ± 0.55) compared with the true images (setting A: -1.10 ± 0.98; setting B: -0.91 ± 0.67, both P < 0.001). The average scores of conformity of the lesion were 1.75 ± 1.07, 2.19 ± 1.04, and 2.48 ± 0.91 for settings A-C, respectively, with significant differences among all settings (all P < 0.001).

CONCLUSIONS: The proposed method for contrast signal extraction showed significant improvements in synthesizing postcontrast images. A relevant proportion of images showing inadequate interchangeability with the reference remains at this dosage.

PMID:39074258 | DOI:10.1097/RLI.0000000000001107

Reproduction. 2024 Jul 1:REP-24-0181. doi: 10.1530/REP-24-0181. Online ahead of print.

ABSTRACT

To enhance surgical testicular sperm retrieval outcome for men with nonobstructive azoospermia, a deep-learning model was developed to identify positive seminiferous tubules by labeling 110 images with sperm-containing tubules sampled during microdissection testicular sperm extraction as training and validation data. After training, the model achieved an average precision of 0.60.

PMID:39074049 | DOI:10.1530/REP-24-0181

IEEE Trans Neural Syst Rehabil Eng. 2024 Jul 29;PP. doi: 10.1109/TNSRE.2024.3434343. Online ahead of print.

ABSTRACT

Deep learning methods have advanced quickly in brain imaging analysis over the past few years, but they are usually restricted by the limited labeled data. Pre-trained model on unlabeled data has presented promising improvement in feature learning in many domains, such as natural language processing. However, this technique is under-explored in brain network analysis. In this paper, we focused on pre-training methods with Transformer networks to leverage existing unlabeled data for brain functional network classification. First, we proposed a Transformer-based neural network, named as BrainNPT, for brain functional network classification. The proposed method leveraged <cls> token as a classification embedding vector for the Transformer model to effectively capture the representation of brain network. Second, we proposed a pre-training framework for BrainNPT model to leverage unlabeled brain network data to learn the structure information of BFNs. The results of classification experiments demonstrated the BrainNPT model without pre-training achieved the best performance with the state-of-the-art models, and the BrainNPT model with pre-training strongly outperformed the state-of-the-art models. The pre-training BrainNPT model improved 8.75% of accuracy compared with the model without pre-training. We further compared the pre-training strategies and the data augmentation methods, analyzed the influence of the parameters of the model, and explained the trained model.

PMID:39074019 | DOI:10.1109/TNSRE.2024.3434343

IEEE Trans Pattern Anal Mach Intell. 2024 Jul 29;PP. doi: 10.1109/TPAMI.2024.3434373. Online ahead of print.

ABSTRACT

Visual segmentation seeks to partition images, video frames, or point clouds into multiple segments or groups. This technique has numerous real-world applications, such as autonomous driving, image editing, robot sensing, and medical analysis. Over the past decade, deep learning-based methods have made remarkable strides in this area. Recently, transformers, a type of neural network based on self-attention originally designed for natural language processing, have considerably surpassed previous convolutional or recurrent approaches in various vision processing tasks. Specifically, vision transformers offer robust, unified, and even simpler solutions for various segmentation tasks. This survey provides a thorough overview of transformer-based visual segmentation, summarizing recent advancements. We first review the background, encompassing problem definitions, datasets, and prior convolutional methods. Next, we summarize a meta-architecture that unifies all recent transformer-based approaches. Based on this meta-architecture, we examine various method designs, including modifications to the meta-architecture and associated applications. We also present several specific subfields, including 3D point cloud segmentation, foundation model tuning, domain-aware segmentation, efficient segmentation, and medical segmentation. Additionally, we compile and re-evaluate the reviewed methods on several well-established datasets. Finally, we identify open challenges in this field and propose directions for future research. The project page can be found at https://github.com/lxtGH/Awesome-Segmentation-With-Transformer.

PMID:39074008 | DOI:10.1109/TPAMI.2024.3434373

IEEE J Biomed Health Inform. 2024 Jul 29;PP. doi: 10.1109/JBHI.2024.3434394. Online ahead of print.

ABSTRACT

Emerging research indicates that the degenerative biomarkers associated with Alzheimer's disease (AD) exhibit a non-random distribution within the cerebral cortex, instead following the structural brain network. The alterations in brain networks occur much earlier than the onset of clinical symptoms, thereby affecting the progression of brain disease. In this context, the utilization of computational methods to ascertain the propagation patterns of neuropathological events would contribute to the comprehension of the pathophysiological mechanism involved in the evolution of AD. Despite the encouraging findings achieved by existing graph-based deep learning approaches in analyzing irregular graph data, their applications in identifying the spreading pathway of neuropathology are limited due to two disadvantages. They include (1) lack of a common brain network as an unbiased reference basis for group comparison, and (2) lack of an appropriate mechanism for the identification of propagation patterns. To this end, we propose a proof-of-concept harmonic wavelet neural network (HWNN) to predict the early stage of AD and localize disease-related significant wavelets, which can be used to characterize the spreading pathways of neuropathological events across the brain network. The extensive experiments constructed on both synthetic and real datasets demonstrate that our proposed method achieves superior performance in classification accuracy and statistical power of identifying propagation patterns, compared with other representative approaches.

PMID:39074003 | DOI:10.1109/JBHI.2024.3434394