Nat Commun. 2024 Dec 18;15(1):10692. doi: 10.1038/s41467-024-55139-4.

ABSTRACT

Optical approaches have made great strides towards the goal of high-speed, energy-efficient computing necessary for modern deep learning and AI applications. Read-in and read-out of data, however, limit the overall performance of existing approaches. This study introduces a multilayer optoelectronic computing framework that alternates between optical and optoelectronic layers to implement matrix-vector multiplications and rectified linear functions, respectively. Our framework is designed for real-time, parallelized operations, leveraging 2D arrays of LEDs and photodetectors connected via independent analog electronics. We experimentally demonstrate this approach using a system with a three-layer network with two hidden layers and operate it to recognize images from the MNIST database with a recognition accuracy of 92% and classify classes from a nonlinear spiral data with 86% accuracy. By implementing multiple layers of a deep neural network simultaneously, our approach significantly reduces the number of read-ins and read-outs required and paves the way for scalable optical accelerators requiring ultra low energy.

PMID:39695133 | DOI:10.1038/s41467-024-55139-4

Brief Bioinform. 2024 Nov 22;26(1):bbae663. doi: 10.1093/bib/bbae663.

ABSTRACT

Cell-cell communications (CCCs) involve signaling from multiple sender cells that collectively impact downstream functional processes in receiver cells. Currently, computational methods are lacking for quantifying the contribution of pairwise combinations of cell types to specific functional processes in receiver cells (e.g. target gene expression or cell states). This limitation has impeded understanding the underlying mechanisms of cancer progression and identifying potential therapeutic targets. Here, we proposed a deep learning-based method, scDCA, to decipher the dominant cell communication assembly (DCA) that have a higher impact on a particular functional event in receiver cells from single-cell RNA-seq data. Specifically, scDCA employed a multi-view graph convolution network to reconstruct the CCCs landscape at single-cell resolution, and then identified DCA by interpreting the model with the attention mechanism. Taking the samples from advanced renal cell carcinoma as a case study, the scDCA was successfully applied and validated in revealing the DCA affecting the crucial gene expression in immune cells. The scDCA was also applied and validated in revealing the DCA responsible for the variation of 14 typical functional states of malignant cells. Furthermore, the scDCA was applied and validated to explore the alteration of CCCs under clinical intervention by comparing the DCA for certain cytotoxic factors between patients with and without immunotherapy. In summary, scDCA provides a valuable and practical tool for deciphering the cell type combinations with the most dominant impact on a specific functional process of receiver cells, which is of great significance for precise cancer treatment. Our data and code are free available at a public GitHub repository: https://github.com/pengsl-lab/scDCA.git.

PMID:39694816 | DOI:10.1093/bib/bbae663

World Neurosurg. 2024 Dec 16:123567. doi: 10.1016/j.wneu.2024.123567. Online ahead of print.

ABSTRACT

OBJECTIVE: This study aims to develop a fully automated, CT-based deep learning（DL） model to segment ossified lesions of the posterior longitudinal ligament (OPLL) and to measure the thickness of the ossified material and calculate the cervical spinal cord compression factor.

MATERIALS AND METHODS: A total of 307 patients were enrolled, with 260 patients from Shanghai Changzheng Hospital, And 47 patients from the Traditional Chinese Medicine Hospital of Southwest Medical University. CT images were used to manually segment the OPLL by four experienced radiologists. The DL model employing a 3D U-Net framework was developed to segment the OPLLs. The system also measures the thickness of the ossified material at its thickest point and the diameter of the spinal canal at the corresponding level. Segmentation performance was evaluated using the Dice Similarity Coefficient (DSC), Average Surface Distance (ASD), and Intra-Class Correlation (ICC) between ground truth and segmentation volumes. Concordance between the radiologists' and the DL system's measurements of the ossified material thickness, residual spinal canal diameter at maximum compression, and cervical spinal cord compression coefficient was assessed in a randomly selected subset of 30 cases from the training set using ICCs and Bland-Altman plots.

RESULTS: The DL system demonstrated average DSC of 0.81, 0.75, and 0.71 for the training, internal validation, and external test sets, respectively. The mean ASD was 1.30 for the training set, 2.35 for the internal validation set, and 2.63 for the external test set. The intraclass correlation coefficient (ICC) values of 0.958 for the thickness of the ossified material and 0.974 for the residual canal diameter measurement.

CONCLUSIONS: The proposed DL model effectively detects and separates ossification foci in OPLL on CT images. It exhibits comparable performance to radiologists in quantifying spinal cord compression metrics.

PMID:39694139 | DOI:10.1016/j.wneu.2024.123567

Ultrasonics. 2024 Dec 14;148:107551. doi: 10.1016/j.ultras.2024.107551. Online ahead of print.

ABSTRACT

Ultrasound imaging with flexible transducers requires the knowledge of shape geometry for effective beamforming, which such geometry is variable and often unknown. The conventional iteration-based shape estimation methods estimate transducer shape with high computational expense. Although deep-learning-based methods are introduced to reduce computation time, their low shape estimation accuracy limits the practical applications. In this paper, we propose a novel deep-learning-based approach, called FlexSANet, for shape estimation in ultrasound imaging with flexible transducers, which rapidly achieves precise shape estimation and then reconstructs high-quality images. First, in-phase/quadrature (I/Q) data are demodulated from raw radio frequency (RF) data to provide comprehensive guidance for the estimation task. A sparse processing mechanism is employed to extract crucial channel signals, resulting in sparse I/Q data and reducing the estimation time. Then, a spatial-aware shape estimation network establishes a one-shot mapping between the sparse I/Q data and the flexible probe shape. Finally, the ultrasound image is reconstructed using the delay-and-sum (DAS) beamformer with estimated shape. Massive comparisons on simulation datasets and in vivo datasets demonstrate the superiority of the proposed shape estimation method in rapidly and accurately estimating the transducer shape, leading to real-time and high-quality imaging. The mean absolute error of element position in shape estimation is below 1/8 wavelengths for simulation and in vivo experiments, indicating minimal element position error. The structural similarity between the ultrasound images reconstructed with real and estimated shapes is above 0.84 for simulation experiments and 0.80 for in vivo experiments, demonstrating superior image quality. More significantly, its estimation time on CPU of only 0.12 s promises clinical application potential of flexible ultrasound transducers.

PMID:39693916 | DOI:10.1016/j.ultras.2024.107551

J Neural Eng. 2024 Dec 18. doi: 10.1088/1741-2552/ada0e5. Online ahead of print.

ABSTRACT

OBJECTIVE: Automatic detection and prediction of epilepsy are crucial for improving patient care and quality of life. However, existing methods typically focus on single-dimensional information and often confuse the periodic and aperiodic components in electrophysiological signals.

APPROACH: We propose a novel deep learning framework that integrates temporal, spatial, and frequency information of EEG signals, in which periodic and aperiodic components are separated in the frequency domain. Specifically, we calculated the periodic and aperiodic components in single channel and the synchronization index of each component between channels. A self-attention mechanism is employed to filter single-channel features by selectively focusing on the most distinguishing features. Then, a hybrid bilinear deep learning network is utilized to capture the spatiotemporal features by combining a convolutional neural network (CNN) and a long short-term memory (LSTM) network. Finally, a bilinear pooling layer is employed to extract second-order features based on interactions between these spatiotemporal features.

MAIN RESULTS: The model achieves exceptional performance，with a detection accuracy of 98.84% on the CHB-MIT dataset, and a prediction accuracy of 98.44% on CHB-MIT and 97.65% on the Kaggle dataset, both with an false positive rate (FPR) of 0.02.

SIGNIFICANCE: This work paves the way for developing real-time, wearable epilepsy prediction devices to improve patient care.

PMID:39693763 | DOI:10.1088/1741-2552/ada0e5

Sci Adv. 2024 Dec 20;10(51):eadq4324. doi: 10.1126/sciadv.adq4324. Epub 2024 Dec 18.

ABSTRACT

Protein-protein interaction (PPI) networks, where nodes represent proteins and edges depict myriad interactions among them, are fundamental to understanding the dynamics within biological systems. Despite their pivotal role in modern biology, reliably discerning patterns from these intertwined networks remains a substantial challenge. The essence of the challenge lies in holistically characterizing the relationships of each node with others in the network and effectively using this information for accurate pattern discovery. In this work, we introduce a self-supervised network embedding framework termed discriminative network embedding (DNE). Unlike conventional methods that primarily focus on direct or limited-order node proximity, DNE characterizes a node both locally and globally by harnessing the contrast between representations from neighboring and distant nodes. Our experimental results demonstrate DNE's superior performance over existing techniques across various critical network analyses, including PPI inference and the identification of protein functional modules. DNE emerges as a robust strategy for node representation in PPI networks, offering promising avenues for diverse biomedical applications.

PMID:39693438 | DOI:10.1126/sciadv.adq4324

PLoS One. 2024 Dec 18;19(12):e0314707. doi: 10.1371/journal.pone.0314707. eCollection 2024.

ABSTRACT

OBJECTIVES: Various imaging features on optical coherence tomography (OCT) are crucial for identifying and defining disease progression. Establishing a consensus on these imaging features is essential, particularly for training deep learning models for disease classification. This study aims to analyze the inter-rater reliability in labeling the quality and common imaging signatures of retinal OCT scans.

METHODS: 500 OCT scans obtained from CIRRUS HD-OCT 5000 devices were displayed at 512x1024x128 resolution on a customizable, in-house annotation software. Each patient's eye was represented by 16 random scans. Two masked reviewers independently labeled the quality and specific pathological features of each scan. Evaluated features included overall image quality, presence of fovea, and disease signatures including subretinal fluid (SRF), intraretinal fluid (IRF), drusen, pigment epithelial detachment (PED), and hyperreflective material. The raw percentage agreement and Cohen's kappa (κ) coefficient were used to evaluate concurrence between the two sets of labels.

RESULTS: Our analysis revealed κ = 0.60 for the inter-rater reliability of overall scan quality, indicating substantial agreement. In contrast, there was slight agreement in determining the cause of poor image quality (κ = 0.18). The binary determination of presence and absence of retinal disease signatures showed almost complete agreement between reviewers (κ = 0.85). Specific retinal pathologies, such as the foveal location of the scan (0.78), IRF (0.63), drusen (0.73), and PED (0.87), exhibited substantial concordance. However, less agreement was found in identifying SRF (0.52), hyperreflective dots (0.41), and hyperreflective foci (0.33).

CONCLUSIONS: Our study demonstrates significant inter-rater reliability in labeling the quality and retinal pathologies on OCT scans. While some features show stronger agreement than others, these standardized labels can be utilized to create automated machine learning tools for diagnosing retinal diseases and capturing valuable pathological features in each scan. This standardization will aid in the consistency of medical diagnoses and enhance the accessibility of OCT diagnostic tools.

PMID:39693322 | DOI:10.1371/journal.pone.0314707

Transl Vis Sci Technol. 2024 Dec 2;13(12):31. doi: 10.1167/tvst.13.12.31.

ABSTRACT

PURPOSE: Accurate diagnosis of retinal disease based on optical coherence tomography (OCT) requires scrutiny of both B-scan and en face images. The aim of this study was to investigate the effectiveness of fusing en face and B-scan images for better diagnostic performance of deep learning models.

METHODS: A multiview fusion network (MVFN) with a decision fusion module to integrate fast-axis and slow-axis B-scans and en face information was proposed and compared with five state-of-the-art methods: a model using B-scans, a model using en face imaging, a model using three-dimensional volume, and two other relevant methods. They were evaluated using the OCTA-500 public dataset and a private multicenter dataset with 2330 cases; cases from the first center were used for training and cases from the second center were used for external validation. Performance was assessed by averaged area under the curve (AUC), accuracy, sensitivity, specificity, and precision.

RESULTS: In the private external test set, our MVFN achieved the highest AUC of 0.994, significantly outperforming the other models (P < 0.01). Similarly, for the OCTA-500 public dataset, our proposed method also outperformed the other methods with the highest AUC of 0.976, further demonstrating its effectiveness. Typical cases were demonstrated using activation heatmaps to illustrate the synergy of combining en face and B-scan images.

CONCLUSIONS: The fusion of en face and B-scan information is an effective strategy for improving the diagnostic accuracy of deep learning models.

TRANSLATIONAL RELEVANCE: Multiview fusion models combining B-scan and en face images demonstrate great potential in improving AI performance for retina disease diagnosis.

PMID:39693092 | DOI:10.1167/tvst.13.12.31

ACS Sens. 2024 Dec 18. doi: 10.1021/acssensors.4c03526. Online ahead of print.

NO ABSTRACT

PMID:39693047 | DOI:10.1021/acssensors.4c03526

Discov Oncol. 2024 Dec 18;15(1):772. doi: 10.1007/s12672-024-01671-0.

ABSTRACT

OBJECTIVE: Recently, the integration of Artificial Intelligence (AI) has significantly enhanced the diagnostic accuracy in breast cancer screening. This study aims to deliver an extensive review of the advancements in AI for breast cancer diagnosis and prognosis through a bibliometric analysis.

METHODOLOGY: Therefore, this study gathered pertinent peer-reviewed research articles from the Scopus database, spanning the years 2000 to 2024. These articles were subsequently subjected to quantitative analysis and visualization through the Bibliometrix R package. Ultimately, potential areas for future research challenges were pinpointed.

RESULTS: This study analyzes the development of Artificial Intelligence (AI) research for breast cancer diagnosis and prognosis from 2000 to 2024, based on 2678 publications sourced from Scopus. A sharp rise in global publication trends is observed between 2018 and 2023, with 2023 producing 456 papers, indicating intensified academic focus. Leading contributors include ZHENG B, with 36 publications, and institutions like RADBOUD UNIVERSITY MEDICAL CENTER and the IEO EUROPEAN INSTITUTE OF ONCOLOGY IRCCS. The USA leads both in publications (473) and total citations (18,530), followed by India with 289 papers. Co-occurrence analysis shows that "mammography" (3171 occurrences) and "artificial intelligence" (1691 occurrences) are among the most frequent keywords, reflecting core themes. Co-citation network analysis identifies foundational works by authors like Lecun Y. and Simonyan K. in advancing AI applications in breast cancer. Institutional and country-level collaboration analysis reveals the USA's significant partnerships with China, the UK, and Canada, driving the global research agenda in this field.

CONCLUSION: In conclusion, this bibliometric review underscores the growing influence of AI, particularly deep learning, in breast cancer diagnosis and treatment research from 2000 to 2024. The United States leads the field in publications and collaborations, with India, Spain, and the Netherlands also making significant contributions. Key institutions and journals have driven advancements, with AI applications focusing on improving diagnostic imaging and early detection. However, challenges like data limitations, regulatory hurdles, and unequal global collaboration persist, requiring further interdisciplinary efforts to enhance AI integration in clinical practice.

PMID:39692996 | DOI:10.1007/s12672-024-01671-0

Brain Inform. 2024 Dec 18;11(1):33. doi: 10.1186/s40708-024-00244-9.

ABSTRACT

Brain age, a biomarker reflecting brain health relative to chronological age, is increasingly used in neuroimaging to detect early signs of neurodegenerative diseases and support personalized treatment plans. Two primary approaches for brain age prediction have emerged: morphometric feature extraction from MRI scans and deep learning (DL) applied to raw MRI data. However, a systematic comparison of these methods regarding performance, interpretability, and clinical utility has been limited. In this study, we present a comparative evaluation of two pipelines: one using morphometric features from FreeSurfer and the other employing 3D convolutional neural networks (CNNs). Using a multisite neuroimaging dataset, we assessed both model performance and the interpretability of predictions through eXplainable Artificial Intelligence (XAI) methods, applying SHAP to the feature-based pipeline and Grad-CAM and DeepSHAP to the CNN-based pipeline. Our results show comparable performance between the two pipelines in Leave-One-Site-Out (LOSO) validation, achieving state-of-the-art performance on the independent test set ( M A E = 3.21 with DNN and morphometric features and M A E = 3.08 with a DenseNet-121 architecture). SHAP provided the most consistent and interpretable results, while DeepSHAP exhibited greater variability. Further work is needed to assess the clinical utility of Grad-CAM. This study addresses a critical gap by systematically comparing the interpretability of multiple XAI methods across distinct brain age prediction pipelines. Our findings underscore the importance of integrating XAI into clinical practice, offering insights into how XAI outputs vary and their potential utility for clinicians.

PMID:39692946 | DOI:10.1186/s40708-024-00244-9

Brain Inform. 2024 Dec 18;11(1):32. doi: 10.1186/s40708-024-00246-7.

ABSTRACT

High-throughput mesoscopic optical imaging technology has tremendously boosted the efficiency of procuring massive mesoscopic datasets from mouse brains. Constrained by the imaging field of view, the image strips obtained by such technologies typically require further processing, such as cross-sectional stitching, artifact removal, and signal area cropping, to meet the requirements of subsequent analyse. However, obtaining a batch of raw array mouse brain data at a resolution of 0.65 × 0.65 × 3 μ m 3 can reach 220TB, and the cropping of the outer contour areas in the disjointed processing still relies on manual visual observation, which consumes substantial computational resources and labor costs. In this paper, we design an efficient deep differential guided filtering module (DDGF) by fusing multi-scale iterative differential guided filtering with deep learning, which effectively refines image details while mitigating background noise. Subsequently, by amalgamating DDGF with deep learning network, we propose a lightweight deep differential guided filtering segmentation network (DDGF-SegNet), which demonstrates robust performance on our dataset, achieving Dice of 0.92, Precision of 0.98, Recall of 0.91, and Jaccard index of 0.86. Building on the segmentation, we utilize connectivity analysis for ascertaining three-dimensional spatial orientation of each brain within the array. Furthermore, we streamline the entire processing workflow by developing an automated pipeline optimized for cluster-based message passing interface(MPI) parallel computation, which reduces the processing time for a mouse brain dataset to a mere 1.1 h, enhancing manual efficiency by 25 times and overall data processing efficiency by 2.4 times, paving the way for enhancing the efficiency of big data processing and parsing for high-throughput mesoscopic optical imaging techniques.

PMID:39692944 | DOI:10.1186/s40708-024-00246-7

Protoplasma. 2024 Dec 18. doi: 10.1007/s00709-024-02019-9. Online ahead of print.

ABSTRACT

Microscopic analyses of cytoskeleton organization are crucial for understanding various cellular activities, including cell proliferation and environmental responses in plants. Traditionally, assessments of cytoskeleton dynamics have been qualitative, relying on microscopy-assisted visual inspection. However, the transition to quantitative digital microscopy has introduced new technical challenges, with segmentation of cytoskeleton structures proving particularly demanding. In this study, we examined the utility of a deep learning-based segmentation method for accurate quantitative evaluation of cytoskeleton organization using confocal microscopic images of the cortical microtubules in tobacco BY-2 cells. The results showed that, although conventional methods sufficed for measurement of cytoskeleton angles and parallelness, the deep learning-based method significantly improved the accuracy of density measurements. To assess the versatility of the method, we extended our analysis to physiologically significant models in the context of changes in cytoskeleton density, namely Arabidopsis thaliana guard cells and zygotes. The deep learning-based method successfully improved the accuracy of cytoskeleton density measurements for quantitative evaluations of physiological changes in both stomatal movement in guard cells and intracellular polarization in elongating zygotes, confirming its utility in these applications. The results demonstrate the effectiveness of deep learning-based segmentation in providing precise and high-throughput measurements of cytoskeleton density, and has the potential to automate and expedite analyses of large-scale image datasets.

PMID:39692866 | DOI:10.1007/s00709-024-02019-9

J Acoust Soc Am. 2024 Dec 1;156(6):4073-4084. doi: 10.1121/10.0034602.

ABSTRACT

Passive acoustic monitoring (PAM) is an increasingly popular tool to study vocalising species. The amount of data generated by PAM studies calls for robust automatic classifiers. Deep learning (DL) techniques have been proven effective in identifying acoustic signals in challenging datasets, but due to their black-box nature their underlying biases are hard to quantify. This study compares human analyst annotations, a multi-hypothesis tracking (MHT) click train classifier and a DL-based acoustic classifier to classify acoustic recordings based on the presence or absence of sperm whale (Physeter macrocephalus) click trains and study the temporal and spatial distributions of the Mediterranean sperm whale subpopulation around the Balearic Islands. The MHT and DL classifiers showed agreements with human labels of 85.7% and 85.0%, respectively, on data from sites they were trained on, but both saw a drop in performance when deployed on a new site. Agreement rates between classifiers surpassed those between human experts. Modeled seasonal and diel variations in sperm whale detections for both classifiers showed compatible results, revealing an increase in occurrence and diurnal activity during the summer and autumn months. This study highlights the strengths and limitations of two automatic classification algorithms to extract biologically useful information from large acoustic datasets.

PMID:39692862 | DOI:10.1121/10.0034602

Abdom Radiol (NY). 2024 Dec 18. doi: 10.1007/s00261-024-04739-1. Online ahead of print.

ABSTRACT

BACKGROUND: When antispasmodics are unavailable, the periodically rotated overlapping parallel lines with enhanced reconstruction (PROPELLER; called BLADE by Siemens Healthineers) or half Fourier single-shot turbo spin echo (HASTE) is clinically used in gynecologic MRI. However, their imaging qualities are limited compared to Turbo Spin Echo (TSE) with antispasmodics. Even with antispasmodics, TSE can be artifact-affected, necessitating a rapid backup sequence.

PURPOSE: This study aimed to investigate the utility of HASTE with deep learning reconstruction and variable flip angle evolution (iHASTE) compared to conventional sequences with and without antispasmodics.

MATERIALS AND METHODS: This retrospective study included MRI scans without antispasmodics for 79 patients who underwent iHASTE, HASTE, and BLADE and MRI scans with antispasmodics for 79 case-control matched patients who underwent TSE. Three radiologists qualitatively evaluated image quality, robustness to artifacts, tissue contrast, and uterine lesion margins. Tissue contrast was also quantitatively evaluated.

RESULTS: Quantitative evaluations revealed that iHASTE exhibited significantly superior tissue contrast to HASTE and BLADE. Qualitative evaluations indicated that iHASTE outperformed HASTE in overall quality. Two of three radiologists judged iHASTE to be significantly superior to BLADE, while two of three judged TSE to be significantly superior to iHASTE. iHASTE demonstrated greater robustness to artifacts than both BLADE and TSE. Lesion margins in iHASTE had lower scores than BLADE and TSE.

CONCLUSION: iHASTE is a viable clinical option in patients undergoing gynecologic MRI with anti-spasmodics. iHASTE may also be considered as a useful add-on sequence in patients undergoing MRI with antispasmodics.

PMID:39692759 | DOI:10.1007/s00261-024-04739-1

Cancer Sci. 2024 Dec 18. doi: 10.1111/cas.16426. Online ahead of print.

ABSTRACT

The depth of invasion plays a critical role in predicting the prognosis of early esophageal cancer, but the reasons behind invasion and the changes occurring in invasive areas are still not well understood. This study aimed to explore the morphological differences between invasive and non-invasive areas in early esophageal cancer specimens that have undergone endoscopic submucosal dissection (ESD), using artificial intelligence (AI) to shed light on the underlying mechanisms. In this study, data from 75 patients with esophageal squamous cell carcinoma (ESCC) were analyzed and endoscopic assessments were conducted to determine submucosal (SM) invasion. An AI model, specifically a Clustering-constrained Attention Multiple Instance Learning model (CLAM), was developed to predict the depth of cancer by training on surface histological images taken from both invasive and non-invasive regions. The AI model highlighted specific image portions, or patches, which were further examined to identify morphological differences between the two types of areas. The 256-pixel AI model demonstrated an average area under the receiver operating characteristic curve (AUC) value of 0.869 and an accuracy (ACC) of 0.788. The analysis of the AI-identified patches revealed that regions with invasion (SM) exhibited greater vascularity compared with non-invasive regions (epithelial). The invasive patches were characterized by a significant increase in the number and size of blood vessels, as well as a higher count of red blood cells (all with p-values <0.001). In conclusion, this study demonstrated that AI could identify critical differences in surface histopathology between non-invasive and invasive regions, particularly highlighting a higher number and larger size of blood vessels in invasive areas.

PMID:39692707 | DOI:10.1111/cas.16426

Technol Cancer Res Treat. 2024 Jan-Dec;23:15330338241308610. doi: 10.1177/15330338241308610.

ABSTRACT

INTRODUCTION: This study aimed to devise a diagnostic algorithm, termed the Refined Radiomics and Deep Learning Features-Guided CatBoost Classifier (RRDLC-Classifier), and evaluate its efficacy in predicting pathological high-grade patterns in patients diagnosed with clinical stage I solid lung adenocarcinoma (LADC).

METHODS: In this retrospective study, a total of 371 patients diagnosed with clinical stage I solid LADC were randomly categorized into training and validation sets in a 7:3 ratio. Uni- and multivariate logistic regression analyses were performed to examine the imaging findings that can be used to predict pathological high-grade patterns meticulously. Employing redundancy and the least absolute shrinkage and selection operator regression, a radiomics model was developed. Subsequently, radiomics refinement and deep learning features were employed using a machine learning algorithm to construct the RRDLC-Classifier, which aims to predict high-grade patterns in clinical stage I solid LADC. Evaluation metrics, such as receiver operating characteristic curves, areas under the curve (AUCs), accuracy, sensitivity, and specificity, were computed for assessment.

RESULTS: The RRDLC-Classifier attained the highest AUC of 0.838 (95% confidence interval [CI]: 0.766-0.911) in predicting high-grade patterns in clinical stage I solid LADC, followed by radiomics with an AUC of 0.779 (95% CI: 0.675-0.883), and imaging findings with an AUC of 0.6 (95% CI: 0.472-0.726).

CONCLUSIONS: This study introduces the RRDLC-Classifier, a novel diagnostic algorithm that amalgamates refined radiomics and deep learning features to predict high-grade patterns in clinical stage I solid LADC. This algorithm may exhibit excellent diagnostic performance, which can facilitate its application in precision medicine.

PMID:39692551 | DOI:10.1177/15330338241308610

Chaos. 2024 Dec 1;34(12):123152. doi: 10.1063/5.0238619.

ABSTRACT

Despite recent advancements in machine learning algorithms, well-established models like the Long Short-Term Memory (LSTM) are still widely used for modeling tasks. This paper introduces an enhanced LSTM variant and explores its capabilities in multiple input single output chaotic system modeling, offering a large-scale analysis that focuses on LSTM gate-level architecture, the effects of noise, non-stationary and dynamic behavior modeling, system parameter drifts, and short- and long-term forecasting. The experimental evaluation is performed on datasets generated using MATLAB, where the Lorenz and Rössler system equations are implemented and simulated in various scenarios. The extended analysis reveals that a simplified, less complex LSTM-based architecture can be successfully employed for accurate chaotic system modeling without the need for complex deep learning methodologies. This new proposed model includes only three of the four standard LSTM gates, with other feedback modifications.

PMID:39689728 | DOI:10.1063/5.0238619

Tuberc Respir Dis (Seoul). 2024 Dec 17. doi: 10.4046/trd.2024.0062. Online ahead of print.

ABSTRACT

Thoracic radiology is a primary field where artificial intelligence (AI) has been extensively researched. Recent advancements in AI demonstrate potential improvements in radiologists' performance. AI facilitates the detection and classification of abnormalities, as well as the quantification of both normal and abnormal anatomical structures. Furthermore, it enables prognostication based on these quantitative values. In this review article, the recent achievements of AI in thoracic radiology will be reviewed, mainly focused on deep learning, and the current limitations and future directions of this cutting-edge technique will be discussed.

PMID:39689720 | DOI:10.4046/trd.2024.0062

Ann Surg Oncol. 2024 Dec 17. doi: 10.1245/s10434-024-16697-5. Online ahead of print.

ABSTRACT

MAIN OBJECTIVES: We aimed at comparing intratumoral and peritumoral deep learning, radiomics, and fusion models in predicting KRAS mutations in rectal cancer using endorectal ultrasound imaging.

METHODS: This study included 304 patients with rectal cancer from Fujian Medical University Union Hospital. The patients were randomly divided into a training group (213 patients) and a test group (91 patients) at a 7:3 ratio. Radiomics and deep learning models were established using primary tumor and peritumoral images. In the optimally performing regions-of-interest, two fusion strategies, a feature-based and a decision-based model, were employed to build the fusion models. The Shapley additive explanation (SHAP) method was used to evaluate the significance of features in the optimal radiomics, deep learning, and fusion models. The performance of each model was assessed using the area under the receiver operating characteristic curve (AUC) and decision curve analysis (DCA).

RESULTS: In the test cohort, both the radiomics and deep learning models exhibited optimal performance with a 10-pixel patch extension, yielding AUC values of 0.824 and 0.856, respectively. The feature-based DLRexpand10_FB model attained the highest AUC (0.896) across all study sets. In addition, the DLRexpand10_FB model demonstrated excellent sensitivity, specificity, and DCA. SHAP analysis underscored the deep learning feature (DL_1) as the most significant factor in the hybrid model.

CONCLUSION: The feature-based fusion model DLRexpand10_FB can be employed to predict KRAS gene mutations based on pretreatment endorectal ultrasound images of rectal cancer. The integration of peritumoral regions enhanced the predictive performance of both the radiomics and deep learning models.

PMID:39690384 | DOI:10.1245/s10434-024-16697-5