Acad Radiol. 2024 Dec 13:S1076-6332(24)00886-9. doi: 10.1016/j.acra.2024.11.042. Online ahead of print.

ABSTRACT

RATIONALE AND OBJECTIVES: To comprehensively assess the feasibility of low-dose computed tomography (LDCT) using deep learning image reconstruction (DLIR) for evaluating pulmonary subsolid nodules, which are challenging due to their susceptibility to noise.

MATERIALS AND METHODS: Patients undergoing both standard-dose CT (SDCT) and LDCT between March and June 2023 were prospectively enrolled. LDCT images were reconstructed with high-strength DLIR (DLIR-H), medium-strength DLIR (DLIR-M), adaptive statistical iterative reconstruction-V level 50% (ASIR-V-50%), and filtered back projection (FBP); SDCT with FBP as the reference standard. Objective assessment, including image noise, contrast-to-noise ratio (CNR), and signal-to-noise ratio (SNR), and subjective assessment using five-point scales by five radiologists were performed. Detection and false-positive rate of subsolid nodules, and morphologic features of nodules were recorded.

RESULTS: 102 patients (mean age, 57.0 ± 12.3 years) with 358 subsolid nodules in SDCT were enrolled. The mean effective dose of SDCT and LDCT were 5.37 ± 0.80mSv and 0.86 ± 0.14mSv, respectively (P < 0.001). DLIR-H showed the lowest noise, highest CNRs, SNRs, and subjective scores among LDCT groups (all P < 0.001), almost approaching comparability with SDCT. The detection rates for DLIR-H, DLIR-M, ASIR-V-50%, and FBP were 76.5%, 76.3%, 83.8%, and 72.1%, respectively (P < 0.001), with false-positive rate of 2.5%, 2.2%, 8.3%, and 1.1%, respectively (P < 0.001). DLIR-H showed the highest detection rates for morphologic features (79.4%-95.2%) compared to DLIR-M (74.6%-88.9%), ASIR-V-50% (72.0%-88.4%), and FBP (66.1%-84.1%) (all P ≤ 0.001).

CONCLUSION: Sub-milliSievert LDCT with DLIR-H offers substantial dose reduction without compromising image quality. It is promising for evaluating subsolid nodules with a high detection rate and better identification of morphologic features.

PMID:39674695 | DOI:10.1016/j.acra.2024.11.042

Brief Bioinform. 2024 Nov 22;26(1):bbae647. doi: 10.1093/bib/bbae647.

ABSTRACT

RNA N$^{6}$-methyladenosine (m$^{6}$A) is a critical epigenetic modification closely related to rice growth, development, and stress response. m$^{6}$A accurate identification, directly related to precision rice breeding and improvement, is fundamental to revealing phenotype regulatory and molecular mechanisms. Faced on rice m$^{6}$A variable-length sequence, to input into the model, the maximum length padding and label encoding usually adapt to obtain the max-length padded sequence for prediction. Although this can retain complete sequence information, resulting in sparse information and invalid padding, reducing feature extraction accuracy. Simultaneously, existing rice-specific m$^{6}$A prediction methods are still at an early stage. To address these issues, we develop a new end-to-end deep learning framework, MFDm$^{6}$ARice, for predicting rice m$^{6}$A sites. In particular, to alleviate sparseness, we construct a multi-kernel feature fusion module to mine essential information in max-length padded sequences by multi-kernel feature extraction function and effectively transfer information through global-local dynamic fusion function. Concurrently, considering the complexity and computational efficiency of high-dimensional features caused by invalid padding, we design a downsampling residual feature embedding module to optimize feature space compression and achieve accurate feature expression and efficient computational performance. Experiments show that MFDm$^{6}$ARice outperforms comparison methods in cross-validation, same- and cross-species independent test sets, demonstrating good robustness and generalization. The application on maize m$^{6}$A indicates the MFDm$^{6}$ARice's scalability. Further investigations have shown that combining different kernel features, focusing on global channel-local spatial, and employing reasonable downsampling and residual connections can improve feature representation and extraction, ensure effective information transfer, and significantly enhance model performance.

PMID:39674264 | DOI:10.1093/bib/bbae647

Environ Res. 2024 Dec 12:120615. doi: 10.1016/j.envres.2024.120615. Online ahead of print.

ABSTRACT

Accurate prediction of influent parameters such as Chemical Oxygen Demand (COD) and Biochemical Oxygen Demand over five days (BOD5) is crucial for optimizing wastewater treatment processes, enhancing efficiency, and reducing costs. Traditional prediction methods struggle to capture the dynamic variations of influent parameters. Mechanistic biochemical models are unable to predict these parameters, and conventional machine learning methods show limited accuracy in forecasting key water quality indicators such as COD and BOD5. This study proposes a hybrid model that combines signal decomposition and deep learning to improve the accuracy of COD and BOD5 predictions. Additionally, a new dynamic feature selection (DFS) mechanism is introduced to optimize feature selection in real-time, reducing model redundancy and enhancing prediction stability. The model achieved R2 values of 0.88 and 0.96 for COD, and 0.75 and 0.93 for BOD5 across two wastewater treatment plants. RMSE and MAE values were significantly reduced, with decreases of 14.93% and 12.55% for COD at WWTP No. 5, and 20.89% and 20.40% for COD at WWTP No. 7. For BOD5, RMSE and MAE decreased by 3.56% and 5.28% at WWTP No. 5, and by 10.06% and 10.20% at WWTP No. 7. These results highlight the effectiveness of the proposed model and DFS mechanism in improving prediction accuracy and model performance. This approach provides valuable insights for wastewater treatment optimization and broader time series forecasting applications.

PMID:39674247 | DOI:10.1016/j.envres.2024.120615

Comput Biol Med. 2024 Dec 13;185:109538. doi: 10.1016/j.compbiomed.2024.109538. Online ahead of print.

ABSTRACT

Alzheimer's dementia (AD) is a neurodegenerative disorder that affects the central nervous system, causing the cells to stop working or die. The quality of life for individuals with AD steadily declines over time. While current treatments can relieve symptoms, a definitive cure remains elusive. However, technological advancements in machine learning (ML) and deep learning (DL) have opened up new possibilities for early AD detection. Early diagnosis is crucial, as trial drugs show promising results in patients who are diagnosed early. This study used a magnetic resonance imaging (MRI) dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI). The dataset consisted of 200 patients who were followed up at different time points and categorized as having AD (50), progressive-mild cognitive impairment to AD (50), stable-mild cognitive impairment (50), or cognitively normal (50). However, the utilization of MRI datasets poses challenges such as high dimensionality, limited training samples, and variability within and between subjects. To overcome these challenges, I propose using convolutional neural networks (CNNs) to extract informative features from an MRI sample. I fine-tune four pretrained models (i.e., SqueezeNet-v1.1, MobileNet-v2, Xception, and Inception-v3) to generate discriminative descriptors of MRI sample characteristics. Additionally, I suggest using the 3D shearlet transform, considering the volumetric properties of MRI data. Before the transformation, I implemented preprocessing protocols such as skull stripping, normalization of image intensity, and spatial cropping. I then summarize the shearlet coefficients using texture-based techniques. Finally, I integrate both deep and shearlet-based features using discriminant correlation analysis (DCA) to yield a robust and computationally efficient classification model. I employ two classifiers, support vector machines (SVMs) and decision tree baggers (DTBs). My objective was to develop a model capable of accurately diagnosing early-stage AD that can facilitate effective intervention and management of the condition. Our feature representation demonstrated high accuracy when applied to AD datasets at three time points. Specifically, accuracies of 94.46%, 92.97%, and 95.44% were achieved 18 months, 12 months, and at the time of stable diagnosis, respectively.

PMID:39674071 | DOI:10.1016/j.compbiomed.2024.109538

Comput Biol Med. 2024 Dec 13;185:109525. doi: 10.1016/j.compbiomed.2024.109525. Online ahead of print.

ABSTRACT

This paper introduces the Spatio-Temporal Cross Network (STCNet), a novel deep learning architecture tailored for robust hand gesture recognition in surface electromyography (sEMG) across multiple subjects. We address the challenges associated with the inter-subject variability and environmental factors such as electrode shift and muscle fatigue, which traditionally undermine the robustness of gesture recognition systems. STCNet integrates a convolutional-recurrent architecture with a spatio-temporal block that extracts features over segmented time intervals, enhancing both spatial and temporal analysis. Additionally, a rolling convolution technique designed to reflect the circular band structure of the sEMG measurement device is incorporated, thus capturing the inherent spatial relationships more effectively. We further propose a subject-aware contrastive learning framework that utilizes both subject and gesture label information to align the representation of vector space. Our comprehensive experimental evaluations demonstrate the superiority of STCNet under aggregated conditions, achieving state-of-the-art performance on benchmark datasets and effectively managing the variability among different subjects. The implemented code can be found at https://github.com/KNU-BrainAI/STCNet.

PMID:39674068 | DOI:10.1016/j.compbiomed.2024.109525

Gait Posture. 2024 Dec 11;117:58-64. doi: 10.1016/j.gaitpost.2024.12.014. Online ahead of print.

ABSTRACT

BACKGROUND: Electromyographic (EMG) recordings indicate that both the flexor digitorum brevis and soleus muscles contribute significantly to the control of standing balance, However, less is known about the adjustments in EMG activity of these two muscles across different postures.

RESEARCH QUESTION: The purpose of our study was to use deep-learning models to distinguish between the EMG activity of the flexor digitorum brevis and soleus muscles across four standing postures.

METHODS: Deep convolutional neural networks were employed to classify standing postures based on the temporal and spatial features embedded in high-density surface EMG signals. The EMG recordings were obtained with grid electrodes placed over the flexor digitorum brevis and soleus muscles of healthy young men during four standing tasks: bipedal, tandem, one-leg, and tip-toe.

RESULTS AND SIGNIFICANCE: Two-way repeated-measures analysis of variance demonstrated that the model achieved significantly greater classification accuracy, particularly during tandem stance, using EMG data from flexor digitorum brevis compared with soleus muscle. Average classification accuracy was 84.6 % for flexor digitorum brevis and 79.1 % for soleus. The classification accuracy of both muscles varied across the four postures. There were significant differences in classification accuracy for flexor digitorum brevis between bipedal and tandem stances compared with one-leg and tip-toe stances. In contrast, the EMG data for soleus were only significantly different between bipedal stance and one-leg stance. These findings indicate that flexor digitorum brevis exhibited more distinct adjustments than soleus in the temporo-spatial features of EMG activity across the four postures.

PMID:39674063 | DOI:10.1016/j.gaitpost.2024.12.014

Comput Biol Chem. 2024 Dec 11;115:108310. doi: 10.1016/j.compbiolchem.2024.108310. Online ahead of print.

ABSTRACT

Binding affinity prediction has been considered as a fundamental task in drug discovery. Despite much effort to improve accuracy of binding affinity prediction, the prior work considered only macro-level features that can represent the characteristics of the whole architecture of a drug and a target protein, and the features from local structure of the drug and the protein tend to be lost. In this paper, we propose a deep learning model that can comprehensively extract the local features of both a drug and a target protein for accurate binding affinity prediction. The proposed model consists of two components named as Multi-Stream CNN and Multi-Stream GCN, each of which is responsible for capturing micro-level characteristics or local features from subsequences of a target protein sequence and subgraph of a drug molecule, respectively. Having multiple streams consisting of different numbers of layers, both the components can compute and preserve the local features with a stream consisting of a single layer. Our evaluation with two popular datasets, Davis and KIBA, demonstrates that the proposed model outperforms all the baseline models using the global features, implying that local features play significant roles of binding affinity prediction.

PMID:39674048 | DOI:10.1016/j.compbiolchem.2024.108310

Neuro Oncol. 2024 Dec 14:noae270. doi: 10.1093/neuonc/noae270. Online ahead of print.

ABSTRACT

BACKGROUND: Accurate intraoperative diagnosis is crucial for differentiating between primary CNS lymphoma (PCNSL) and other CNS entities, guiding surgical decision-making, but represents significant challenges due to overlapping histomorphological features, time constraints, and differing treatment strategies. We combined stimulated Raman histology (SRH) with deep learning to address this challenge.

METHODS: We imaged unprocessed, label-free tissue samples intraoperatively using a portable Raman scattering microscope, generating virtual H&E-like images within less than three minutes. We developed a deep learning pipeline called RapidLymphoma based on a self-supervised learning strategy to (1) detect PCNSL, (2) differentiate from other CNS entities, and (3) test the diagnostic performance in a prospective international multicenter cohort and two additional independent test cohorts. We trained on 54,000 SRH patch images sourced from surgical resections and stereotactic-guided biopsies, including various CNS neoplastic/non-neoplastic lesions. Training and test data were collected from four tertiary international medical centers. The final histopathological diagnosis served as ground-truth.

RESULTS: In the prospective test cohort of PCNSL and non-PCNSL entities (n=160), RapidLymphoma achieved an overall balanced accuracy of 97.81% ±0.91, non-inferior to frozen section analysis in detecting PCNSL (100% vs. 77.77%). The additional test cohorts (n=420, n=59) reached balanced accuracy rates of 95.44% ±0.74 and 95.57% ±2.47 in differentiating IDH-wildtype diffuse gliomas and various brain metastasis from PCNSL. Visual heatmaps revealed RapidLymphoma's capabilities to detect class-specific histomorphological key features.

CONCLUSIONS: RapidLymphoma proves reliable and valid for intraoperative PCNSL detection and differentiation from other CNS entities. It provides visual feedback within three minutes, enabling fast clinical decision-making and subsequent treatment strategy planning.

PMID:39673805 | DOI:10.1093/neuonc/noae270

Childs Nerv Syst. 2024 Dec 14;41(1):48. doi: 10.1007/s00381-024-06674-4.

ABSTRACT

PURPOSE: The treatment of hydrocephalus aims to facilitate optimal brain development and improve the overall condition of patients. To further evaluate the postoperative recovery process in individuals undergoing hydrocephalus treatment, we investigated the interplay between brain parenchymal and ventricular volumes, alongside neurocognitive parameters.

METHODS: In this study, 52 children under the age of 10 undergoing hydrocephalus treatment were included. All participants underwent T1w MR images and Gesell developmental schedule assessments. Initially, we investigated the correlation between patients' brain development and motor assessment scores. This analysis explored the association between cognition and both brain parenchymal and ventricular sizes. Furthermore, we investigated these relationships in the contexts of communicating and obstructive hydrocephalus. Finally, to quantitatively evaluate patients' brain development using more detailed texture information from imaging, we employed three different classification models for prediction. To compare their performances, we assessed these classification frameworks using a fourfold cross-validation method.

RESULTS: Leveraging the deep learning framework, both pre- and postoperative T1w MR images have demonstrated a significant predictive value in estimating patients' brain development, with the accuracy of 0.808 for postoperative images. In the statistical analysis, we identified a correlation between developmental assessments in children with communicating hydrocephalus and postoperative brain parenchymal volume.

CONCLUSION: The findings indicate that postoperative evaluation of brain development is more closely associated with brain parenchymal and ventricular volumes than the Evans index. Additionally, deep learning frameworks exhibit promising potential as effective tools for accurately predicting patients' postoperative recovery.

PMID:39673623 | DOI:10.1007/s00381-024-06674-4

Eur J Nucl Med Mol Imaging. 2024 Dec 14. doi: 10.1007/s00259-024-07032-x. Online ahead of print.

ABSTRACT

PURPOSE: Clinical whole-body (WB) PET images can be compensated for respiratory motion using data-driven gating (DDG). However, PET DDG images may still exhibit motion artefacts at the diaphragm if the CT is acquired in a different respiratory phase than the PET image. This study evaluates the combined use of PET DDG and a deep-learning model (AIR-PETCT) for elastic registration of CT (WarpCT) to the non attenuation- and non scatter-corrected PET image (PET NAC), enabling improved PET reconstruction.

METHODS: The validation cohort included 20 patients referred for clinical FDG PET/CT, undergoing two CT scans: a free respiration CTfree and an end-expiration breath-hold CTex. AIR-PETCT registered each CT to the PET NAC and PET DDG NAC images. The image quality of PET and PET DDG images reconstructed using CTs and WarpCTs was evaluated by three blinded readers. Additionally, a clinical impact cohort of 20 patients with significant "banana" artefacts from FDG, PSMA, and DOTATOC scans was assessed for image quality and tumor-to-background ratios.

RESULTS: AIR-PETCT was robust and generated consistent WarpCTs when registering different CTs to the same PET NAC. The use of WarpCT instead of CT consistently led to equivalent or improved PET image quality. The algorithm significantly reduced "banana" artefacts and improved lesion-to-background ratios around the diaphragm. The blinded clinicians clearly preferred PET DDG images reconstructed using WarpCT.

CONCLUSION: AIR-PETCT effectively reduces respiratory motion artefacts from PET images, while improving lesion contrast. The combination of PET DDG and WarpCT holds promise for clinical application, improving PET image evaluation and diagnostic confidence.

PMID:39673603 | DOI:10.1007/s00259-024-07032-x

Curr Probl Diagn Radiol. 2024 Dec 10:S0363-0188(24)00232-9. doi: 10.1067/j.cpradiol.2024.12.004. Online ahead of print.

ABSTRACT

BACKGROUND: Radiologists increasingly use artificial intelligence (AI) to enhance diagnostic accuracy and optimize workflows. However, many lack the technical skills to effectively apply machine learning (ML) and deep learning (DL) algorithms, limiting the accessibility of these methods to radiology researchers who could otherwise benefit from them. Large language models (LLMs), such as GPT-4o, may serve as virtual advisors, offering tailored algorithm recommendations for specific research needs. This study evaluates GPT-4o's effectiveness as a recommender system to enhance radiologists' understanding and implementation of AI in research.

INTERVENTION: GPT-4o was used to recommend ML and DL algorithms based on specific details provided by researchers, including dataset characteristics, modality types, data sizes, and research objectives. The model acted as a virtual advisor, guiding researchers in selecting the most appropriate models for their studies.

METHODS: The study systematically evaluated GPT-4o's recommendations for clarity, task alignment, model diversity, and baseline selection. Responses were graded to assess the model's ability to meet the needs of radiology researchers.

RESULTS: GPT-4o effectively recommended appropriate ML and DL algorithms for various radiology tasks, including segmentation, classification, and regression in medical imaging. The model suggested a diverse range of established and innovative algorithms, such as U-Net, Random Forest, Attention U-Net, and EfficientNet, aligning well with accepted practices.

CONCLUSION: GPT-4o shows promise as a valuable tool for radiologists and early career researchers by providing clear and relevant AI and ML algorithm recommendations. Its ability to bridge the knowledge gap in AI implementation could democratize access to advanced technologies, fostering innovation and improving radiology research quality. Further studies should explore integrating LLMs into routine workflows and their role in ongoing professional development.

PMID:39672727 | DOI:10.1067/j.cpradiol.2024.12.004

Magn Reson Imaging. 2024 Dec 11:110281. doi: 10.1016/j.mri.2024.110281. Online ahead of print.

ABSTRACT

Magnetic Resonance Imaging (MRI) stands out as a notable non-invasive method for medical imaging assessments, widely employed in early medical diagnoses due to its exceptional resolution in portraying soft tissue structures. However, the MRI method faces challenges with its inherently slow acquisition process, stemming from the sequential sampling in k-space and limitations in traversal speed due to physiological and hardware constraints. Compressed Sensing in MRI (CS-MRI) accelerates image acquisition by utilizing greatly under-sampled k-space information. Despite its advantages, conventional CS-MRI encounters issues such as sluggish iterations and artefacts at higher acceleration factors. Recent advancements integrate deep learning models into CS-MRI, inspired by successes in various computer vision domains. It has drawn significant attention from the MRI community because of its great potential for image reconstruction from undersampled k-space data in fast MRI. This paper proposes a lightweight Adaptive Spatial-Channel Attention EfficientNet B3-based Generative Adversarial Network (ASCA-EffNet GAN) for fast, high-quality MR image reconstruction from greatly under-sampled k-space information in CS-MRI. The proposed GAN employs a U-net generator with ASCA-based EfficientNet B3 for encoder blocks and a ResNet decoder. The discriminator is a binary classifier with ASCA-based EfficientNet B3, a fully connected layer and a sigmoid layer. The EfficientNet B3 utilizes a compound scaling strategy that achieves a balance amongst model depth, width, and resolution, resulting in optimal performance with a reduced number of parameters. Furthermore, the adaptive attention mechanisms in the proposed ASCA-EffNet GAN effectively capture spatial and channel-wise features, contributing to detailed anatomical structure reconstruction. Experimental evaluations on the dataset demonstrate ASCA-EffNet GAN's superior performance across various metrics, surpassing conventional reconstruction methods. Hence, ASCA-EffNet GAN showcases remarkable reconstruction capabilities even under high under-sampling rates, making it suitable for clinical applications.

PMID:39672285 | DOI:10.1016/j.mri.2024.110281

Comput Biol Med. 2024 Dec 12;185:109534. doi: 10.1016/j.compbiomed.2024.109534. Online ahead of print.

ABSTRACT

This systematic literature review explores the intersection of neuroscience and deep learning in the context of decoding motor imagery Electroencephalogram (EEG) signals to enhance the quality of life for individuals with motor disabilities. Currently, the most used non-invasive method for measuring brain activity is the EEG, due to its high temporal resolution, user-friendliness, and safety. A Brain Computer Interface (BCI) framework can be made using these signals which can provide a new communication channel to people that are suffering from motor disabilities or other neurological disorders. However, implementing EEG-based BCI systems in real-world scenarios for motor imagery recognition presents challenges, primarily due to the inherent variability among individuals and low signal-to-noise ratio (SNR) of EEG signals. To assist researchers in navigating this complex problem, a comprehensive review article is presented, summarizing the key findings from relevant studies since 2017. This review primarily focuses on the datasets, preprocessing methods, feature extraction techniques, and deep learning models employed by various researchers. This review aims to contribute valuable insights and serve as a resource for researchers, practitioners, and enthusiasts interested in the combination of neuroscience and deep learning, ultimately hoping to contribute to advancements that bridge the gap between the human mind and machine interfaces.

PMID:39672015 | DOI:10.1016/j.compbiomed.2024.109534

Mil Med Res. 2024 Dec 14;11(1):77. doi: 10.1186/s40779-024-00580-1.

ABSTRACT

Ovarian cancer (OC) remains one of the most lethal gynecological malignancies globally. Despite the implementation of various medical imaging approaches for OC screening, achieving accurate differential diagnosis of ovarian tumors continues to pose significant challenges due to variability in image performance, resulting in a lack of objectivity that relies heavily on the expertise of medical professionals. This challenge can be addressed through the emergence and advancement of radiomics, which enables high-throughput extraction of valuable information from conventional medical images. Furthermore, radiomics can integrate with genomics, a novel approach termed radiogenomics, which allows for a more comprehensive, precise, and personalized assessment of tumor biological features. In this review, we present an extensive overview of the application of radiomics and radiogenomics in diagnosing and predicting ovarian tumors. The findings indicate that artificial intelligence methods based on imaging can accurately differentiate between benign and malignant ovarian tumors, as well as classify their subtypes. Moreover, these methods are effective in forecasting survival rates, treatment outcomes, metastasis risk, and recurrence for patients with OC. It is anticipated that these advancements will function as decision-support tools for managing OC while contributing to the advancement of precision medicine.

PMID:39673071 | DOI:10.1186/s40779-024-00580-1

Environ Sci Pollut Res Int. 2024 Dec 14. doi: 10.1007/s11356-024-35712-6. Online ahead of print.

ABSTRACT

This study aimed to integrate game theory and deep learning algorithms with the InVEST Ecosystem Services Model (IESM) for Sediment Retention (SR) modeling in the Kasilian watershed, Iran. The Kasilian watershed is characterized by multiple sub-watersheds, which vary in their environmental conditions and SR potential, with a total of 19 sub-watersheds. The research was carried out in four phases: mapping SR using the IESM, implementing the Fallback bargaining algorithm based on game theory, applying deep learning algorithms (CNN, LSTM, RNN), and performing statistical analysis for optimal model selection. Based on the results, the analysis of geo-environmental criteria indicated that sub-watersheds with poor conditions regarding rain erosivity, soil erodibility, LS, elevation, and land use faced greater challenges in SR. Utilizing the Fallback bargaining algorithm for sub-watershed prioritization revealed that sub-watershed 5 emerged as having the highest SR potential due to high rain erosivity and a significant LS factor. Spatial SR mapping via game theory algorithm demonstrated that northern sub-watersheds in the Kasilian watershed had greater SR potential. Deep learning algorithms were also utilized for SR distribution modeling, where the RNN model was deemed optimal, yielding error statistics of MAE: 0.05, MSE: 0.04, R2: 0.79, RMSE: 0.20, and AUC: 0.97. The SR distribution patterns demonstrated that RNN and LSTM algorithms exhibited similar classification outcomes, differing from those of the CNN algorithm. The prioritization of sub-watersheds using various approaches revealed that the Fallback bargaining algorithm showed a 47% similarity with the InVEST model results. In contrast, deep learning models such as CNN, LSTM, and ARANN exhibited 84%, 79%, and 79% similarity, respectively. These findings supported SR zonation maps, reinforcing that deep learning models outperformed the game theory algorithm. The Alpha Diversity Indices (ADI) confirmed that the outputs from the LSTM and RNN models showed identical changes across all indices. Minimal variations in the other approaches suggested that all five methods yielded similar results based on diversity indices (including Taxa, Dominance, Simpson, and Equitability), indicating no significant differences among them when compared to the InVEST model in sediment modeling.

PMID:39673030 | DOI:10.1007/s11356-024-35712-6

J Imaging Inform Med. 2024 Dec 13. doi: 10.1007/s10278-024-01344-y. Online ahead of print.

ABSTRACT

Point-of-care ultrasound (POCUS) stands as a safe, portable, and cost-effective imaging modality for swift bedside patient examinations. Specifically, lung ultrasonography (LUS) has proven useful in evaluating both acute and chronic pulmonary conditions. Despite its clinical value, automatic LUS interpretation remains relatively unexplored, particularly in multi-label contexts. This work proposes a novel deep learning (DL) framework tailored for interpreting lung POCUS videos, whose outputs are the finding(s) present in these videos (such as A-lines, B-lines, or consolidations). The pipeline, based on a residual (2+1)D architecture, initiates with a pre-processing routine for video masking and standardisation, and employs a semi-supervised approach to harness available unlabeled data. Additionally, we introduce an ensemble modeling strategy that aggregates outputs from models trained to predict distinct label sets, thereby leveraging the hierarchical nature of LUS findings. The proposed framework and its building blocks were evaluated through extensive experiments with both multi-class and multi-label models, highlighting its versatility. In a held-out test set, the categorical proposal, suited for expedite triage, achieved an average F1-score of 92.4%, while the multi-label proposal, helpful for patient management and referral, achieved an average F1-score of 70.5% across five relevant LUS findings. Overall, the semi-supervised methodology contributed significantly to improved performance, while the proposed hierarchy-aware ensemble provided moderate additional gains.

PMID:39673011 | DOI:10.1007/s10278-024-01344-y

J Imaging Inform Med. 2024 Dec 13. doi: 10.1007/s10278-024-01355-9. Online ahead of print.

ABSTRACT

The global burden of lung diseases is a pressing issue, particularly in developing nations with limited healthcare access. Accurate diagnosis of lung conditions is crucial for effective treatment, but diagnosing lung ailments using medical imaging techniques like chest radiograph images and CT scans is challenging due to the complex anatomical intricacies of the lungs. Deep learning methods, particularly convolutional neural networks (CNN), offer promising solutions for automated disease classification using imaging data. This research has the potential to significantly improve healthcare access in developing countries with limited medical resources, providing hope for better diagnosis and treatment of lung diseases. The study employed a diverse range of CNN models for training, including a baseline model and transfer learning models such as VGG16, VGG19, InceptionV3, and ResNet50. The models were trained using image datasets sourced from the NIH and COVID-19 repositories containing 8000 chest radiograph images depicting four lung conditions (lung opacity, COVID-19, pneumonia, and pneumothorax) and 2000 healthy chest radiograph images, with a ten-fold cross-validation approach. The VGG19-based model outperformed the baseline model in diagnosing lung diseases with an average accuracy of 0.995 and 0.996 on validation and external test datasets. The proposed model also outperformed published lung-disease prediction models; these findings underscore the superior performance of the VGG19 model compared to other architectures in accurately classifying and detecting lung diseases from chest radiograph images. This study highlights AI's potential, especially CNNs like VGG19, in improving diagnostic accuracy for lung disorders, promising better healthcare outcomes. The predictive model is available on GitHub at https://github.com/PGlab-NIPER/Lung_disease_classification .

PMID:39673008 | DOI:10.1007/s10278-024-01355-9

Eur Spine J. 2024 Dec 14. doi: 10.1007/s00586-024-08609-8. Online ahead of print.

ABSTRACT

OBJECTIVE: To construct a nomogram model based on magnetic resonance imaging (MRI) radiomics combined with clinical characteristics and evaluate its role and value in predicting the prognosis of patients with cervical spinal cord injury (cSCI).

METHODS: In this study, we assessed the prognosis of 168 cSCI patients using the American Spinal Injury Association (ASIA) scale and the Functional Independence Measure (FIM) scale. The study involved extracting radiomics features using both manually defined metrics and features derived through deep learning via transfer learning methods from MRI sequences, specifically T1-weighted and T2-weighted images (T1WI & T2WI). The feature selection was performed employing the least absolute shrinkage and selection operator (Lasso) regression across both radiomics and deep transfer learning datasets. Following this selection process, a deep learning radiomics signature was established. This signature, in conjunction with clinical data, was incorporated into a predictive model. The efficacy of the models was appraised using the area under the receiver operating characteristic curve (AUC), calibration curve and decision curve analysis (DCA) to assess their diagnostic performance.

RESULTS: Comparing the effectiveness of the models by linking the AUC of each model, we chose the best-performance radiomics model with clinical model to create the final nomogram. Our analysis revealed that, in the testing cohort, the combined model achieved an AUC of 0.979 for the ASIA and 0.947 for the FIM. The training cohort showed more promising performance, with an AUC of 0.957 for ASIA and 1.000 for FIM. Furthermore, the calibration curve showed that the predicted probability of the nomogram was consistent with the actual incidence rate and the DCA curve validated its effectiveness as a prognostic tool in a clinical setting.

CONCLUSION: We constructed a combined model that can be used to help predict the prognosis of cSCI patients with radiomics and clinical characteristics, and further provided guidance for clinical decision-making by generating a nomogram.

PMID:39672993 | DOI:10.1007/s00586-024-08609-8

Med Biol Eng Comput. 2024 Dec 14. doi: 10.1007/s11517-024-03255-0. Online ahead of print.

ABSTRACT

Echocardiography is a primary tool for cardiac diagnosis. Accurate delineation of the left ventricle is a prerequisite for echocardiography-based clinical decision-making. In this work, we propose an echocardiographic left ventricular segmentation method based on the diffusion probability model, which is named EchoSegDiff. The EchoSegDiff takes an encoder-decoder structure in the reverse diffusion process. A diffusion encoder residual block (DEResblock) based on the atrous pyramid squeeze attention (APSA) block is coined as the main module of the encoder, so that the EchoSegDiff can catch multiscale features effectively. A novel feature fusion module (FFM) is further proposed, which can adaptively fuse the features from encoder and decoder to reduce semantic gap between encoder and decoder. The proposed EchoSegDiff is validated on two publicly available echocardiography datasets. In terms of left ventricular segmentation performance, it outperforms other state-of-the-art networks. The segmentation accuracy on the two datasets reached 93.69% and 89.95%, respectively. This demonstrates the excellent potential of EchoSegDiff in the task of left ventricular segmentation in echocardiography.

PMID:39672990 | DOI:10.1007/s11517-024-03255-0

Sci Rep. 2024 Dec 13;14(1):30459. doi: 10.1038/s41598-024-80099-6.

ABSTRACT

Drought is a natural disaster that can affect a larger area over time. Damage caused by the drought can only be reduced through its accurate prediction. In this context, we proposed a hybrid stacked model for rainfall prediction, which is crucial for effective drought forecasting and management. In the first layer of stacked models, Bi-directional LSTM is used to extract the features, and then in the second layer, the LSTM model will make the predictions. The model captures complex temporal dependencies by processing multivariate time series data in both forward and backward directions using bi-directional LSTM layers. Trained with the Mean Squared Error loss and Adam optimizer, the model demonstrates improved forecasting accuracy, offering significant potential for proactive drought management.

PMID:39672936 | DOI:10.1038/s41598-024-80099-6