Int J Radiat Oncol Biol Phys. 2024 Dec 10:S0360-3016(24)03711-8. doi: 10.1016/j.ijrobp.2024.11.104. Online ahead of print.

ABSTRACT

PURPOSE: This study aimed to design and evaluate a prior-knowledge-guided U-Net (PK-UNet) for automatic clinical target volume (CTV) segmentation in postmastectomy radiotherapy for breast cancer.

METHODS AND MATERIALS: A total of 102 computed tomography (CT) scans from breast cancer patients who underwent postmastectomy were retrospectively collected. Of these, 80 scans were used for training with 5-fold cross-validation, and 22 scans for independent testing. The CTV included the chest wall, supraclavicular region, and axillary group III. The proposed PK-UNet method employs a two-stage auto-segmentation process. Initially, the localization network categorizes CT slices based on the anatomical information of the CTV and generates prior knowledge labels. These outputs, along with the CT images, were fed into the final segmentation network. Quantitative evaluation was conducted using the mean Dice similarity coefficient (DSC), 95% Hausdorff distance (95HD), average surface distance (ASD), surface Dice similarity coefficient (sDSC). A four-level objective scale evaluation was performed by two experienced radiation oncologists in a randomized, double-blind manner.

RESULTS: Quantitative evaluations revealed that PK-UNet significantly outperformed state-of-the-art (SOTA) segmentation methods (P < 0.01), with a mean DSC of 0.90 ± 0.02 and a 95HD of 2.82 ± 1.29 mm. The mean ASD of PK-UNet was 0.91 ± 0.22 mm and the sDSC was 0.84 ± 0.07, significantly surpassing the performance of AdwU-Net (P < 0.01) and showing comparable results to other models. Clinical evaluation confirmed the efficacy of PK-UNet, with 81.8% of the predicted contours being acceptable for clinical application. The advantages of the auto-segmentation capability of PK-UNet were most evident in the superior and inferior slices and slices with discontinuities at the junctions of different subregions. The average manual correction time was reduced to 1.02 min, compared to 18.20 min for manual contouring leading to a 94.4% reduction in working time.

CONCLUSION: This study introduced the pioneering integration of prior medical knowledge into a deep learning framework for postmastectomy radiotherapy. This strategy addresses the challenges of CTV segmentation in postmastectomy radiotherapy and improves clinical workflow efficiency.

PMID:39667584 | DOI:10.1016/j.ijrobp.2024.11.104

J Dent. 2024 Dec 10:105526. doi: 10.1016/j.jdent.2024.105526. Online ahead of print.

ABSTRACT

OBJECTIVES: Periapical lesions are not always evident on radiographic scans. Sometimes, asymptomatic or initial periapical lesions on cone-beam computed tomography (CBCT) could be missed by inexperienced dentists, especially when the scan has a large field of view and is not for endodontic treatment purposes. Previously, numerous algorithms have been introduced to assist radiographic assessment and diagnosis in the field of endodontics. This study aims to investigate the efficacy of CBCT-SAM, a new artificial intelligence (AI) model, in identifying periapical lesions on CBCT.

METHODS: Model training and validation in this study was performed using 185 CBCT scans with confirmed periapical lesions. Manual segmentation labels were prepared by a trained operator and validated by a maxillofacial radiologist. The diagnostic and segmentation performances of four AI models were evaluated and compared: CBCT-SAM, CBCT-SAM without progressive prediction refinement module(PPR), and two previously developed models: Modified U-Net and PAL-Net. Accuracy was used to evaluated the diagnostic performance of the models, and accuracy, sensitivity, specificity, precision and Dice Similarity Coefficient (DSC) were used to evaluate the models' segmentation performance.

RESULTS: CBCT-SAM achieved an average diagnostic accuracy of 98.92% ± 010.37% and an average segmentation accuracy of 99.65% ± 0.66%. The average sensitivity, specificity, precision and DSC were 72.36 ± 21.61%, 99.87% ± 0.11%, 0.73 ± 0.21 and 0.70 ± 0.19. CBCT-SAM and PAL-Net performed significantly better than Modified U-Net in segmentation accuracy (p= 0.023, p = 0.041), sensitivity (p = 0.000, p = 0.002), and DSC (p=0.001, p=0.004). There is no significant difference between CBCT-SAM, CBCT-SAM without PPR and PAL-Net. However, with PPR incorporated into the model, CBCT-SAM slightly surpassed PAL-Net in the diagnostic and segmentation tasks.

CONCLUSIONS: CBCT-SAM is capable of providing expert-level assistance in the identification of periapical lesions on CBCT.

CLINICAL SIGNIFICANCE: The application of artificial intelligence could increase dentists' chairside diagnostic accuracy and efficiency. By assisting radiographic assessment, such as periapical lesions on CBCT, it help reduce the chance of missed diagnosis by human errors and facilitates early detection and treatment of dental pathologies at the early stage.

PMID:39667487 | DOI:10.1016/j.jdent.2024.105526

J Acoust Soc Am. 2024 Dec 1;156(6):3958-3969. doi: 10.1121/10.0034599.

ABSTRACT

Hearing impairment is often characterized by poor speech-in-noise recognition. State-of-the-art laboratory-based noise-reduction technology can eliminate background sounds from a corrupted speech signal and improve intelligibility, but it can also hinder environmental sound recognition (ESR), which is essential for personal independence and safety. This paper presents a time-frequency mask, the ideal compressed mask (ICM), that aims to provide listeners with improved speech intelligibility without substantially reducing ESR. This is accomplished by limiting the maximum attenuation that the mask performs. Speech intelligibility and ESR for hearing-impaired and normal-hearing listeners were measured using stimuli that had been processed by ICMs with various levels of maximum attenuation. This processing resulted in significantly improved intelligibility while retaining high ESR performance for both types of listeners. It was also found that the same level of maximum attenuation provided the optimal balance of intelligibility and ESR for both listener types. It is argued that future deep-learning-based noise reduction algorithms may provide better outcomes by balancing the levels of the target speech and the background environmental sounds, rather than eliminating all signals except for the target speech. The ICM provides one such simple solution for frequency-domain models.

PMID:39666959 | DOI:10.1121/10.0034599

JCO Glob Oncol. 2024 Dec;10:e2400146. doi: 10.1200/GO.24.00146. Epub 2024 Dec 12.

ABSTRACT

PURPOSE: The burden of cervical cancer in India is enormous, with more than 60,000 deaths being reported in 2020. The key intervention in the WHO's global strategy for the elimination of cervical cancer is to aim for the treatment and care of 90% of women diagnosed with cervical lesions. The current screen-and-treat approach as an option for resource-limited health care systems where screening of the cervix with visual inspection with acetic acid application (VIA) is followed by immediate ablative treatment by nurses in the case of a positive test. This approach often results in overtreatment, owing to the subjective nature of the test. Unnecessary treatments can be diminished with the use of emerging computer-assisted visual evaluation technology, using artificial intelligence (AI) tool to triage VIA-positive women. The aim of this study was (1) to develop a VIA-AI tool using cervical images to identify and categorize the VIA-screen-positive areas for eligibility and suitability for ablative treatment, and (2) to understand the efficacy of the VIA-AI tool in guiding the nurses to decide on treatment eligibility in the screen-and-treat cervical screening program.

METHODS: This was an exploratory, interventional study. The VIA-AI tool was developed using deep-learning AI from the image bank collected in our previously conducted screening programs. This VIA-AI tool was then pilot-tested in an ongoing nurse-led VIA screening program.

RESULTS: A comparative assessment of the cervical features performed in all women using the VIA-AI tool showed clinical accuracy of 76%. The perceived challenge rate for false positives was 20%.

CONCLUSION: This novel cervical image-based VIA-AI algorithm showed promising results in real-life settings, and could help minimize overtreatment in single-visit VIA screening and treatment programs in resource-constrained situations.

PMID:39666915 | DOI:10.1200/GO.24.00146

Plant J. 2024 Dec 12. doi: 10.1111/tpj.17190. Online ahead of print.

ABSTRACT

Due to its excellent performance in processing large amounts of data and capturing complex non-linear relationships, deep learning has been widely applied in many fields of plant biology. Here we first review the application of deep learning in analyzing genome sequences to predict gene expression, chromatin interactions, and epigenetic features (open chromatin, transcription factor binding sites, and methylation sites) in plants. Then, current motif mining and functional component design and synthesis based on generative adversarial networks, large models, and attention mechanisms are elaborated in detail. The progress of protein structure and function prediction, genomic prediction, and large model applications based on deep learning is also discussed. Finally, this work provides prospects for the future development of deep learning in plants with regard to multiple omics data, algorithm optimization, large language models, sequence design, and intelligent breeding.

PMID:39666835 | DOI:10.1111/tpj.17190

PLoS One. 2024 Dec 12;19(12):e0315646. doi: 10.1371/journal.pone.0315646. eCollection 2024.

ABSTRACT

A general-purpose method of emphasizing abnormal lesions in chest radiographs, named EGGPALE (Extrapolative, Generative and General-Purpose Abnormal Lesion Emphasizer), is presented. The proposed EGGPALE method is composed of a flow-based generative model and L-infinity-distance-based extrapolation in a latent space. The flow-based model is trained using only normal chest radiographs, and an invertible mapping function from the image space to the latent space is determined. In the latent space, a given unseen image is extrapolated so that the image point moves away from the normal chest X-ray hyperplane. Finally, the moved point is mapped back to the image space and the corresponding emphasized image is created. The proposed method was evaluated by an image interpretation experiment with nine radiologists and 1,000 chest radiographs, of which positive suspected lung cancer cases and negative cases were validated by computed tomography examinations. The sensitivity of EGGPALE-processed images showed +0.0559 average improvement compared with that of the original images, with -0.0192 deterioration of average specificity. The area under the receiver operating characteristic curve of the ensemble of nine radiologists showed a statistically significant improvement. From these results, the feasibility of EGGPALE for enhancing abnormal lesions was validated. Our code is available at https://github.com/utrad-ical/Eggpale.

PMID:39666722 | DOI:10.1371/journal.pone.0315646

PLoS One. 2024 Dec 12;19(12):e0313871. doi: 10.1371/journal.pone.0313871. eCollection 2024.

ABSTRACT

Accurate medium- to long-term runoff forecasting is of great significance for flood control, drought mitigation, comprehensive water resource management, and ecological restoration. However, runoff formation is a complex process influenced by various natural and anthropogenic factors, resulting in nonlinearity, nonstationarity, and long prediction periods, which complicate forecasting efforts. Traditional statistical models, which primarily focus on individual runoff sequences, struggle to integrate multi-source data, limiting their predictive accuracy. This narrow approach overlooks the multifaceted variables influencing runoff, resulting in incomplete and less reliable predictions. To address these challenges, we selected and integrated Random Forest (RF), Support Vector Regression (SVR), and Multilayer Perceptron Regression (MLPR) to develop two coupled intelligent prediction models-RF-SVR and RF-MLPR-due to their complementary strengths. RF effectively removes collinear and redundant information from high-dimensional data, while SVR and MLPR handle nonlinearity and nonstationarity, offering enhanced generalization capabilities. Specifically, MLPR, with its deep learning structure, can extract more complex latent information from data, making it particularly suitable for long-term forecasting. The proposed models were tested in the Yalong River Basin (YLRB), where accurate medium- to long-term runoff forecasts are essential for ecological management, flood control, and optimal water resource allocation. The results demonstrate the following: (1) The impact of atmospheric circulation indices on YLRB runoff exhibits a one-month lag, providing crucial insights for water resource scheduling and flood prevention. (2) The coupled models effectively eliminate collinearity and redundant variables, improving prediction accuracy across all forecast periods. (3) Compared to single baseline models, the coupled models demonstrated significant performance improvements across six evaluation metrics. For instance, the RF-MLPR model achieved a 3.7%-6.5% improvement in the Nash-Sutcliffe efficiency (NSE) metric across four hydrological stations compared to the RF-SVR model. (4) Prediction accuracy decreased with longer forecast periods, with the R2 value dropping from 0.8886 for a 1-month forecast to 0.6358 for a 12-month forecast, indicating the increasing challenge of long-term predictions due to greater uncertainty and the accumulation of influencing factors over time. (5) The RF-MLPR model outperformed the RF-SVR model, demonstrating a superior ability to capture the complex, nonlinear relationships inherent in the data. For example, in terms of the R2 metric, the RF-MLPR model's performance at the Jinping hydrological station improved by 6.5% compared to the RF-SVR model. Similarly, at the Lianghekou station, for a one-month lead prediction period, the RF-MLPR model's R2 value was 7.9% higher than that of the RF-SVR model. The significance of this research lies not only in its contribution to improving hydrological prediction accuracy but also in its broader applicability. The proposed coupled prediction models provide practical tools for water resource management, flood control planning, and drought mitigation in regions with similar hydrological characteristics. Furthermore, the framework's flexibility in parameterization and its ability to integrate multi-source data offer valuable insights for interdisciplinary applications across environmental sciences, meteorology, and climate prediction, making it a globally relevant contribution to addressing water management challenges under changing climatic conditions.

PMID:39666703 | DOI:10.1371/journal.pone.0313871

Transl Vis Sci Technol. 2024 Dec 2;13(12):16. doi: 10.1167/tvst.13.12.16.

ABSTRACT

PURPOSE: Eye rubbing is considered to play a significant role in the progression of keratoconus and of corneal ectasia following refractive surgery. To our knowledge, no tool performs an objective quantitative evaluation of eye rubbing using a device that is familiar to typical patients. We introduce here an innovative solution for objectively quantifying and preventing eye rubbing. It consists of an application that uses a deep-learning artificial intelligence (AI) algorithm deployed on a smartwatch.

METHODS: A Samsung Galaxy Watch 4 smartwatch collected motion data from eye rubbing and everyday activities, including readings from the gyroscope, accelerometer, and linear acceleration sensors. The training of the model was carried out using two deep-learning algorithms, long short-term memory (LSTM) and gated recurrent unit (GRU), as well as four machine learning algorithms: random forest, K-nearest neighbors (KNN), support vector machine (SVM), and XGBoost.

RESULTS: The model achieved an accuracy of 94%. The developed application could recognize, count, and display the number of eye rubbings carried out. The GRU model and XGBoost algorithm also showed promising performance.

CONCLUSIONS: Automated detection of eye rubbing by deep-learning AI has been proven to be feasible. This approach could radically improve the management of patients with keratoconus and those undergoing refractive surgery. It could detect and quantify eye rubbing and help to reduce it by sending alerts directly to the patient.

TRANSLATIONAL RELEVANCE: This proof of concept could confirm one of the most prominent paradigms in keratoconus management, the role of abnormal eye rubbing, while providing the means to challenge or even negate it by offering the first automated and objective tool for detecting eye rubbing.

PMID:39666356 | DOI:10.1167/tvst.13.12.16

Radiol Med. 2024 Dec 12. doi: 10.1007/s11547-024-01939-z. Online ahead of print.

ABSTRACT

PURPOSE: Assessment of the stability of intracranial aneurysms is important in the clinic but remains challenging. The aim of this study was to construct a deep learning model (DLM) to identify unstable aneurysms on computed tomography angiography (CTA) images.

METHODS: The clinical data of 1041 patients with 1227 aneurysms were retrospectively analyzed from August 2011 to May 2021. Patients with aneurysms were divided into unstable (ruptured, evolving and symptomatic aneurysms) and stable (fortuitous, nonevolving and asymptomatic aneurysms) groups and randomly divided into training (833 patients with 991 aneurysms) and internal validation (208 patients with 236 aneurysms) sets. One hundred and ninety-seven patients with 229 aneurysms from another hospital were included in the external validation set. Six models based on a convolutional neural network (CNN) or logistic regression were constructed on the basis of clinical, morphological and deep learning (DL) features. The area under the curve (AUC), accuracy, sensitivity and specificity were calculated to evaluate the discriminating ability of the models.

RESULTS: The AUCs of Models A (clinical), B (morphological) and C (DL features from the CTA image) in the external validation set were 0.5706, 0.9665 and 0.8453, respectively. The AUCs of Model D (clinical and DL features), Model E (clinical and morphological features) and Model F (clinical, morphological and DL features) in the external validation set were 0.8395, 0.9597 and 0.9696, respectively.

CONCLUSIONS: The CNN-based DLM, which integrates clinical, morphological and DL features, outperforms other models in predicting IA stability. The DLM has the potential to assess IA stability and support clinical decision-making.

PMID:39666223 | DOI:10.1007/s11547-024-01939-z

J Neurol. 2024 Dec 12;272(1):37. doi: 10.1007/s00415-024-12810-6.

ABSTRACT

BACKGROUND: Acute ischemic stroke (AIS) is a major cause of morbidity and mortality, with hemorrhagic transformation (HT) further worsening outcomes. Traditional scoring systems have limited predictive accuracy for HT in AIS. Recent research has explored machine learning (ML) and deep learning (DL) algorithms for stroke management. This study evaluates and compares the effectiveness of ML and DL algorithms in predicting HT post-AIS, benchmarking them against conventional models.

METHODS: A systematic search was conducted across PubMed, Embase, Web of Science, Scopus, and IEEE, initially yielding 1421 studies. After screening, 24 studies met the inclusion criteria. The Prediction Model Risk of Bias Assessment Tool (PROBAST) was used to assess the quality of these studies, and a qualitative synthesis was performed due to heterogeneity in the study design.

RESULTS: The included studies featured diverse ML and DL algorithms, with Logistic Regression (LR), Support Vector Machine (SVM), and Random Forest (RF) being the most common. Gradient boosting (GB) showed superior performance. Median Area Under the Curve (AUC) values were 0.91 for GB, 0.83 for RF, 0.77 for LR, and 0.76 for SVM. Neural networks had a median AUC of 0.81 and convolutional neural networks (CNNs) had a median AUC of 0.91. ML techniques outperformed conventional models, particularly those integrating clinical and imaging data.

CONCLUSIONS: ML and DL models significantly surpass traditional scoring systems in predicting HT. These advanced models enhance clinical decision-making and improve patient outcomes. Future research should address data expansion, imaging protocol standardization, and model transparency to enhance stroke outcomes further.

PMID:39666168 | DOI:10.1007/s00415-024-12810-6

Insights Imaging. 2024 Dec 12;15(1):298. doi: 10.1186/s13244-024-01872-9.

ABSTRACT

OBJECTIVES: To develop and externally validate a fully automated diagnostic convolutional neural network (CNN) model for cirrhosis based on liver MRI and serum biomarkers.

METHODS: This multicenter retrospective study included consecutive patients receiving pathological evaluation of liver fibrosis stage and contrast-enhanced liver MRI between March 2010 and January 2024. On the training dataset, an MRI-based CNN model was constructed for cirrhosis against pathology, and then a combined model was developed integrating the CNN model and serum biomarkers. On the testing datasets, the area under the receiver operating characteristic curve (AUC) was computed to compare the diagnostic performance of the combined model with that of aminotransferase-to-platelet ratio index (APRI), fibrosis-4 index (FIB-4), and radiologists. The influence of potential confounders on the diagnostic performance was evaluated by subgroup analyses.

RESULTS: A total of 1315 patients (median age, 54 years; 1065 men; training, n = 840) were included, 855 (65%) with pathological cirrhosis. The CNN model was constructed on pre-contrast T1- and T2-weighted imaging, and the combined model was developed integrating the CNN model, age, and eight serum biomarkers. On the external testing dataset, the combined model achieved an AUC of 0.86, which outperformed FIB-4, APRI and two radiologists (AUC: 0.67 to 0.73, all p < 0.05). Subgroup analyses revealed comparable diagnostic performances of the combined model in patients with different sizes of focal liver lesions.

CONCLUSION: Based on pre-contrast T1- and T2-weighted imaging, age, and serum biomarkers, the combined model allowed diagnosis of cirrhosis with moderate accuracy, independent of the size of focal liver lesions.

CRITICAL RELEVANCE STATEMENT: The fully automated convolutional neural network model utilizing pre-contrast MR imaging, age and serum biomarkers demonstrated moderate accuracy, outperforming FIB-4, APRI, and radiologists, independent of size of focal liver lesions, potentially facilitating noninvasive diagnosis of cirrhosis pending further validation.

KEY POINTS: This fully automated convolutional neural network (CNN) model, using pre-contrast MRI, age, and serum biomarkers, diagnoses cirrhosis. The CNN model demonstrated an external testing dataset AUC of 0.86, independent of the size of focal liver lesions. The CNN model outperformed aminotransferase-to-platelet ratio index, fibrosis-4 index, and radiologists, potentially facilitating noninvasive diagnosis of cirrhosis.

PMID:39666107 | DOI:10.1186/s13244-024-01872-9

Insights Imaging. 2024 Dec 12;15(1):297. doi: 10.1186/s13244-024-01869-4.

ABSTRACT

INTRODUCTION: Artificial intelligence (AI) in radiology is a rapidly evolving field. In breast imaging, AI has already been applied in a real-world setting and multiple studies have been conducted in the area. The aim of this analysis is to identify the most influential publications on the topic of artificial intelligence in breast imaging.

METHODS: A retrospective bibliometric analysis was conducted on artificial intelligence in breast radiology using the Web of Science database. The search strategy involved searching for the keywords 'breast radiology' or 'breast imaging' and the various keywords associated with AI such as 'deep learning', 'machine learning,' and 'neural networks'.

RESULTS: From the top 100 list, the number of citations per article ranged from 30 to 346 (average 85). The highest cited article titled 'Artificial Neural Networks In Mammography-Application To Decision-Making In The Diagnosis Of Breast-Cancer' was published in Radiology in 1993. Eighty-three of the articles were published in the last 10 years. The journal with the greatest number of articles was Radiology (n = 22). The most common country of origin was the United States (n = 51). Commonly occurring topics published were the use of deep learning models for breast cancer detection in mammography or ultrasound, radiomics in breast cancer, and the use of AI for breast cancer risk prediction.

CONCLUSION: This study provides a comprehensive analysis of the top 100 most-cited papers on the subject of artificial intelligence in breast radiology and discusses the current most influential papers in the field.

CLINICAL RELEVANCE STATEMENT: This article provides a concise summary of the top 100 most-cited articles in the field of artificial intelligence in breast radiology. It discusses the most impactful articles and explores the recent trends and topics of research in the field.

KEY POINTS: Multiple studies have been conducted on AI in breast radiology. The most-cited article was published in the journal Radiology in 1993. This study highlights influential articles and topics on AI in breast radiology.

PMID:39666106 | DOI:10.1186/s13244-024-01869-4

Trends Mol Med. 2024 Dec 11:S1471-4914(24)00310-1. doi: 10.1016/j.molmed.2024.11.009. Online ahead of print.

ABSTRACT

Deep learning has revolutionized cancer diagnostics, shifting from pixel-based image analysis to more comprehensive, patient-centric care. This opinion article explores recent advancements in neural network architectures, highlighting their evolution in biomedical research and their impact on medical imaging interpretation and multimodal data integration. We emphasize the need for domain-specific artificial intelligence (AI) systems capable of handling complex clinical tasks, advocating for the development of multimodal large language models that can integrate diverse data sources. These models have the potential to significantly enhance the precision and efficiency of cancer diagnostics, transforming AI from a supplementary tool into a core component of clinical decision-making, ultimately improving patient outcomes and advancing cancer care.

PMID:39665958 | DOI:10.1016/j.molmed.2024.11.009

J Voice. 2024 Dec 10:S0892-1997(24)00388-6. doi: 10.1016/j.jvoice.2024.11.008. Online ahead of print.

ABSTRACT

Parkinson's disease (PD) is a neurodegenerative disorder affecting motor and non-motor functions, including speech. This study evaluates the effectiveness of the audio spectrogram transformer (AST) model in detecting PD through vocal biomarkers, hypothesizing that its self-attention mechanism would better capture PD related speech impairments compared to traditional deep learning approaches. Speech recordings from 150 participants (100 from PC-GITA: 50 PD, 50 healthy controls (HC); 50 from Italian Parkinson's voice and speech (ITA): 28 PD, 22 HC) were analyzed using the AST model and compared against established architectures including VGG16, VGG19, ResNet18, ResNet34, vision transformer, and swin transformer. Audio preprocessing included sampling rate standardization to 16 kHz and amplitude normalization. The AST model achieved superior classification performance across all datasets: 97.14% accuracy on ITA, 91.67% on Parkinson's Colombian - Grupo de Investigación en Telecomunicaciones Aplicadas (PC-GITA), and 92.73% on the combined dataset. Performance remained consistent across different speech tasks, with particularly strong results in sustained vowel analysis (precision: 0.97 ± 0.03, recall: 0.96 ± 0.03). The model demonstrated robust cross-lingual generalization, outperforming traditional architectures by 5%-10% in accuracy. These results suggest that the AST model provides a reliable, non-invasive method for PD detection through voice analysis, with strong performance across different languages and speech tasks. The model's success in cross-lingual generalization indicates potential for broader clinical application, though validation across more diverse populations is needed for clinical implementation.

PMID:39665946 | DOI:10.1016/j.jvoice.2024.11.008

Acad Radiol. 2024 Dec 10:S1076-6332(24)00882-1. doi: 10.1016/j.acra.2024.11.037. Online ahead of print.

ABSTRACT

PURPOSE: This study aimed to develop and evaluate a machine learning model combining clinical, radiomics, and deep learning features derived from PET/CT imaging to predict lymph node metastasis (LNM) in patients with non-small cell lung cancer (NSCLC). The model's interpretability was enhanced using Shapley additive explanations (SHAP).

METHODS: A total of 248 NSCLC patients who underwent preoperative PET/CT scans were included and divided into training, test, and external validation sets. Radiomics features were extracted from segmented tumor regions on PET/CT images, and deep learning features were generated using the ResNet50 architecture. Feature selection was performed using minimum-redundancy maximum-relevance (mRMR), and the least absolute shrinkage and selection operator (LASSO) algorithm. Four models-clinical, radiomics, deep learning radiomics (DL_radiomics), and combined model-were constructed using the XGBoost algorithm and evaluated based on diagnostic performance metrics, including area under the receiver operating characteristic curve (AUC), accuracy, F1 score, sensitivity, and specificity. Shapley Additive exPlanations (SHAP) was used for model interpretability.

RESULTS: The combined model achieved the highest AUC in the test set (AUC=0.853), outperforming the clinical (AUC=0.758), radiomics (AUC=0.831), and DL_radiomics (AUC=0.834) models. Decision curve analysis (DCA) demonstrated that the combined model offered greater clinical net benefits. SHAP was used for global interpretation, and the summary plot indicated that the features ct_original_glrlm_LongRunHighGrayLevelEmphasis, and pet_gradient_glcm_lmc1 were the most important for the model's predictions.

CONCLUSION: The combined model, combining clinical, radiomics, and deep learning features from PET/CT, significantly improved the accuracy of LNM prediction in NSCLC patients. SHAP-based interpretability provided valuable insights into the model's decision-making process, enhancing its potential clinical application for preoperative decision-making in NSCLC.

PMID:39665892 | DOI:10.1016/j.acra.2024.11.037

Radiol Cardiothorac Imaging. 2024 Dec;6(6):e240005. doi: 10.1148/ryct.240005.

ABSTRACT

Purpose To measure the benefit of single-phase CT, inspiratory-expiratory CT, and clinical data for convolutional neural network (CNN)-based chronic obstructive pulmonary disease (COPD) staging. Materials and Methods This retrospective study included inspiratory and expiratory lung CT images and spirometry measurements acquired between November 2007 and April 2011 from 8893 participants (mean age, 59.6 years ± 9.0 [SD]; 53.3% [4738 of 8893] male) in the COPDGene phase I cohort (ClinicalTrials.gov: NCT00608764). CNNs were trained to predict spirometry measurements (forced expiratory volume in 1 second [FEV1], FEV1 percent predicted, and ratio of FEV1 to forced vital capacity [FEV1/FVC]) using clinical data and either single-phase or multiphase CT. Spirometry predictions were then used to predict Global Initiative for Chronic Obstructive Lung Disease (GOLD) stage. Agreement between CNN-predicted and reference standard spirometry measurements and GOLD stage was assessed using intraclass correlation coefficient (ICC) and compared using bootstrapping. Accuracy for predicting GOLD stage, within-one GOLD stage, and GOLD 0 versus 1-4 was calculated. Results CNN-predicted and reference standard spirometry measurements showed moderate to good agreement (ICC, 0.66-0.79), which improved by inclusion of clinical data (ICC, 0.70-0.85; P ≤ .04), except for FEV1/FVC in the inspiratory-phase CNN model with clinical data (P = .35) and FEV1 in the expiratory-phase CNN model with clinical data (P = .33). Single-phase CNN accuracies for GOLD stage, within-one stage, and diagnosis ranged from 59.8% to 84.1% (682-959 of 1140), with moderate to good agreement (ICC, 0.68-0.70). Accuracies of CNN models using inspiratory and expiratory images ranged from 60.0% to 86.3% (684-984 of 1140), with moderate to good agreement (ICC, 0.72). Inclusion of clinical data improved agreement and accuracy for both the single-phase CNNs (ICC, 0.72; P ≤ .001; accuracy, 65.2%-85.8% [743-978 of 1140]) and inspiratory-expiratory CNNs (ICC, 0.77-0.78; P ≤ .001; accuracy, 67.6%-88.0% [771-1003 of 1140]), except expiratory CNN with clinical data (no change in GOLD stage ICC; P = .08). Conclusion CNN-based COPD diagnosis and staging using single-phase CT provides comparable accuracy with inspiratory-expiratory CT when provided clinical data relevant to staging. Keywords: Convolutional Neural Network, Chronic Obstructive Pulmonary Disease, CT, Severity Staging, Attention Map Supplemental material is available for this article. © RSNA, 2024.

PMID:39665633 | DOI:10.1148/ryct.240005

J Chem Phys. 2024 Dec 14;161(22):225102. doi: 10.1063/5.0241246.

ABSTRACT

Molecular dynamics simulations are crucial for understanding the structural and dynamical behavior of biomolecular systems, including the impact of their environment. However, there is a gap between the time scale of these simulations and that of real-world experiments. To address this problem, various enhanced simulation methods have been developed. In addition, there has been a significant advancement of the force fields used for simulations associated with the explicit treatment of electronic polarizability. In this study, we apply oscillating chemical potential grand canonical Monte Carlo and machine learning methods to determine reaction coordinates combined with metadynamics simulations to explore the role of Mg2+ distribution and electronic polarizability in the context of the classical Drude oscillator polarizable force field on the stability of the twister ribozyme. The introduction of electronic polarizability along with the details of the distribution of Mg2+ significantly stabilizes the simulations with respect to sampling the crystallographic conformation. The introduction of electronic polarizability leads to increased stability over that obtained with the additive CHARMM36 FF reported in a previous study, allowing for a distribution of a wider range of ions to stabilize twister. Specific interactions contributing to stabilization are identified, including both those observed in the crystal structures and additional experimentally unobserved interactions. Interactions of Mg2+ with the bases are indicated to make important contributions to stabilization. Notably, the presence of specific interactions between the Mg2+ ions and bases or the non-bridging phosphate oxygens (NBPOs) leads to enhanced dipole moments of all three moieties. Mg2+-NBPO interactions led to enhanced dipoles of the phosphates but, interestingly, not in all the participating ions. The present results further indicate the importance of electronic polarizability in stabilizing RNA in molecular simulations and the complicated nature of the relationship of Mg2+-RNA interactions with the polarization response of the bases and phosphates.

PMID:39665326 | DOI:10.1063/5.0241246

Adv Sci (Weinh). 2024 Dec 12:e2404755. doi: 10.1002/advs.202404755. Online ahead of print.

ABSTRACT

Aortic stenosis (AS) is the most common valvular heart disease in developed countries. High-fidelity preclinical models can improve AS management by enabling therapeutic innovation, early diagnosis, and tailored treatment planning. However, their use is currently limited by complex workflows necessitating lengthy expert-driven manual operations. Here, we propose an AI-powered computational framework for accelerated and democratized patient-specific modeling of AS hemodynamics from computed tomography (CT). First, we demonstrate that the automated meshing algorithms can generate task-ready geometries for both computational and benchtop simulations with higher accuracy and 100 times faster than existing approaches. Then, we show that the approach can be integrated with fluid-structure interaction and soft robotics models to accurately recapitulate a broad spectrum of clinical hemodynamic measurements of diverse AS patients. The efficiency and reliability of these algorithms make them an ideal complementary tool for personalized high-fidelity modeling of AS biomechanics, hemodynamics, and treatment planning.

PMID:39665137 | DOI:10.1002/advs.202404755

Front Plant Sci. 2024 Nov 27;15:1502314. doi: 10.3389/fpls.2024.1502314. eCollection 2024.

ABSTRACT

Accurately identifying apple diseases is essential to control their spread and support the industry. Timely and precise detection is crucial for managing the spread of diseases, thereby improving the production and quality of apples. However, the development of algorithms for analyzing complex leaf images remains a significant challenge. Therefore, in this study, a lightweight deep learning model is designed from scratch to identify the apple leaf condition. The developed framework comprises two stages. First, the designed 37-layer model was employed to assess the condition of apple leaves (healthy or diseased). Second, transfer learning was used for further subclassification of the disease class (e.g., rust, complex, scab, and frogeye leaf spots). The trained lightweight model was reused because the model trained with correlated images facilitated transfer learning for further classification of the disease class. A dataset available online was used to validate the proposed two-stage framework, resulting in a classification rate of 98.25% for apple leaf condition identification and an accuracy of 98.60% for apple leaf disease diagnosis. Furthermore, the results confirm that the proposed model is lightweight and involves relatively fewer learnable parameters in comparison with other pre-trained deep learning models.

PMID:39665107 | PMC:PMC11631600 | DOI:10.3389/fpls.2024.1502314

Alzheimers Dement (Amst). 2024 Dec 11;16(4):e70037. doi: 10.1002/dad2.70037. eCollection 2024 Oct-Dec.

ABSTRACT

INTRODUCTION: This study evaluates the clinical value of a deep learning-based artificial intelligence (AI) system that performs rapid brain volumetry with automatic lobe segmentation and age- and sex-adjusted percentile comparisons.

METHODS: Fifty-five patients-17 with Alzheimer's disease (AD), 18 with frontotemporal dementia (FTD), and 20 healthy controls-underwent cranial magnetic resonance imaging scans. Two board-certified neuroradiologists (BCNR), two board-certified radiologists (BCR), and three radiology residents (RR) assessed the scans twice: first without AI support and then with AI assistance.

RESULTS: AI significantly improved diagnostic accuracy for AD (area under the curve -AI: 0.800, +AI: 0.926, p < 0.05), with increased correct diagnoses (p < 0.01) and reduced errors (p < 0.03). BCR and RR showed notable performance gains (BCR: p < 0.04; RR: p < 0.02). For the diagnosis FTD, overall consensus (p < 0.01), BCNR (p < 0.02), and BCR (p < 0.05) recorded significantly more correct diagnoses.

DISCUSSION: AI-assisted volumetry improves diagnostic performance in differentiating AD and FTD, benefiting all reader groups, including BCNR.

HIGHLIGHTS: Artificial intelligence (AI)-supported brain volumetry significantly improved the diagnostic accuracy for Alzheimer's disease (AD) and frontotemporal dementia (FTD), with notable performance gains across radiologists of varying expertise levels.The presented AI tool is readily clinically available and reduces brain volumetry processing time from 12 to 24 hours to under 5 minutes, with full integration into picture archiving and communication systems, streamlining the workflow and facilitating real-time clinical decision making.AI-supported rapid brain volumetry has the potential to improve early diagnosis and to improve patient management.

PMID:39665087 | PMC:PMC11632536 | DOI:10.1002/dad2.70037