Sci Rep. 2025 Feb 25;15(1):6758. doi: 10.1038/s41598-025-90780-z.

ABSTRACT

Diabetic foot ulcers (DFUs) are a common and serious complication of diabetes, presenting as open sores or wounds on the sole. They result from impaired blood circulation and neuropathy associated with diabetes, increasing the risk of severe infections and even amputations if untreated. Early detection, effective wound care, and diabetes management are crucial to prevent and treat DFUs. Artificial intelligence (AI), particularly through deep learning, has revolutionized DFU diagnosis and treatment. This work introduces the DFU_XAI framework to enhance the interpretability of deep learning models for DFU labeling and localization, ensuring clinical relevance. The framework evaluates six advanced models-Xception, DenseNet121, ResNet50, InceptionV3, MobileNetV2, and Siamese Neural Network (SNN)-using interpretability techniques like SHAP, LIME, and Grad-CAM. Among these, the SNN model excelled with 98.76% accuracy, 99.3% precision, 97.7% recall, 98.5% F1-score, and 98.6% AUC. Grad-CAM heat maps effectively identified ulcer locations, aiding clinicians with precise and visually interpretable insights. The DFU_XAI framework integrates explainability into AI-driven healthcare, enhancing trust and usability in clinical settings. This approach addresses challenges of transparency in AI for DFU management, offering reliable and efficient solutions to this critical healthcare issue. Traditional DFU methods are labor-intensive and costly, highlighting the transformative potential of AI-driven systems.

PMID:40000748 | DOI:10.1038/s41598-025-90780-z

J Imaging Inform Med. 2025 Feb 25. doi: 10.1007/s10278-025-01436-3. Online ahead of print.

ABSTRACT

Ultrasound is widely used to monitor and diagnose thyroid nodules, but accurately segmenting these nodules in ultrasound images remains a challenge due to the presence of noise and artifacts, which often blur nodule boundaries. While several deep learning algorithms have been developed for this task, their performance is frequently suboptimal. In this study, we introduce the use of EfficientNet-B7 as the backbone for the DeepLabV3+ architecture in thyroid nodule segmentation, marking its first application in this area. We evaluated the proposed method using a dataset from the First Affiliated Hospital of Zhengzhou University, along with two public datasets. The results demonstrate high performance, with a pixel accuracy (PA) of 97.67%, a Dice similarity coefficient of 0.8839, and an Intersection over Union (IoU) of 79.69%. These outcomes outperform most traditional segmentation networks.

PMID:40000546 | DOI:10.1007/s10278-025-01436-3

Insights Imaging. 2025 Feb 25;16(1):48. doi: 10.1186/s13244-025-01923-9.

ABSTRACT

OBJECTIVES: To develop and validate a deep learning model based on nnU-Net combined with radiomics to achieve autosegmentation of gastric cancer (GC) and preoperative prediction via the Lauren classification.

METHODS: Patients with a pathological diagnosis of GC were retrospectively enrolled in three medical centers. The nnU-Net autosegmentation model was developed using manually segmented datasets and evaluated by the Dice similarity coefficient (DSC). The CT images were processed by the nnU-Net model to obtain autosegmentation results and extract radiomic features. The least absolute shrinkage and selection operator (LASSO) method selects optimal features for calculating the Radscore and constructing a radiomic model. Clinical characteristics and the Radscore were integrated to construct a combined model. Model performance was evaluated via the receiver operating characteristic (ROC) curve.

RESULTS: A total of 433 GC patients were divided into the training set, internal validation set, external test set-1, and external test set-2. The nnU-Net model achieved a DSC of 0.79 in the test set. The areas under the curve (AUCs) of the internal validation set, external test set-1, and external test set-2 were 0.84, 0.83, and 0.81, respectively, for the radiomic model; and 0.81, 0.81, and 0.82, respectively, for the combined model. The AUCs of the radiomic and combined models showed no statistically significant difference (p > 0.05). The radiomic model was selected as the optimal model.

CONCLUSIONS: The nnU-Net model can efficiently and accurately achieve automatic segmentation of GCs. The radiomic model can preoperatively predict the Lauren classification of GC with high accuracy.

CRITICAL RELEVANCE STATEMENT: This study highlights the potential of nnU-Net combined with radiomics to noninvasively predict the Lauren classification in gastric cancer patients, enhancing personalized treatment strategies and improving patient management.

KEY POINTS: The Lauren classification influences gastric cancer treatment and prognosis. The nnU-Net model reduces doctors' manual segmentation errors and workload. Radiomics models aid in preoperative Lauren classification prediction for patients with gastric cancer.

PMID:40000513 | DOI:10.1186/s13244-025-01923-9

Acad Radiol. 2025 Feb 24:S1076-6332(25)00104-7. doi: 10.1016/j.acra.2025.02.002. Online ahead of print.

ABSTRACT

RATIONALE AND OBJECTIVES: This study assesses the image quality of temporal bone ultra-high-resolution (UHR) Computed tomography (CT) scans in adults and children using hybrid iterative reconstruction (HIR) and a novel, vendor-specific deep learning-based reconstruction (DLR) algorithm called AiCE Inner Ear.

MATERIAL AND METHODS: In a retrospective, single-center study (February 1-July 30, 2023), UHR-CT scans of 57 temporal bones of 35 patients (5 children, 23 male) with at least one anatomical unremarkable temporal bone were included. There is an adult computed tomography dose index volume (CTDIvol 25.6 mGy) and a pediatric protocol (15.3 mGy). Images were reconstructed using HIR at normal resolution (0.5-mm slice thickness, 512² matrix) and UHR (0.25-mm, 1024² and 2048² matrix) as well as with a vendor-specific DLR advanced intelligent clear-IQ engine inner ear (AiCE Inner Ear) at UHR (0.25-mm, 1024² matrix). Three radiologists evaluated 18 anatomic structures using a 5-point Likert scale. Signal-to-noise (SNR) and contrast-to-noise ratio (CNR) were measured automatically.

RESULTS: In the adult protocol subgroup (n=30; median age: 51 [11-89]; 19 men) and the pediatric protocol subgroup (n=5; median age: 2 [1-3]; 4 men), UHR-CT with DLR significantly improved subjective image quality (p<0.024), reduced noise (p<0.001), and increased CNR and SNR (p<0.001). DLR also enhanced visualization of key structures, including the tendon of the stapedius muscle (p<0.001), tympanic membrane (p<0.009), and basal aspect of the osseous spiral lamina (p<0.018).

CONCLUSION: Vendor-specific DLR-enhanced UHR-CT significantly improves temporal bone image quality and diagnostic performance.

PMID:40000329 | DOI:10.1016/j.acra.2025.02.002

Acad Radiol. 2025 Feb 24:S1076-6332(25)00108-4. doi: 10.1016/j.acra.2025.02.007. Online ahead of print.

ABSTRACT

PURPOSE: This meta-analysis aims to evaluate the diagnostic performance of CT-based artificial intelligence (AI) in diagnosing cervical lymph node metastasis (LNM) of papillary thyroid cancer (PTC).

METHODS: A systematic search was conducted in PubMed, Embase, and Web of Science databases through December 2024, following PRISMA-DTA guidelines. Studies evaluating CT-based AI models for diagnosing cervical LNM in patients with pathologically confirmed PTC were included. The methodological quality was assessed using a modified QUADAS-2 tool. A bivariate random-effects model was used to calculate pooled sensitivity, specificity, and area under the curve (AUC). Heterogeneity was evaluated using I2 statistics, and meta-regression analyses were performed to explore potential sources of heterogeneity.

RESULTS: 17 studies comprising 1778 patients in internal validation sets and 4072 patients in external validation sets were included. In internal validation sets, AI demonstrated a sensitivity of 0.80 (95% CI: 0.71-0.86), specificity of 0.79 (95% CI: 0.73-0.84), and AUC of 0.86 (95% CI: 0.83-0.89). Radiologists suggested comparable performance with sensitivity of 0.77 (95% CI: 0.64-0.87), specificity of 0.79 (95% CI: 0.72-0.85), and AUC of 0.85 (95% CI: 0.81-0.88). Subgroup analyses revealed that deep learning methods outperformed machine learning in sensitivity (0.86 vs 0.72, P<0.05). No significant publication bias was found in internal validation sets for AI diagnosis (P=0.78).

CONCLUSION: CT-based AI showed comparable diagnostic performance to radiologists for detecting cervical LNM in PTC patients, with deep learning models showing superior sensitivity. AI could potentially serve as a valuable diagnostic support tool, though further prospective validation is warranted. Limitations include high heterogeneity among studies and insufficient external validation in diverse populations.

PMID:40000328 | DOI:10.1016/j.acra.2025.02.007

Sheng Wu Yi Xue Gong Cheng Xue Za Zhi. 2024 Dec 25;41(6):1293-1300. doi: 10.7507/1001-5515.202404004.

ABSTRACT

Gastric tumors are neoplastic lesions that occur in the stomach, posing a great threat to human health. Gastric cancer represents the malignant form of gastric tumors, and early detection and treatment are crucial for patient recovery. Endoscopic examination is the primary method for diagnosing gastric tumors. Deep learning techniques can automatically extract features from endoscopic images and analyze them, significantly improving the detection rate of gastric cancer and serving as an important tool for auxiliary diagnosis. This paper reviews relevant literature in recent years, presenting the application of deep learning methods in the classification, object detection, and segmentation of gastric tumor endoscopic images. In addition, this paper also summarizes several computer-aided diagnosis (CAD) systems and multimodal algorithms related to gastric tumors, highlights the issues with current deep learning methods, and provides an outlook on future research directions, aiming to promote the clinical application of deep learning methods in the endoscopic diagnosis of gastric tumors.

PMID:40000222 | DOI:10.7507/1001-5515.202404004

Breast Cancer Res. 2025 Feb 25;27(1):29. doi: 10.1186/s13058-025-01983-1.

ABSTRACT

BACKGROUND: Radiomics and AI have been widely used in breast cancer imaging, but a comprehensive systematic analysis is lacking. Therefore, this study aims to conduct a bibliometrics analysis in this field to discuss its research status and frontier hotspots and provide a reference for subsequent research.

METHODS: Publications related to AI, radiomics, and breast cancer imaging were searched in the Web of Science Core Collection. CiteSpace plotted the relevant co-occurrence network according to authors and keywords. VOSviewer and Pajek were used to draw relevant co-occurrence maps according to country and institution. In addition, R was used to conduct bibliometric analysis of relevant authors, countries/regions, journals, keywords, and annual publications and citations based on the collected information.

RESULTS: A total of 2,701 Web of Science Core Collection publications were retrieved, including 2,486 articles (92.04%) and 215 reviews (7.96%). The number of publications increased rapidly after 2018. The United States of America (n = 17,762) leads in citations, while China (n = 902) leads in the number of publications. Sun Yat-sen University (n = 75) had the largest number of publications. Bin Zheng (n = 28) was the most published author. Nico Karssemeijer (n = 72.1429) was the author with the highest average citations. "Frontiers in Oncology" was the journal with the most publications, and "Radiology" had the highest IF. The keywords with the most frequent occurrence were "breast cancer", "deep learning", and "classification". The topic trends in recent years were "explainable AI", "neoadjuvant chemotherapy", and "lymphovascular invasion".

CONCLUSION: The application of radiomics and AI in breast cancer imaging has received extensive attention. Future research hotspots may mainly focus on the progress of explainable AI in the technical field and the prediction of lymphovascular invasion and neoadjuvant chemotherapy efficacy in clinical application.

PMID:40001088 | DOI:10.1186/s13058-025-01983-1

Radiat Oncol. 2025 Feb 25;20(1):26. doi: 10.1186/s13014-025-02603-0.

ABSTRACT

BACKGROUND: Deformable registration plays an important role in the accurate delineation of tumors. Most of the existing deep learning methods ignored two issues that can lead to inaccurate registration, including the limited field of view in MR scans and the different scanning angles that can exist between multimodal images. The purpose of this study is to improve the registration accuracy between CT and MR for nasopharyngeal carcinoma cases.

METHODS: 269 cases were enrolled in the study, and 188 cases were designated for training, while a separate set of 81 cases was reserved for testing. Each case had a CT volume and a T1-MR volume. The treatment table was removed from their CT images. The CycleFCNs model was used for deformable registration, and two strategies including adaptive mask registration strategy and weight allocation strategy were adopted for training. Dice similarity coefficient, Hausdorff distance, precision, and recall were calculated for normal tissues of CT-MR image pairs, before and after the registration. Three deformable registration methods including RayStation, Elastix, and VoxelMorph were compared with the proposed method.

RESULTS: The registration results of RayStation and Elastix are essentially consistent. Upon employing the VoxelMorph model and the proposed method for registration, a clear trend of increased dice similarity coefficient and decreased hausdorff distance can be observed. It is noteworthy that for the temporomandibular joint, pituitary, optic nerve, and optic chiasma, the proposed method has improved the average dice similarity coefficient from 0.86 to 0.91, 0.87 to 0.93, 0.85 to 0.89, and 0.77 to 0.83, respectively, as compared to RayStation. Additionally, within the same anatomical structures, the average hausdorff distance has been decreased from 2.98 mm to 2.28 mm, 1.83 mm to 1.53 mm, 3.74 mm to 3.56 mm, and 5.94 mm to 5.87 mm. Compared to the original CycleFCNs model, the improved model has significantly enhanced the dice similarity coefficient of the brainstem, pituitary gland, and optic nerve (P < 0.001).

CONCLUSIONS: The proposed method significantly improved the registration accuracy for multi-modal medical images in NPC cases. These findings have important clinical implications, as increased registration accuracy can lead to more precise tumor segmentation, optimized treatment planning, and ultimately, improved patient outcomes.

PMID:40001040 | DOI:10.1186/s13014-025-02603-0

BMC Cancer. 2025 Feb 25;25(1):341. doi: 10.1186/s12885-025-13711-1.

ABSTRACT

OBJECTIVE: We aimed to develop a preoperative clinical radiomics survival prediction model based on the radiomics features via deep learning to provide a reference basis for preoperative assessment and treatment decisions for patients with gallbladder carcinoma (GBC).

METHODS: A total of 168 GBC patients who underwent preoperative upper abdominal enhanced CT from one high-volume medical center between January 2011 to December 2020 were retrospectively analyzed. The region of interest (ROI) was manually outlined by two physicians using 3D Slicer software to establish a nnU-Net model. The DeepSurv survival prediction model was developed by combining radiomics features and preoperative clinical variables.

RESULTS: A total of 1502 radiomics features were extracted from the ROI results based on the nnU-Net model and manual segmentation, and 13 radiomics features were obtained through the 4-step dimensionality reduction methods, respectively. The C-index and AUC of 1-, 2-, and 3-year survival prediction for the nnU-Net based clinical radiomics DeepSurv model was higher than clinical and nnU-Net based radiomics DeepSurv models in the training and testing sets, and close to manual based clinical radiomics DeepSurv model. Delong-test was performed on the AUC of 1-, 2-, and 3-year survival prediction for the two preoperative clinical radiomics DeepSurv prediction models in the testing set, and the results showed that the two models had the same prediction efficiency (all P > 0.05).

CONCLUSIONS: By using the DeepSurv model via nnU-Net segmentation, postoperative survival outcomes for individual gallbladder carcinoma patients could be assessed and stratified, which can provide references for preoperative diagnosis and treatment decisions.

PMID:40001024 | DOI:10.1186/s12885-025-13711-1

BMC Med Imaging. 2025 Feb 25;25(1):61. doi: 10.1186/s12880-025-01602-7.

ABSTRACT

BACKGROUND: To compare the influence of rectal susceptibility artifacts on the subjective evaluation and deep learning (DL) in prostate cancer (PCa) diagnosis.

METHODS: This retrospective two-center study included 1052 patients who underwent MRI and biopsy due to clinically suspected PCa between November 2019 and November 2023. The extent of rectal artifacts in these patients' images was evaluated using the Likert four-level method. The PCa diagnosis was performed by six radiologists and an automated PCa diagnosis DL method. The performance of DL and radiologists was evaluated using the area under the receiver operating characteristic curve (AUC) and the area under the multi-reader multi-case receiver operating characteristic curve, respectively.

RESULTS: Junior radiologists and DL demonstrated statistically significantly higher AUCs in patients without artifacts compared to those with artifacts (R1: 0.73 vs. 0.64; P = 0.01; R2: 0.74 vs. 0.67; P = 0.03; DL: 0.77 vs. 0.61; P < 0.001). In subgroup analysis, no statistically significant differences in the AUC were observed among different grades of rectal artifacts for both all radiologists (0.08 ≤ P ≤ 0.90) and DL models (0.12 ≤ P ≤ 0.96). The AUC for DL without artifacts significantly exceeded those with artifacts in both the peripheral zone (PZ) and transitional zone (TZ) (DLPZ: 0.78 vs. 0.61; P = 0.003; DLTZ: 0.73 vs. 0.59; P = 0.011). Conversely, there were no statistically significant differences in AUC with and without artifacts for all radiologists in PZ and TZ (0.08 ≤ P ≤ 0.98).

CONCLUSIONS: Rectal susceptibility artifacts have significant negative effects on subjective evaluation of junior radiologists and DL.

CLINICAL TRIAL NUMBER: Not applicable.

PMID:40000986 | DOI:10.1186/s12880-025-01602-7

Sci Rep. 2025 Feb 25;15(1):6820. doi: 10.1038/s41598-025-91553-4.

ABSTRACT

Monitoring the daily activity levels of black cattle is a crucial aspect of their well-being. The rapid advancements in artificial intelligence have transformed computer vision applications, including object detection, segmentation, and tracking. This has led to more effective and precise monitoring techniques for livestock. In modern cattle farms, video monitoring is essential for analyzing behavior, evaluating health, and predicting estrus events in precision farming. This paper introduces the novel Customized Multi-Camera Multi-Cattle Tracking (MCMCT) system. This unique approach uses four cameras to overcome the challenges of detecting and tracking black cattle in complex open ranch environments. The MCMCT system enhances a tracking-by-detection model with the YOLO v8 segmentation model as the detection backbone network to develop a precision black cattle monitoring system. Single-camera setups in real-world datasets of our open ranches, covering 23.3 m x 20 m with 55 cattle, have limitations in capturing all necessary details. Therefore, a multi-camera solution provides better coverage and more accurate behavior detection of cattle. The effectiveness of the MCMCT system is demonstrated through experimental results, with the YOLOv8-MCMCT system achieving an average Multi-Object Tracking Accuracy (MOTA) of 95.61% across 10 cases of 4 cameras at a processing speed of 30 frames per second. This high accuracy is a testament to the performance of the proposed MCMCT system. Additionally, integrating the Segment Anything Model (SAM) with YOLOv8 enhances the system's capability by automating cattle mask region extraction, reducing the need for manual labeling. Comparative analysis with state-of-the-art deep learning-based tracking methods, including Bot-sort, Byte-track, and OC-sort, further highlights the MCMCT's performance in multi-cattle tracking within complex natural scenes. The advanced algorithms and capabilities of the MCMCT system make it a valuable tool for non-contact automatic livestock monitoring in precision cattle farming. Its adaptability ensures effective performance across varied ranch environments without extensive retraining. This research significantly contributes to livestock monitoring, offering a robust solution for tracking black cattle and enhancing overall agricultural efficiency and management.

PMID:40000894 | DOI:10.1038/s41598-025-91553-4

Commun Med (Lond). 2025 Feb 25;5(1):50. doi: 10.1038/s43856-025-00770-5.

ABSTRACT

BACKGROUND: Huntington's disease, a neurodegenerative disorder, impairs both upper and lower limb function, typically assessed in clinical settings. However, wearable sensors offer the opportunity to monitor real-world data that complements clinical assessments, providing a more comprehensive understanding of disease symptoms.

METHODS: In this study, we monitor upper limb function in individuals with Huntington's disease (HD, n = 16), prodromal HD (pHD, n = 7), and controls (CTR, n = 16) using a wrist-worn wearable sensor over a 7-day period. Goal-directed hand movements are detected through a deep learning model, and kinematic features of each movement are analyzed. The collected data is used to predict disease groups and clinical scores using statistical and machine learning models.

RESULTS: Here we show that significant differences in goal-directed movement features exist between the groups. Additionally, several of these features strongly correlate with clinical scores. Classification models accurately distinguish between HD, pHD, and CTR individuals, achieving a balanced accuracy of 67% and a recall of 0.72 for the HD group. Regression models effectively predict clinical scores.

CONCLUSIONS: This study demonstrates the potential of wearable sensors and machine learning to monitor upper limb function in Huntington's disease, offering a tool for early detection, remote monitoring, and assessing treatment efficacy in clinical trials.

PMID:40000872 | DOI:10.1038/s43856-025-00770-5

Sci Rep. 2025 Feb 25;15(1):6702. doi: 10.1038/s41598-025-91203-9.

ABSTRACT

Generally, music genres have not new established framework, since they are often determined by the composer's background by cultural or historical impact and geographical origin. In this work, a new methodology is presented based on deep learning and metaheuristic algorithms to enhance the performance in music style categorization. The model consists of two main parts: a pre-trained model, a ZFNet, through which high level features are extracted from audio signals and a ResNeXt model for classification. A fractional-order-based variant of the Grey Lag Goose Optimization (FGLGO) algorithm is used to optimize the parameters of ResNeXt to boost the performance of the model. A dual-path recurrent network is employed for real-time music generation and evaluate the model on two benchmark datasets, ISMIR2004 and extended Ballroom, compared to the state-of-the-art models included CNN, PRCNN, BiLSTM and BiRNN. Experimental results show that with accuracy rates of 0.918 on the extended Ballroom dataset and 0.954 on the ISMIR2004 dataset, the proposed model improves accuracy and efficiency incrementally over existing models.

PMID:40000796 | DOI:10.1038/s41598-025-91203-9

Sci Rep. 2025 Feb 25;15(1):6798. doi: 10.1038/s41598-025-91329-w.

ABSTRACT

Ozone pollution affects food production, human health, and the lives of individuals. Due to rapid industrialization and urbanization, Liaocheng has experienced increasing of ozone concentration over several years. Therefore, ozone has become a major environmental problem in Liaocheng City. Long short-term memory (LSTM) and artificial neural network (ANN) models are established to predict ozone concentrations in Liaocheng City from 2014 to 2023. The results show a general improvement in the accuracy of the LSTM model compared to the ANN model. Compared to the ANN, the LSTM has an increase in determination coefficient (R2), value from 0.6779 to 0.6939, a decrease in root mean square error (RMSE) value from 27.9895 μg/m3 to 27.2140 μg/m3 and a decrease in mean absolute error (MAE) value from 21.6919 μg/m3 to 20.8825 μg/m3. The prediction accuracy of the LSTM is superior to the ANN in terms of R, RMSE, and MAE. In summary, LSTM is a promising technique for predicting ozone concentrations. Moreover, by leveraging historical data and LSTM enables accurate predictions of future ozone concentrations on a global scale. This model will open up new avenues for controlling and mitigating ozone pollution.

PMID:40000767 | DOI:10.1038/s41598-025-91329-w

Sci Rep. 2025 Feb 25;15(1):6800. doi: 10.1038/s41598-025-91362-9.

ABSTRACT

This study was conducted to develop and validate a novel deep reinforcement learning (DRL) algorithm incorporating the segment anything model (SAM) to enhance the accuracy of automatic contouring organs at risk during radiotherapy for cervical cancer patients. CT images were collected from 150 cervical cancer patients treated at our hospital between 2021 and 2023. Among these images, 122 CT images were used as a training set for the algorithm training of the DRL model based on the SAM model, and 28 CT images were used for the test set. The model's performance was evaluated by comparing its segmentation results with the ground truth (manual contouring) obtained through manual contouring by expert clinicians. The test results were compared with the contouring results of commercial automatic contouring software based on the deep learning (DL) algorithm model. The Dice similarity coefficient (DSC), 95th percentile Hausdorff distance, average symmetric surface distance (ASSD), and relative absolute volume difference (RAVD) were used to quantitatively assess the contouring accuracy from different perspectives, enabling the contouring results to be comprehensively and objectively evaluated. The DRL model outperformed the DL model across all evaluated metrics. DRL achieved higher median DSC values, such as 0.97 versus 0.96 for the left kidney (P < 0.001), and demonstrated better boundary accuracy with lower HD95 values, e.g., 14.30 mm versus 17.24 mm for the rectum (P < 0.001). Moreover, DRL exhibited superior spatial agreement (median ASSD: 1.55 mm vs. 1.80 mm for the rectum, P < 0.001) and volume prediction accuracy (median RAVD: 10.25 vs. 10.64 for the duodenum, P < 0.001). These findings indicate that integrating SAM with RL (reinforcement learning) enhances segmentation accuracy and consistency compared to conventional DL methods. The proposed approach introduces a novel training strategy that improves performance without increasing model complexity, demonstrating its potential applicability in clinical practice.

PMID:40000766 | DOI:10.1038/s41598-025-91362-9

Sci Rep. 2025 Feb 25;15(1):6739. doi: 10.1038/s41598-025-91275-7.

ABSTRACT

The rapid growth of e-commerce has made product recommendation systems essential for enhancing customer experience and driving business success. This research proposes an advanced recommendation framework that integrates sentiment analysis (SA) and collaborative filtering (CF) to improve recommendation accuracy and user satisfaction. The methodology involves feature-level sentiment analysis with a multi-step pipeline: data preprocessing, feature extraction using a log-term frequency-based modified inverse class frequency (LFMI) algorithm, and sentiment classification using a Multi-Layer Attention-based Encoder-Decoder Temporal Convolution Neural Network (MLA-EDTCNet). To address class imbalance issues, a Modified Conditional Generative Adversarial Network (MCGAN) generates balanced oversamples. Furthermore, the Ocotillo Optimization Algorithm (OcOA) fine-tunes the model parameters to ensure optimal performance by balancing exploration and exploitation during training. The integrated system predicts sentiment polarity-positive, negative, or neutral-and combines these insights with CF to provide personalized product recommendations. Extensive experiments conducted on an Amazon product dataset demonstrate that the proposed approach outperforms state-of-the-art models in accuracy, precision, recall, F1-score, and AUC. By leveraging SA and CF, the framework delivers recommendations tailored to user preferences while enhancing engagement and satisfaction. This research highlights the potential of hybrid deep learning techniques to address critical challenges in recommendation systems, including class imbalance and feature extraction, offering a robust solution for modern e-commerce platforms.

PMID:40000752 | DOI:10.1038/s41598-025-91275-7

Phys Med Biol. 2025 Feb 25. doi: 10.1088/1361-6560/adba39. Online ahead of print.

ABSTRACT

Objective:To assess the performance of a probabilistic deep learning based algorithm for predicting inter-fraction anatomical changes in head and neck patients.&#xD;&#xD;Approach:A probabilistic daily anatomy model for head and neck patients (DAMHN) is built on the variational autoencoder architecture. The model approximates the generative joint conditional probability distribution of the repeat computed tomography (rCT) images and their corresponding masks on the planning CT images (pCT) and their masks. The model outputs deformation vector fields, which are used to produce possible rCTs and associated masks. The dataset is composed of 93 patients (i.e., 315 pCT - rCT pairs), 9 (i.e., 27 pairs) of which were set aside for final testing. The performance of the model is assessed based on the reconstruction accuracy and the generative performance for the set aside patients. &#xD;&#xD;Main results:The model achieves a DICE score of 0.83 and an image similarity score (NCC) of 0.60 on the test set. The generated parotid glands, spinal cord and constrictor muscle volume change distributions and center of mass shift distributions were also assessed. For all organs, the medians of the distributions are close to the true ones, and the distributions are broad enough to encompass the real observed changes. Moreover, the generated images display anatomical changes in line with the literature reported ones, such as the medial shifts of the parotids glands. &#xD;&#xD;Significance:DAMHNis capable of generating realistic anatomies observed during the course of the treatment and has applications in anatomical robust optimization, treatment planning based on plan library approaches and robustness evaluation against inter-fractional changes.

PMID:39999567 | DOI:10.1088/1361-6560/adba39

Comput Biol Med. 2025 Feb 24;188:109857. doi: 10.1016/j.compbiomed.2025.109857. Online ahead of print.

ABSTRACT

Explainability in Medical Computer Vision is one of the most sensible implementations of Artificial Intelligence nowadays in healthcare. In this work, we propose a novel Deep Learning architecture for eXplainable Artificial Intelligence, specially designed for medical diagnostic. The proposed approach leverages Variational Autoencoders properties to produce linear modifications of images in a lower-dimensional embedded space, and then reconstructs these modifications into non-linear explanations in the original image space. The proposed approach is based on global and local regularisation of the latent space, which stores visual and semantic information about images. Specifically, a multi-objective genetic algorithm is designed for searching explanations, finding individuals that can misclassify the classification output of the network while producing the minimum number of changes in the image descriptor. The genetic algorithm is able to search for explanations without defining any hyperparameters, and uses only one individual to provide a complete explanation of the whole image. Furthermore, the explanations found by the proposed approach are compared with state-of-the-art eXplainable Artificial Intelligence systems and the results show an improvement in the precision of the explanation between 56.39 and 7.23 percentage points.

PMID:39999495 | DOI:10.1016/j.compbiomed.2025.109857

Comput Biol Med. 2025 Feb 24;188:109835. doi: 10.1016/j.compbiomed.2025.109835. Online ahead of print.

ABSTRACT

The integration of Artificial Intelligence (AI) with the Internet of Medical Things (IoMT) has revolutionized disease prediction and detection, but challenges such as data heterogeneity, privacy concerns, and model generalizability hinder its full potential in healthcare. This review examines these challenges and evaluates the effectiveness of AI-IoMT techniques in predicting chronic and terminal diseases, including cardiovascular conditions, Alzheimer's disease, and cancers. We analyze a range of Machine Learning (ML) and Deep Learning (DL) approaches (e.g., XGBoost, Random Forest, CNN, LSTM), alongside advanced strategies like federated learning, transfer learning, and blockchain, to improve model robustness, data security, and interoperability. Findings highlight that transfer learning and ensemble methods enhance model adaptability across clinical settings, while blockchain and federated learning effectively address privacy and data standardization. Ultimately, the review emphasizes the importance of data harmonization, secure frameworks, and multi-disease models as critical research directions for scalable, comprehensive AI-IoMT solutions in healthcare.

PMID:39999492 | DOI:10.1016/j.compbiomed.2025.109835

Hepatology. 2025 Feb 25. doi: 10.1097/HEP.0000000000001283. Online ahead of print.

ABSTRACT

BACKGROUND AIMS: The prognoses and therapeutic responses of patients with intrahepatic cholangiocarcinoma (iCCA) depend on spatial interactions among tumor microenvironment (TME) components. However, the spatial TME characteristics of iCCA remain poorly understood. The aim of this study was to generate a comprehensive spatial atlas of iCCA using artificial intelligence-assisted spatial multiomics patterns and to identify spatial features associated with prognosis and immunotherapy.

APPROACH RESULTS: Spatial multiomics, including imaging mass cytometry (IMC, n=155 in-house), spatial proteomics (n=155 in-house), spatial transcriptomics (n=4 in-house), multiplex immunofluorescence (mIF, n=20 in-house), single-cell RNA sequencing (scRNA-seq, n=9 in-house and n=34 public), bulk RNA-seq (n=244 public), and bulk proteomics (n=110 in-house and n=214 public), were employed to elucidate the spatial TME of iCCA. More than 1.06 million cells were resolved, and the findings revealed that spatial topology, including cellular deposition patterns, cellular communities, and intercellular communications, profoundly correlates with the prognosis of iCCA patients. Specifically, CD163hi M2-like resident-tissue macrophages suppress anti-tumor immunity by directly interacting with CD8+ T cells, resulting in poorer patient survival. Additionally, five spatial subtypes with distinct prognoses were identified, and potential therapeutic options were generated for these subtypes. Furthermore, a spatial TME deep learning system was developed to predict the prognosis of iCCA patients with high accuracy from a single 1-mm2 tumor sample.

CONCLUSIONS: This study offers preliminary insights into the spatial TME ecosystem of iCCA, providing valuable foundations for precise patient classification and the development of personalized treatment strategies.

PMID:39999448 | DOI:10.1097/HEP.0000000000001283