JACC Adv. 2024 Mar 20;3(9):100889. doi: 10.1016/j.jacadv.2024.100889. eCollection 2024 Sep.

NO ABSTRACT

PMID:39372463 | PMC:PMC11451070 | DOI:10.1016/j.jacadv.2024.100889

JACC Adv. 2024 Jun 12;3(9):100998. doi: 10.1016/j.jacadv.2024.100998. eCollection 2024 Sep.

ABSTRACT

BACKGROUND: Recent studies suggest that cardiac amyloidosis (CA) is significantly underdiagnosed. For rare diseases like CA, the optimal selection of cases and controls for artificial intelligence model training is unknown and can significantly impact model performance.

OBJECTIVES: This study evaluates the performance of electrocardiogram (ECG) waveform-based artificial intelligence models for CA screening and assesses impact of different criteria for defining cases and controls.

METHODS: Using a primary cohort of ∼1.3 million ECGs from 341,989 patients, models were trained using different case and control definitions. Case definitions included ECGs from patients with an amyloidosis diagnosis by International Classification of Diseases-9/10 code, patients with CA, and patients seen in CA clinic. Models were then tested on test cohorts with identical selection criteria as well as a Cedars-Sinai general patient population cohort.

RESULTS: In matched held-out test data sets, different model AUCs ranged from 0.660 (95% CI: 0.642-0.736) to 0.898 (95% CI: 0.868-0.924). However, algorithms exhibited variable generalizability when tested on a Cedars-Sinai general patient population cohort, with AUCs dropping to 0.467 (95% CI: 0.443-0.491) to 0.898 (95% CI: 0.870-0.923). Models trained on more well-curated patient cases resulted in higher AUCs on similarly constructed test cohorts. However, all models performed similarly in the overall Cedars-Sinai general patient population cohort. A model trained with International Classification of Diseases 9/10 cases and population controls matched for age and sex resulted in the best screening performance.

CONCLUSIONS: Models performed similarly in population screening, regardless of stringency of cases used during training, showing that institutions without dedicated amyloid clinics can train meaningful models on less curated CA cases. Additionally, AUC or other metrics alone are insufficient in evaluating deep learning algorithm performance. Instead, evaluation in the most clinically meaningful population is key.

PMID:39372462 | PMC:PMC11450940 | DOI:10.1016/j.jacadv.2024.100998

Int Dent J. 2024 Oct 5:S0020-6539(24)01492-8. doi: 10.1016/j.identj.2024.09.002. Online ahead of print.

ABSTRACT

PURPOSE: This study investigated the potential application of large language models (LLMs) in dental education and practice, with a focus on ChatGPT and Claude3-Opus. Using the Korean Dental Licensing Examination (KDLE) as a benchmark, we aimed to assess the capabilities of these models in the dental field.

METHODS: This study evaluated three LLMs: GPT-3.5, GPT-4 (version: March 2024), and Claude3-Opus (version: March 2024). We used the KDLE questionnaire from 2019 to 2023 as inputs to the LLMs and then used the outputs from the LLMs as the corresponding answers. The total scores for individual subjects were obtained and compared. We also compared the performance of LLMs with those of individuals who underwent the exams.

RESULTS: Claude3-Opus performed best among the considered LLMs, except in 2019 when ChatGPT-4 performed best. Claude3-Opus and ChatGPT-4 surpassed the cut-off scores in all the years considered; this indicated that Claude3-Opus and ChatGPT-4 passed the KDLE, whereas ChatGPT-3.5 did not. However, all LLMs considered performed worse than humans, represented here by dental students in Korea. On average, the best-performing LLM annually achieved 85.4% of human performance.

CONCLUSION: Using the KDLE as a benchmark, our study demonstrates that although LLMs have not yet reached human-level performance in overall scores, both Claude3-Opus and ChatGPT-4 exceed the cut-off scores and perform exceptionally well in specific subjects.

CLINICAL RELEVANCE: Our findings will aid in evaluating the feasibility of integrating LLMs into dentistry to improve the quality and availability of dental services by offering patient information that meets the basic competency standards of a dentist.

PMID:39370338 | DOI:10.1016/j.identj.2024.09.002

J Appl Clin Med Phys. 2024 Oct 6:e14547. doi: 10.1002/acm2.14547. Online ahead of print.

ABSTRACT

PURPOSE: In the current clinical diagnostic process, the gold standard for lymph node metastasis (LNM) diagnosis is histopathological examination following surgical lymphadenectomy. Developing a non-invasive and preoperative method for predicting LNM is necessary and holds significant clinical importance.

METHODS: We develop a ranking attention multiple instance learning (RA-MIL) model that integrates convolutional neural networks (CNNs) and ranking attention pooling to diagnose LNM from T2 MRI. Our RA-MIL model applies the CNNs to derive imaging features from 2D MRI slices and employs ranking attention pooling to create patient-level feature representation for diagnostic classification. Based on the MIL and attention theory, informative regions of top-ranking MRI slices from LNM-positive patients are visualized to enhance the interpretability of automatic LNM prediction. This retrospective study collected 300 female patients with cervical cancer who underwent T2-weighted magnetic resonance imaging (MRI) scanning and histopathological diagnosis from one hospital (289 patients) and one open-source dataset (11 patients).

RESULTS: Our RA-MIL model delivers promising LNM prediction performance, achieving the area under the receiver operating characteristic curve (AUC) of 0.809 on the internal test set and 0.833 on the public dataset. Experiments show significant improvements in LNM status prediction using the proposed RA-MIL model compared with other state-of-the-art (SOTA) comparative deep learning models.

CONCLUSIONS: The developed RA-MIL model has the potential to serve as a non-invasive auxiliary tool for preoperative LNM prediction, offering visual interpretability regarding informative MRI slices and regions in LNM-positive patients.

PMID:39369718 | DOI:10.1002/acm2.14547

Med Phys. 2024 Oct 6. doi: 10.1002/mp.17451. Online ahead of print.

ABSTRACT

BACKGROUND: The I-SPY 2 trial is a national-wide, multi-institutional clinical trial designed to evaluate multiple new therapeutic drugs for high-risk breast cancer. Previous studies suggest that pathological complete response (pCR) is a viable indicator of long-term outcomes of neoadjuvant chemotherapy for high-risk breast cancer. While pCR can be assessed during surgery after the chemotherapy, early prediction of pCR before the completion of the chemotherapy may facilitate personalized treatment management to achieve an improved outcome. Notably, the acquisition of dynamic contrast-enhanced magnetic resonance (DCEMR) images at multiple time points during the I-SPY 2 trial opens up the possibility of achieving early pCR prediction.

PURPOSE: In this study, we investigated the feasibility of the early prediction of pCR to neoadjuvant chemotherapy using multi-time point DCEMR images and clinical data acquired in the I-SPY2 trial. The prediction uncertainty was also quantified to allow physicians to make patient-specific decisions on treatment plans based on the level of associated uncertainty.

METHODS: The dataset used in our study included 624 patients with DCEMR images acquired at 3 time points before the completion of the chemotherapy: pretreatment (T0), after 3 cycles of treatment (T1), and after 12 cycles of treatment (T2). A convolutional long short-term memory (LSTM) network-based deep learning model, which integrated multi-time point deep image representations with clinical data, including tumor subtypes, was developed to predict pCR. The performance of the model was evaluated via the method of nested 5-fold cross validation. Moreover, we also quantified prediction uncertainty for each patient through test-time augmentation. To investigate the relationship between predictive performance and uncertainty, the area under the receiver operating characteristic curve (AUROC) was assessed on subgroups of patients stratified by the uncertainty score.

RESULTS: By integrating clinical data and DCEMR images obtained at three-time points before treatment completion, the AUROC reached 0.833 with a sensitivity of 0.723 and specificity of 0.800. This performance was significantly superior (p < 0.01) to models using only images (AUROC = 0.706) or only clinical data (AUROC = 0.746). After stratifying the patients into eight subgroups based on the uncertainty score, we found that group #1, with the lowest uncertainty, had a superior AUROC of 0.873. The AUROC decreased to 0.637 for group #8, which had the highest uncertainty.

CONCLUSIONS: The results indicate that our convolutional LSTM network-based deep learning model can be used to predict pCR earlier before the completion of chemotherapy. By combining clinical data and multi-time point deep image representations, our model outperforms models built solely on clinical or image data. Estimating prediction uncertainty may enable physicians to prioritize or disregard predictions based on their associated uncertainties. This approach could potentially enhance the personalization of breast cancer therapy.

PMID:39369684 | DOI:10.1002/mp.17451

Artif Intell Med. 2024 Sep 30;157:102993. doi: 10.1016/j.artmed.2024.102993. Online ahead of print.

ABSTRACT

The integration of different imaging modalities, such as structural, diffusion tensor, and functional magnetic resonance imaging, with deep learning models has yielded promising outcomes in discerning phenotypic characteristics and enhancing disease diagnosis. The development of such a technique hinges on the efficient fusion of heterogeneous multimodal features, which initially reside within distinct representation spaces. Naively fusing the multimodal features does not adequately capture the complementary information and could even produce redundancy. In this work, we present a novel joint self-supervised and supervised contrastive learning method to learn the robust latent feature representation from multimodal MRI data, allowing the projection of heterogeneous features into a shared common space, and thereby amalgamating both complementary and analogous information across various modalities and among similar subjects. We performed a comparative analysis between our proposed method and alternative deep multimodal learning approaches. Through extensive experiments on two independent datasets, the results demonstrated that our method is significantly superior to several other deep multimodal learning methods in predicting abnormal neurodevelopment. Our method has the capability to facilitate computer-aided diagnosis within clinical practice, harnessing the power of multimodal data. The source code of the proposed model is publicly accessible on GitHub: https://github.com/leonzyzy/Contrastive-Network.

PMID:39369634 | DOI:10.1016/j.artmed.2024.102993

Comput Methods Programs Biomed. 2024 Oct 3;257:108454. doi: 10.1016/j.cmpb.2024.108454. Online ahead of print.

ABSTRACT

BACKGROUND AND OBJECTIVE: Integrating domain knowledge into deep learning models can improve their effectiveness and increase explainability. This study aims to enhance the classification performance of electrocardiograms (ECGs) by customizing specific guided mechanisms based on the characteristics of different cardiac abnormalities.

METHODS: Two novel guided attention mechanisms, Guided Spatial Attention (GSA) and CAM-based spatial guided attention mechanism (CGAM), were introduced. Different attention guidance labels were created based on clinical knowledge for four ECG abnormality classification tasks: ST change detection, premature contraction identification, Wolf-Parkinson-White syndrome (WPW) classification, and atrial fibrillation (AF) detection. The models were trained and evaluated separately for each classification task. Model explainability was quantified using Shapley values.

RESULTS: GSA improved the F1 score of the model by 5.74%, 5%, 8.96%, and 3.91% for ST change detection, premature contraction identification, WPW classification, and AF detection, respectively. Similarly, CGAM exhibited improvements of 3.89%, 5.40%, 8.21%, and 1.80% for the respective tasks. The combined use of GSA and CGAM resulted in even higher improvements of 6.26%, 5.58%, 8.85%, and 4.03%, respectively. Moreover, when all four tasks were conducted simultaneously, a notable overall performance boost was achieved, demonstrating the broad adaptability of the proposed model. The quantified Shapley values demonstrated the effectiveness of the guided attention mechanisms in enhancing the model's explainability.

CONCLUSIONS: The guided attention mechanisms, utilizing domain knowledge, effectively directed the model's attention, leading to improved classification performance and explainability. These findings have significant implications in facilitating accurate automated ECG classification.

PMID:39369585 | DOI:10.1016/j.cmpb.2024.108454

Seizure. 2024 Sep 25;122:80-86. doi: 10.1016/j.seizure.2024.09.019. Online ahead of print.

ABSTRACT

INTRODUCTION: Interictal epileptiform discharges (IEDs) in electroencephalograms (EEGs) are an important biomarker for epilepsy. Currently, the gold standard for IED detection is the visual analysis performed by experts. However, this process is expert-biased, and time-consuming. Developing fast, accurate, and robust detection methods for IEDs based on EEG may facilitate epilepsy diagnosis. We aim to assess the performance of deep learning (DL) and classic machine learning (ML) algorithms in classifying EEG segments into IED and non-IED categories, as well as distinguishing whether the entire EEG contains IED or not.

METHODS: We systematically searched PubMed, Embase, and Web of Science following PRISMA guidelines. We excluded studies that only performed the detection of IEDs instead of binary segment classification. Risk of Bias was evaluated with Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2). Meta-analysis with the overall area under the Summary Receiver Operating Characteristic (SROC), sensitivity, and specificity as effect measures, was performed with R software.

RESULTS: A total of 23 studies, comprising 3,629 patients, were eligible for synthesis. Eighteen models performed discharge-level classification, and 6 whole-EEG classification. For the IED-level classification, 3 models were validated in an external dataset with more than 50 patients and achieved a sensitivity of 84.9 % (95 % CI: 82.3-87.2) and a specificity of 68.7 % (95 % CI: 7.9-98.2). Five studies reported model performance using both internal validation (cross-validation) and external datasets. The meta-analysis revealed higher performance for internal validation, with 90.4 % sensitivity and 99.6 % specificity, compared to external validation, which showed 78.1 % sensitivity and 80.1 % specificity.

CONCLUSION: Meta-analysis showed higher performance for models validated with resampling methods compared to those using external datasets. Only a minority of models use more robust validation techniques, which often leads to overfitting.

PMID:39369555 | DOI:10.1016/j.seizure.2024.09.019

Comput Biol Med. 2024 Oct 5;183:109218. doi: 10.1016/j.compbiomed.2024.109218. Online ahead of print.

ABSTRACT

Traditionally, the classification of HIV-1 M group subtypes has depended on statistical methods constrained by sample sizes. Here HIV-1-M-SPBEnv was proposed as the first deep learning-based method for classifying HIV-1 M group subtypes via env gene sequences. This approach overcomes sample size challenges by utilizing artificial molecular evolution techniques to generate a synthetic dataset suitable for machine learning. Employing a convolutional Autoencoder embedded with two residual blocks and two transpose residual blocks, followed by a full connected neural network block, HIV-1-M-SPBEnv simplifies complex, high-dimensional DNA sequence data into concise, information-rich, low-dimensional representations, achieving exceptional classification accuracy. Through independent data set validation, the precision, accuracy, recall and F1 score of the HIV-1-M-SPBEnv model predictions were all 100 %, confirming its capability to accurately identify all 12 subtypes of the HIV-1 M group. Deployed through a web server, it provides seamless HIV-1 M group subtype prediction capabilities for researchers and clinicians. HIV-1-M-SPBEnv web server is accessible at http://www.hivsubclass.com and all the code is available at https://github.com/pengsihua2023/HIV-1-M-SPBEnv.

PMID:39369547 | DOI:10.1016/j.compbiomed.2024.109218

Eur J Med Chem. 2024 Sep 29;280:116912. doi: 10.1016/j.ejmech.2024.116912. Online ahead of print.

ABSTRACT

Deep learning has gained increasing attention in recent years, yielding promising results in hit screening and molecular optimization. Herein, we employed an efficient strategy based on multiple deep learning techniques to optimize Wee1 inhibitors, which involves activity interpretation, scaffold-based molecular generation, and activity prediction. Starting from our in-house Wee1 inhibitor GLX0198 (IC50 = 157.9 nM), we obtained three optimized compounds (IC50 = 13.5 nM, 33.7 nM, and 47.1 nM) out of five picked molecules. Further minor modifications on these compounds led to the identification of potent Wee1 inhibitors with desirable inhibitory effects on multiple cancer cell lines. Notably, the best compound 13 exhibited superior cancer cell inhibition, with IC50 values below 100 nM in all tested cancer cells. These results suggest that deep learning can greatly facilitate decision-making at the stage of molecular optimization.

PMID:39369485 | DOI:10.1016/j.ejmech.2024.116912

Commun Biol. 2024 Oct 5;7(1):1271. doi: 10.1038/s42003-024-06964-2.

ABSTRACT

Image-based spatial transcriptomic sequencing technologies have enabled the measurement of gene expression at single-cell resolution, but with a limited number of genes. Current computational approaches attempt to overcome these limitations by imputing missing genes, but face challenges regarding prediction accuracy and identification of cell populations due to the neglect of gene-gene relationships. In this context, we present stImpute, a method to impute spatial transcriptomics according to reference scRNA-seq data based on the gene network constructed from the protein language model ESM-2. Specifically, stImpute employs an autoencoder to create gene expression embeddings for both spatial transcriptomics and scRNA-seq data, which are used to identify the nearest neighboring cells between scRNA-seq and spatial transcriptomics datasets. According to the neighbored cells, the gene expressions of spatial transcriptomics cells are imputed through a graph neural network, where nodes are genes, and edges are based on cosine similarity between the ESM-2 embeddings of the gene-encoding proteins. The gene prediction uncertainty is further measured through a deep learning model. stImpute was shown to consistently outperform state-of-the-art methods across multiple datasets concerning imputation and clustering. stImpute also demonstrates robustness in producing consistent results that are insensitive to model parameters.

PMID:39369061 | DOI:10.1038/s42003-024-06964-2

Sci Rep. 2024 Oct 5;14(1):23199. doi: 10.1038/s41598-024-73695-z.

ABSTRACT

Deep neural networks are increasingly used in medical imaging for tasks such as pathological classification, but they face challenges due to the scarcity of high-quality, expert-labeled training data. Recent efforts have utilized pre-trained contrastive image-text models like CLIP, adapting them for medical use by fine-tuning the model with chest X-ray images and corresponding reports for zero-shot pathology classification, thus eliminating the need for pathology-specific annotations. However, most studies continue to use the same contrastive learning objectives as in the general domain, overlooking the multi-labeled nature of medical image-report pairs. In this paper, we propose a new fine-tuning strategy that includes positive-pair loss relaxation and random sentence sampling. We aim to improve the performance of zero-shot pathology classification without relying on external knowledge. Our method can be applied to any pre-trained contrastive image-text encoder and easily transferred to out-of-domain datasets without further training, as it does not use external data. Our approach consistently improves overall zero-shot pathology classification across four chest X-ray datasets and three pre-trained models, with an average macro AUROC increase of 4.3%. Additionally, our method outperforms the state-of-the-art and marginally surpasses board-certified radiologists in zero-shot classification for the five competition pathologies in the CheXpert dataset.

PMID:39369048 | DOI:10.1038/s41598-024-73695-z

Sci Rep. 2024 Oct 5;14(1):23179. doi: 10.1038/s41598-024-72763-8.

ABSTRACT

Colonoscopy is widely recognized as the most effective method for the detection of colon polyps, which is crucial for early screening of colorectal cancer. Polyp identification and segmentation in colonoscopy images require specialized medical knowledge and are often labor-intensive and expensive. Deep learning provides an intelligent and efficient approach for polyp segmentation. However, the variability in polyp size and the heterogeneity of polyp boundaries and interiors pose challenges for accurate segmentation. Currently, Transformer-based methods have become a mainstream trend for polyp segmentation. However, these methods tend to overlook local details due to the inherent characteristics of Transformer, leading to inferior results. Moreover, the computational burden brought by self-attention mechanisms hinders the practical application of these models. To address these issues, we propose a novel CNN-Transformer hybrid model for polyp segmentation (CTHP). CTHP combines the strengths of CNN, which excels at modeling local information, and Transformer, which excels at modeling global semantics, to enhance segmentation accuracy. We transform the self-attention computation over the entire feature map into the width and height directions, significantly improving computational efficiency. Additionally, we design a new information propagation module and introduce additional positional bias coefficients during the attention computation process, which reduces the dispersal of information introduced by deep and mixed feature fusion in the Transformer. Extensive experimental results demonstrate that our proposed model achieves state-of-the-art performance on multiple benchmark datasets for polyp segmentation. Furthermore, cross-domain generalization experiments show that our model exhibits excellent generalization performance.

PMID:39369043 | DOI:10.1038/s41598-024-72763-8

Sci Rep. 2024 Oct 5;14(1):23237. doi: 10.1038/s41598-024-73665-5.

ABSTRACT

In the domain of medical imaging, the advent of deep learning has marked a significant progression, particularly in the nuanced area of periodontal disease diagnosis. This study specifically targets the prevalent issue of scarce labeled data in medical imaging. We introduce a novel unsupervised few-shot learning algorithm, meticulously crafted for classifying periodontal diseases using a limited collection of dental panoramic radiographs. Our method leverages UNet architecture for generating regions of interest (RoI) from radiographs, which are then processed through a Convolutional Variational Autoencoder (CVAE). This approach is pivotal in extracting critical latent features, subsequently clustered using an advanced algorithm. This clustering is key in our methodology, enabling the assignment of labels to images indicative of periodontal diseases, thus circumventing the challenges posed by limited datasets. Our validation process, involving a comparative analysis with traditional supervised learning and standard autoencoder-based clustering, demonstrates a marked improvement in both diagnostic accuracy and efficiency. For three real-world validation datasets, our UNet-CVAE architecture achieved up to average 14% higher accuracy compared to state-of-the-art supervised models including the vision transformer model when trained with 100 labeled images. This study not only highlights the capability of unsupervised learning in overcoming data limitations but also sets a new benchmark for diagnostic methodologies in medical AI, potentially transforming practices in data-constrained scenarios.

PMID:39369017 | DOI:10.1038/s41598-024-73665-5

Nat Commun. 2024 Oct 5;15(1):8649. doi: 10.1038/s41467-024-51653-7.

ABSTRACT

Electrolytes play a critical role in designing next-generation battery systems, by allowing efficient ion transfer, preventing charge transfer, and stabilizing electrode-electrolyte interfaces. In this work, we develop a differentiable geometric deep learning (GDL) model for chemical mixtures, DiffMix, which is applied in guiding robotic experimentation and optimization towards fast-charging battery electrolytes. In particular, we extend mixture thermodynamic and transport laws by creating GDL-learnable physical coefficients. We evaluate our model with mixture thermodynamics and ion transport properties, where we show improved prediction accuracy and model robustness of DiffMix than its purely data-driven variants. Furthermore, with a robotic experimentation setup, Clio, we improve ionic conductivity of electrolytes by over 18.8% within 10 experimental steps, via differentiable optimization built on DiffMix gradients. By combining GDL, mixture physics laws, and robotic experimentation, DiffMix expands the predictive modeling methods for chemical mixtures and enables efficient optimization in large chemical spaces.

PMID:39369004 | DOI:10.1038/s41467-024-51653-7

Sci Data. 2024 Oct 5;11(1):1088. doi: 10.1038/s41597-024-03944-3.

ABSTRACT

Accurate detection and classification of lung malignancies are crucial for early diagnosis, treatment planning, and patient prognosis. Conventional histopathological analysis is time-consuming, limiting its clinical applicability. To address this, we present a dataset of 691 high-resolution (1200 × 1600 pixels) histopathological lung images, covering adenocarcinomas, squamous cell carcinomas, and normal tissues from 45 patients. These images are subdivided into three differentiation levels for both pathological types: well, moderately, and poorly differentiated, resulting in seven classes for classification. The dataset includes images at 20x and 40x magnification, reflecting real clinical diversity. We evaluated image classification using deep neural network and multiple instance learning approaches. Each method was used to classify images at 20x and 40x magnification into three superclasses. We achieved accuracies between 81% and 92%, depending on the method and resolution, demonstrating the dataset's utility.

PMID:39368979 | DOI:10.1038/s41597-024-03944-3

Asian J Surg. 2024 Oct 4:S1015-9584(24)02201-2. doi: 10.1016/j.asjsur.2024.09.156. Online ahead of print.

ABSTRACT

OBJECTIVE: To delve deeper into the study of degenerative diseases, it becomes imperative to investigate whether deep-learning reconstruction (DLR) can improve the evaluation of white matter hyperintensity (WMH) on 3.0T scanners, and compare its lesion detection capabilities with conventional reconstruction (CR).

METHODS: A total of 131 participants (mean age, 46 years ±17; 46 men) were included in the study. The images of these participants were evaluated by readers blinded to clinical data. Two readers independently assessed subjective image indicators on a 4-point scale. The severity of WMH was assessed by four raters using the Fazekas scale. To evaluate the relative detection capabilities of each method, we employed the Wilcoxon signed rank test to compare scores between the DLR and the CR group. Additionally, we assessed interrater reliability using weighted k statistics and intraclass correlation coefficient to test consistency among the raters.

RESULTS: In terms of subjective image scoring, the DLR group exhibited significantly better scores compared to the CR group (P < 0.001). Regarding the severity of WMH, the DL group demonstrated superior performance in detecting lesions. Majority readers agreed that the DL group provided clearer visualization of the lesions compared to the conventional group.

CONCLUSION: DLR exhibits notable advantages over CR, including subjective image quality, lesion detection sensitivity, and inter reader reliability.

PMID:39368951 | DOI:10.1016/j.asjsur.2024.09.156

BMC Cancer. 2024 Oct 5;24(1):1229. doi: 10.1186/s12885-024-13018-7.

ABSTRACT

OBJECTIVE: To evaluate the diagnostic efficacy of a deep learning (DL) model based on PET/CT images for distinguishing and predicting various pathological subtypes of invasive lung adenocarcinoma.

METHODS: A total of 250 patients diagnosed with invasive lung cancer were included in this retrospective study. The pathological subtypes of the cancer were recorded. PET/CT images were analyzed, including measurements and recordings of the short and long diameters on the maximum cross-sectional plane of the CT image, the density of the lesion, and the associated imaging signs. The SUVmax, SUVmean, and the lesion's long and short diameters on the PET image were also measured. A manual diagnostic model was constructed to analyze its diagnostic performance across different pathological subtypes. The acquired images were first denoised, followed by data augmentation to expand the dataset. The U-Net network architecture was then employed for feature extraction and network segmentation. The classification network was based on the ResNet residual network to address the issue of gradient vanishing in deep networks. Batch normalization was applied to ensure the feature matrix followed a distribution with a mean of 0 and a variance of 1. The images were divided into training, validation, and test sets in a ratio of 6:2:2 to train the model. The deep learning model was then constructed to analyze its diagnostic performance across different pathological subtypes.

RESULTS: Statistically significant differences (P < 0.05) were observed among the four different subtypes in PET/CT imaging performance. The AUC and diagnostic accuracy of the manual diagnostic model for different pathological subtypes were as follows: APA: 0.647, 0.664; SPA: 0.737, 0.772; PPA: 0.698, 0.780; LPA: 0.849, 0.904. Chi-square tests indicated significant statistical differences among these subtypes (P < 0.05). The AUC and diagnostic accuracy of the deep learning model for the different pathological subtypes were as follows: APA: 0.854, 0.864; SPA: 0.930, 0.936; PPA: 0.878, 0.888; LPA: 0.900, 0.920. Chi-square tests also indicated significant statistical differences among these subtypes (P < 0.05). The Delong test showed that the diagnostic performance of the deep learning model was superior to that of the manual diagnostic model (P < 0.05).

CONCLUSIONS: The deep learning model based on PET/CT images exhibits high diagnostic efficacy in distinguishing and diagnosing various pathological subtypes of invasive lung adenocarcinoma, demonstrating the significant potential of deep learning techniques in accurately identifying and predicting disease subgroups.

PMID:39369213 | DOI:10.1186/s12885-024-13018-7

Abdom Radiol (NY). 2024 Oct 5. doi: 10.1007/s00261-024-04597-x. Online ahead of print.

ABSTRACT

Focal liver lesions (FLL) are common incidental findings in abdominal imaging. While the majority of FLLs are benign and asymptomatic, some can be malignant or pre-malignant, and need accurate detection and classification. Current imaging techniques, such as computed tomography (CT) and magnetic resonance imaging (MRI), play a crucial role in assessing these lesions. Artificial intelligence (AI), particularly deep learning (DL), offers potential solutions by analyzing large data to identify patterns and extract clinical features that aid in the early detection and classification of FLLs. This manuscript reviews the diagnostic capacity of AI-based algorithms in processing CT and MRIs to detect benign and malignant FLLs, with an emphasis in the characterization and classification of these lesions and focusing on differentiating benign from pre-malignant and potentially malignant lesions. A comprehensive literature search from January 2010 to April 2024 identified 45 relevant studies. The majority of AI systems employed convolutional neural networks (CNNs), with expert radiologists providing reference standards through manual lesion delineation, and histology as the gold standard. The studies reviewed indicate that AI-based algorithms demonstrate high accuracy, sensitivity, specificity, and AUCs in detecting and characterizing FLLs. These algorithms excel in differentiating between benign and malignant lesions, optimizing diagnostic protocols, and reducing the needs of invasive procedures. Future research should concentrate on the expansion of data sets, the improvement of model explainability, and the validation of AI tools across a range of clinical setting to ensure the applicability and reliability of such tools.

PMID:39369107 | DOI:10.1007/s00261-024-04597-x

Sci Rep. 2024 Oct 5;14(1):23239. doi: 10.1038/s41598-024-74214-w.

ABSTRACT

Network is an essential tool today, and the Intrusion Detection System (IDS) can ensure the safe operation. However, with the explosive growth of data, current methods are increasingly struggling as they often detect based on a single scale, leading to the oversight of potential features in the extensive traffic data, which may result in degraded performance. In this work, we propose a novel detection model utilizing multi-scale transformer namely IDS-MTran. In essence, the collaboration of multi-scale traffic features broads the pattern coverage of intrusion detection. Firstly, we employ convolution operators with various kernels to generate multi-scale features. Secondly, to enhance the representation of features and the interaction between branches, we propose Patching with Pooling (PwP) to serve as a bridge. Next, we design multi-scale transformer-based backbone to model the features at diverse scales, extracting potential intrusion trails. Finally, to fully capitalize these multi-scale branches, we propose the Cross Feature Enrichment (CFE) to integrate and enrich features, and then output the results. Sufficient experiments show that compared with other models, the proposed method can distinguish different attack types more effectively. Specifically, the accuracy on three common datasets NSL-KDD, CIC-DDoS 2019 and UNSW-NB15 has all exceeded 99%, which is more accurate and stable.

PMID:39369065 | DOI:10.1038/s41598-024-74214-w