PLoS One. 2024 Feb 13;19(2):e0298132. doi: 10.1371/journal.pone.0298132. eCollection 2024.

ABSTRACT

PURPOSE: Measurements of macular pigment optical density (MPOD) using the autofluorescence spectroscopy yield underestimations of actual values in eyes with cataracts. Previously, we proposed a correction method for this error using deep learning (DL); however, the correction performance was validated through internal cross-validation. This cross-sectional study aimed to validate this approach using an external validation dataset.

METHODS: MPODs at 0.25°, 0.5°, 1°, and 2° eccentricities and macular pigment optical volume (MPOV) within 9° eccentricity were measured using SPECTRALIS (Heidelberg Engineering, Heidelberg, Germany) in 197 (training dataset inherited from our previous study) and 157 eyes (validating dataset) before and after cataract surgery. A DL model was trained to predict the corrected value from the pre-operative value using the training dataset, and we measured the discrepancy between the corrected value and the actual postoperative value. Subsequently, the prediction performance was validated using a validation dataset.

RESULTS: Using the validation dataset, the mean absolute values of errors for MPOD and MPOV corrected using DL ranged from 8.2 to 12.4%, which were lower than values with no correction (P < 0.001, linear mixed model with Tukey's test). The error depended on the autofluorescence image quality used to calculate MPOD. The mean errors in high and moderate quality images ranged from 6.0 to 11.4%, which were lower than those of poor quality images.

CONCLUSION: The usefulness of the DL correction method was validated. Deep learning reduced the error for a relatively good autofluorescence image quality. Poor-quality images were not corrected.

PMID:38349916 | DOI:10.1371/journal.pone.0298132

J Am Med Inform Assoc. 2024 Feb 13:ocae025. doi: 10.1093/jamia/ocae025. Online ahead of print.

ABSTRACT

OBJECTIVE: Deep-learning techniques, particularly the Transformer model, have shown great potential in enhancing the prediction performance of longitudinal health records. Previous methods focused on fixed-time risk prediction, however, time-to-event prediction is often more appropriate for clinical scenarios. Here, we present STRAFE, a generalizable survival analysis Transformer-based architecture for electronic health records.

MATERIALS AND METHODS: The input for STRAFE is a sequence of visits with SNOMED-CT codes in OMOP-CDM format. A Transformer-based architecture was developed to calculate probabilities of the occurrence of the event in each of 48 months. Performance was evaluated using a real-world claims dataset of over 130 000 individuals with stage 3 chronic kidney disease (CKD).

RESULTS: STRAFE showed improved mean absolute error (MAE) compared to other time-to-event algorithms in predicting the time to deterioration to stage 5 CKD. Additionally, STRAFE showed an improved area under the receiver operating curve compared to binary outcome algorithms. We show that STRAFE predictions can improve the positive predictive value of high-risk patients by 3-fold. Finally, we suggest a novel visualization approach to predictions on a per-patient basis.

DISCUSSION: Time-to-event predictions are the most appropriate approach for clinical predictions. Our deep-learning algorithm outperformed not only other time-to-event prediction algorithms but also fixed-time algorithms, possibly due to its ability to train on censored data. We demonstrated possible clinical usage by identifying the highest-risk patients.

CONCLUSIONS: The ability to accurately identify patients at high risk and prioritize their needs can result in improved health outcomes, reduced costs, and more efficient use of resources.

PMID:38349850 | DOI:10.1093/jamia/ocae025

IEEE Trans Neural Syst Rehabil Eng. 2024 Feb 13;PP. doi: 10.1109/TNSRE.2024.3364752. Online ahead of print.

ABSTRACT

OBJECTIVE: Brain-computer interfaces (BCIs) can enable direct communication with assistive devices by recording and decoding signals from the brain. To achieve high performance, many electrodes will be used, such as the recently developed invasive BCIs with channel numbers up to hundreds or even thousands. For those high-throughput BCIs, channel selection is important to reduce signal redundancy and invasiveness while maintaining decoding performance. However, such endeavour is rarely reported for invasive BCIs, especially those using deep learning methods.

METHODS: Two deep learning-based methods, referred to as Gumbel and STG, were proposed in this paper. They were evaluated using the Stereo-electroencephalography (SEEG) signals, and compared with three other methods, including manual selection, mutual information-based method (MI), and all channels (all channels without selection). The task is to classify the SEEG signals into five movements using channels selected by each method.

RESULTS: When 10 channels were selected, the mean classification accuracies using Gumbel, STG (referred to as STG-10), manual selection, and MI selection were 65%, 60%, 60%, and 47%, respectively, whilst the accuracy was 59% using all channels (no selection). In addition, an investigation of the selected channels showed that Gumbel and STG have successfully identified the pre-central and post-central areas, which are closely related to motor control.

CONCLUSION: Both Gumbel and STG successfully selected the informative channels in SEEG recordings while maintaining decoding accuracy.

SIGNIFICANCE: This study enables future high-throughput BCIs using deep learning methods, to identify useful channels and reduce computing and wireless transmission pressure.

PMID:38349834 | DOI:10.1109/TNSRE.2024.3364752

IEEE Trans Neural Syst Rehabil Eng. 2024 Feb 13;PP. doi: 10.1109/TNSRE.2024.3365713. Online ahead of print.

ABSTRACT

Cross-subject electroencephalogram (EEG) based seizure subtype classification is very important in precise epilepsy diagnostics. Deep learning is a promising solution, due to its ability to automatically extract latent patterns. However, it usually requires a large amount of training data, which may not always be available in clinical practice. This paper proposes Multi-Branch Mutual-Distillation (MBMD) Transformer for cross-subject EEG-based seizure subtype classification, which can be effectively trained from small labeled data. MBMD Transformer replaces all even-numbered encoder blocks of the vanilla Vision Transformer by our designed multi-branch encoder blocks. A mutual-distillation strategy is proposed to transfer knowledge between the raw EEG data and its wavelets of different frequency bands. Experiments on two public EEG datasets demonstrated that our proposed MBMD Transformer outperformed several traditional machine learning and state-of-the-art deep learning approaches. To our knowledge, this is the first work on knowledge distillation for EEG-based seizure subtype classification.

PMID:38349833 | DOI:10.1109/TNSRE.2024.3365713

IEEE J Biomed Health Inform. 2024 Feb 13;PP. doi: 10.1109/JBHI.2024.3365051. Online ahead of print.

ABSTRACT

When decoding neuroelectrophysiological signals represented by Magnetoencephalography (MEG), deep learning models generally achieve high predictive performance but lack the ability to interpret their predicted results. This limitation prevents them from meeting the essential requirements of reliability and ethical-legal considerations in practical applications. In contrast, intrinsically interpretable models, such as decision trees, possess self-evident interpretability while typically sacrificing accuracy. To effectively combine the respective advantages of both deep learning and intrinsically interpretable models, an MEG transfer approach through feature attribution-based knowledge distillation is pioneered, which transforms deep models (teacher) into highly accurate intrinsically interpretable models (student). The resulting models provide not only intrinsic interpretability but also high predictive performance, besides serving as an excellent approximate proxy to understand the inner workings of deep models. In the proposed approach, post-hoc feature knowledge derived from post-hoc interpretable algorithms, specifically feature attribution maps, is introduced into knowledge distillation for the first time. By guiding intrinsically interpretable models to assimilate this knowledge, the transfer of MEG decoding information from deep models to intrinsically interpretable models is implemented. Experimental results demonstrate that the proposed approach outperforms the benchmark knowledge distillation algorithms. This approach successfully improves the prediction accuracy of Soft Decision Tree by a maximum of 8.28%, reaching almost equivalent or even superior performance to deep teacher models. Furthermore, the model-agnostic nature of this approach offers broad application potential.

PMID:38349827 | DOI:10.1109/JBHI.2024.3365051

Radiol Med. 2024 Feb 13. doi: 10.1007/s11547-024-01787-x. Online ahead of print.

ABSTRACT

INTRODUCTION: Low back pain is a global health issue causing disability and missed work days. Commonly used MRI scans including T1-weighted and T2-weighted images provide detailed information of the spine and surrounding tissues. Artificial intelligence showed promise in improving image quality and simultaneously reducing scan time. This study evaluates the performance of deep learning (DL)-based T2 turbo spin-echo (TSE, T2DLR) and T1 TSE (T1DLR) in lumbar spine imaging regarding acquisition time, image quality, artifact resistance, and diagnostic confidence.

MATERIAL AND METHODS: This retrospective monocentric study included 60 patients with lower back pain who underwent lumbar spinal MRI between February and April 2023. MRI parameters and DL reconstruction (DLR) techniques were utilized to acquire images. Two neuroradiologists independently evaluated image datasets based on various parameters using a 4-point Likert scale.

RESULTS: Accelerated imaging showed significantly less image noise and artifacts, as well as better image sharpness, compared to standard imaging. Overall image quality and diagnostic confidence were higher in accelerated imaging. Relevant disk herniations and spinal fractures were detected in both DLR and conventional images. Both readers favored accelerated imaging in the majority of examinations. The lumbar spine examination time was cut by 61% in accelerated imaging compared to standard imaging.

CONCLUSION: In conclusion, the utilization of deep learning-based image reconstruction techniques in lumbar spinal imaging resulted in significant time savings of up to 61% compared to standard imaging, while also improving image quality and diagnostic confidence. These findings highlight the potential of these techniques to enhance efficiency and accuracy in clinical practice for patients with lower back pain.

PMID:38349416 | DOI:10.1007/s11547-024-01787-x

Int J Surg. 2024 Feb 13. doi: 10.1097/JS9.0000000000001194. Online ahead of print.

ABSTRACT

BACKGROUND: Muscle invasive bladder cancer (MIBC) has a poor prognosis even after radical cystectomy (RC). Postoperative survival stratification based on radiomics and deep learning algorithms may be useful for treatment decision-making and follow-up management. This study was aimed to develop and validate a deep learning (DL) model based on preoperative CT for predicting post-cystectomy overall survival in patients with MIBC.

METHODS: MIBC patients who underwent RC were retrospectively included from four centers, and divided into the training, internal validation and external validation sets. A deep learning model incorporated the convolutional block attention module (CBAM) was built for predicting overall survival using preoperative CT images. We assessed the prognostic accuracy of the DL model and compared it with classic handcrafted radiomics model and clinical model. Then, a deep learning radiomics nomogram (DLRN) was developed by combining clinicopathological factors, radiomics score (Rad-score) and deep learning score (DL-score). Model performance was assessed by C-index, KM curve, and time-dependent ROC curve.

RESULTS: A total of 405 patients with MIBC were included in this study. The DL-score achieved a much higher C-index than Rad-score and clinical model (0.690 vs. 0.652 vs. 0.618 in the internal validation set, and 0.658 vs. 0.601 vs. 0.610 in the external validation set). After adjusting for clinicopathologic variables, the DL-score was identified as a significantly independent risk factor for OS by the multivariate Cox regression analysis in all sets (all P<0.01). The DLRN further improved the performance, with a C-index of 0.713 (95%CI: 0.627-0.798) in the internal validation set and 0.685 (95%CI: 0.586-0.765) in external validation set, respectively.

CONCLUSIONS: A DL model based on preoperative CT can predict survival outcome of patients with MIBC, which may help in risk stratification and guide treatment decision-making and follow-up management.

PMID:38349205 | DOI:10.1097/JS9.0000000000001194

Brief Bioinform. 2024 Jan 22;25(2):bbae031. doi: 10.1093/bib/bbae031.

ABSTRACT

Single-cell RNA sequencing (scRNA-seq) has emerged as a powerful tool to gain biological insights at the cellular level. However, due to technical limitations of the existing sequencing technologies, low gene expression values are often omitted, leading to inaccurate gene counts. Existing methods, including advanced deep learning techniques, struggle to reliably impute gene expressions due to a lack of mechanisms that explicitly consider the underlying biological knowledge of the system. In reality, it has long been recognized that gene-gene interactions may serve as reflective indicators of underlying biology processes, presenting discriminative signatures of the cells. A genomic data analysis framework that is capable of leveraging the underlying gene-gene interactions is thus highly desirable and could allow for more reliable identification of distinctive patterns of the genomic data through extraction and integration of intricate biological characteristics of the genomic data. Here we tackle the problem in two steps to exploit the gene-gene interactions of the system. We first reposition the genes into a 2D grid such that their spatial configuration reflects their interactive relationships. To alleviate the need for labeled ground truth gene expression datasets, a self-supervised 2D convolutional neural network is employed to extract the contextual features of the interactions from the spatially configured genes and impute the omitted values. Extensive experiments with both simulated and experimental scRNA-seq datasets are carried out to demonstrate the superior performance of the proposed strategy against the existing imputation methods.

PMID:38349062 | DOI:10.1093/bib/bbae031

Brief Bioinform. 2024 Jan 22;25(2):bbae027. doi: 10.1093/bib/bbae027.

ABSTRACT

The recent development of deep learning methods have undoubtedly led to great improvement in various machine learning tasks, especially in prediction tasks. This type of methods have also been adapted to answer various problems in bioinformatics, including automatic genome annotation, artificial genome generation or phenotype prediction. In particular, a specific type of deep learning method, called graph neural network (GNN) has repeatedly been reported as a good candidate to predict phenotypes from gene expression because its ability to embed information on gene regulation or co-expression through the use of a gene network. However, up to date, no complete and reproducible benchmark has ever been performed to analyze the trade-off between cost and benefit of this approach compared to more standard (and simpler) machine learning methods. In this article, we provide such a benchmark, based on clear and comparable policies to evaluate the different methods on several datasets. Our conclusion is that GNN rarely provides a real improvement in prediction performance, especially when compared to the computation effort required by the methods. Our findings on a limited but controlled simulated dataset shows that this could be explained by the limited quality or predictive power of the input biological gene network itself.

PMID:38349060 | DOI:10.1093/bib/bbae027

Brief Bioinform. 2024 Jan 22;25(2):bbae040. doi: 10.1093/bib/bbae040.

ABSTRACT

Efficient and accurate recognition of protein-DNA interactions is vital for understanding the molecular mechanisms of related biological processes and further guiding drug discovery. Although the current experimental protocols are the most precise way to determine protein-DNA binding sites, they tend to be labor-intensive and time-consuming. There is an immediate need to design efficient computational approaches for predicting DNA-binding sites. Here, we proposed ULDNA, a new deep-learning model, to deduce DNA-binding sites from protein sequences. This model leverages an LSTM-attention architecture, embedded with three unsupervised language models that are pre-trained on large-scale sequences from multiple database sources. To prove its effectiveness, ULDNA was tested on 229 protein chains with experimental annotation of DNA-binding sites. Results from computational experiments revealed that ULDNA significantly improves the accuracy of DNA-binding site prediction in comparison with 17 state-of-the-art methods. In-depth data analyses showed that the major strength of ULDNA stems from employing three transformer language models. Specifically, these language models capture complementary feature embeddings with evolution diversity, in which the complex DNA-binding patterns are buried. Meanwhile, the specially crafted LSTM-attention network effectively decodes evolution diversity-based embeddings as DNA-binding results at the residue level. Our findings demonstrated a new pipeline for predicting DNA-binding sites on a large scale with high accuracy from protein sequence alone.

PMID:38349057 | DOI:10.1093/bib/bbae040

J Sci Food Agric. 2024 Feb 13. doi: 10.1002/jsfa.13381. Online ahead of print.

ABSTRACT

BACKGROUND: Monitoring and controlling the moisture content throughout the Tencha drying processing procedure is crucial for ensuring its quality. Workers often rely on their senses to perceive the moisture content, leading to relative subjectivity and low reproducibility. The traditional drying methods for measuring moisture content is destructive to samples. This research was conducted using computer vision combined with deep learning for detecting moisture content during the Tencha drying process. Different color space components of Tencha drying samples' image were first extracted by computer vision. The color components were preprocessed using MinMax and Z-score. Subsequently, one-dimensional convolutional neural network (1D-CNN), partial least squares, and backpropagation artificial neural network models were built and compared.

RESULTS: The 1D-CNN model and Z-score preprocessing achieved superior predictive accuracy, with correlation coefficient of prediction (Rp ) = 0.9548 for moisture content. Furthermore, the migration of moisture content during the Tencha drying process was eventually visualized by mapping its spatial and temporal distributions.

CONCLUSION: The results indicated computer vision combined with 1D-CNN was feasible for moisture prediction during Tencha drying process. This study provides technical support for the industrial and intelligent production of Tencha. This article is protected by copyright. All rights reserved.

PMID:38349009 | DOI:10.1002/jsfa.13381

Pol Przegl Chir. 2023 Nov 24;96(0):114-117. doi: 10.5604/01.3001.0053.9857.

ABSTRACT

&lt;b&gt;&lt;br&gt;Introduction:&lt;/b&gt; Artificial intelligence (AI) is an emerging technology with vast potential for use in several fields of medicine. However, little is known about the application of AI in treatment decisions for patients with polytrauma. In this systematic review, we investigated the benefits and performance of AI in predicting the management of patients with polytrauma and trauma.&lt;/br&gt; &lt;b&gt;&lt;br&gt;Methods:&lt;/b&gt; This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Studies were extracted from the PubMed and Google Scholar databases from their inception until November 2022, using the search terms "Artificial intelligence," "polytrauma," and "decision." Seventeen articles were identified and screened for eligibility. Animal studies, review articles, systematic reviews, meta-analyses, and studies that did not involve polytrauma or severe trauma management decisions were excluded. Eight studies were eligible for final review.&lt;/br&gt; &lt;b&gt;&lt;br&gt;Results:&lt;/b&gt; Eight studies focusing on patients with trauma, including two on military trauma, were included. The AI applications were mainly implemented for predictions and/or decisions on shock, bleeding, and blood transfusion. Few studies predicted death/survival. The identification of trauma patients using AI was proposed in a previous study. The overall performance of AI was good (six studies), excellent (one study), and acceptable (one study).&lt;/br&gt; &lt;b&gt;&lt;br&gt;Discussion:&lt;/b&gt; AI demonstrated satisfactory performance in decision-making and management prediction in patients with polytrauma/severe trauma, especially in situations of shock/bleeding.&lt;/br&gt; &lt;b&gt;&lt;br&gt;Importance:&lt;/b&gt; The present study serves as a basis for further research to develop practical AI applications for the management of patients with trauma.&lt;/br&gt.

PMID:38348980 | DOI:10.5604/01.3001.0053.9857

Int J Surg. 2024 Feb 12. doi: 10.1097/JS9.0000000000001161. Online ahead of print.

ABSTRACT

BACKGROUND: Tumor-stroma interactions, as indicated by tumor-stroma ratio (TSR), offer valuable prognostic stratification information. Current histological assessment of TSR is limited by tissue accessibility and spatial heterogeneity. We aimed to develop a multitask deep learning (MDL) model to noninvasively predict TSR and prognosis in colorectal cancer (CRC).

MATERIALS AND METHODS: In this retrospective study including 2268 patients with resected CRC recruited from four centers, we developed an MDL model using preoperative CT images for the simultaneous prediction of TSR and overall survival. Patients in the training cohort (n=956) and internal validation cohort (IVC, n=240) were randomly selected from center I. Patients in the external validation cohort1(EVC1, n=509), EVC2 (n=203), and EVC3 (n=360) were recruited from other three centers. Model performance was evaluated with respect to discrimination and calibration. Furthermore, we evaluated whether the model could predict the benefit from adjuvant chemotherapy.

RESULTS: The MDL model demonstrated strong TSR discrimination, yielding areas under the receiver operating curves (AUCs) of 0.855 (95%CI, 0.800-0.910), 0.838(95% CI, 0.802-0.874), and 0.857(95% CI, 0.804-0.909) in the three validation cohorts, respectively. The MDL model was also able to predict overall survival and disease-free survival across all cohorts. In multivariable Cox analysis, the MDL score (MDLS) remained an independent prognostic factor after adjusting for clinicopathological variables (all P<0.05). For stage II and stage III disease, patients with a high MDLS benefited from adjuvant chemotherapy (hazard ratio [HR] 0.391 [95%CI, 0.230-0.666], P=0.0003; HR=0.467[95%CI, 0.331-0.659], P<0.0001, respectively), whereas those with a low MDLS did not.

CONCLUSION: The multitask DL model based on preoperative CT images effectively predicted TSR status and survival in CRC patients, offering valuable guidance for personalized treatment. Prospective studies are needed to confirm its potential to select patients who might benefit from chemotherapy.

PMID:38348900 | DOI:10.1097/JS9.0000000000001161

Int J Surg. 2024 Feb 12. doi: 10.1097/JS9.0000000000001186. Online ahead of print.

ABSTRACT

OBJECTIVES: We aimed to assess the performance of a deep learning (DL) model, based on a combination of ultrasound (US) and mammography (MG) images, for predicting malignancy in breast lesions categorized as Breast Imaging Reporting and Data System (BI-RADS) US 4A in diagnostic patients with dense breasts.

METHODS: A total of 992 patients were randomly allocated into the training cohort and the test cohort at a proportion of 4:1. Another, 218 patients were enrolled to form a prospective validation cohort. The DL model was developed by incorporating both US and MG images. The predictive performance of the combined DL model for malignancy was evaluated by sensitivity, specificity and area under the receiver operating characteristic curve (AUC). The combined DL model was then compared to a clinical nomogram model and to the DL model trained using US image only and to that trained MG image only.

RESULTS: The combined DL model showed satisfactory diagnostic performance for predicting malignancy in breast lesions, with an AUC of 0.940 (95% confidence interval [95%CI], 0.874~1.000) in the test cohort, and an AUC of 0.906 (95%CI, 0.817~0.995) in the validation cohort, which was significantly higher than the clinical nomogram model, and the DL model for US or MG alone (P<0.05).

CONCLUSIONS: The study developed an objective DL model combining both US and MG imaging features, which was proven to be more accurate for predicting malignancy in the BI-RADS US 4A breast lesions of patients with dense breasts. This model may then be used to more accurately guide clinicians' choices about whether performing biopsies in breast cancer diagnosis.

PMID:38348891 | DOI:10.1097/JS9.0000000000001186

Head Neck. 2024 Feb 13. doi: 10.1002/hed.27629. Online ahead of print.

ABSTRACT

BACKGROUND: The preservation of parathyroid glands is crucial in endoscopic thyroid surgery to prevent hypocalcemia and related complications. However, current methods for identifying and protecting these glands have limitations. We propose a novel technique that has the potential to improve the safety and efficacy of endoscopic thyroid surgery.

PURPOSE: Our study aims to develop a deep learning model called PTAIR 2.0 (Parathyroid gland Artificial Intelligence Recognition) to enhance parathyroid gland recognition during endoscopic thyroidectomy. We compare its performance against traditional surgeon-based identification methods.

MATERIALS AND METHODS: Parathyroid tissues were annotated in 32 428 images extracted from 838 endoscopic thyroidectomy videos, forming the internal training cohort. An external validation cohort comprised 54 full-length videos. Six candidate algorithms were evaluated to select the optimal one. We assessed the model's performance in terms of initial recognition time, identification duration, and recognition rate and compared it with the performance of surgeons.

RESULTS: Utilizing the YOLOX algorithm, we developed PTAIR 2.0, which demonstrated superior performance with an AP50 score of 92.1%. The YOLOX algorithm achieved a frame rate of 25.14 Hz, meeting real-time requirements. In the internal training cohort, PTAIR 2.0 achieved AP50 values of 94.1%, 98.9%, and 92.1% for parathyroid gland early prediction, identification, and ischemia alert, respectively. Additionally, in the external validation cohort, PTAIR outperformed both junior and senior surgeons in identifying and tracking parathyroid glands (p < 0.001).

CONCLUSION: The AI-driven PTAIR 2.0 model significantly outperforms both senior and junior surgeons in parathyroid gland identification and ischemia alert during endoscopic thyroid surgery, offering potential for enhanced surgical precision and patient outcomes.

PMID:38348564 | DOI:10.1002/hed.27629

Front Med (Lausanne). 2024 Jan 29;11:1323516. doi: 10.3389/fmed.2024.1323516. eCollection 2024.

ABSTRACT

BACKGROUND: Artificial intelligence-assisted gastroscopy (AIAG) based on deep learning has been validated in various scenarios, but there is a lack of studies regarding diagnosing neoplasms under white light endoscopy. This study explored the potential role of AIAG systems in enhancing the ability of endoscopists to diagnose gastric tumor lesions under white light.

METHODS: A total of 251 patients with complete pathological information regarding electronic gastroscopy, biopsy, or ESD surgery in Xi'an Gaoxin Hospital were retrospectively collected and comprised 64 patients with neoplasm lesions (excluding advanced cancer) and 187 patients with non-neoplasm lesions. The diagnosis competence of endoscopists with intermediate experience and experts was compared for gastric neoplasms with or without the assistance of AIAG, which was developed based on ResNet-50.

RESULTS: For the 251 patients with difficult clinical diagnoses included in the study, compared with endoscopists with intermediate experience, AIAG's diagnostic competence was much higher, with a sensitivity of 79.69% (79.69% vs. 72.50%, p = 0.012) and a specificity of 73.26% (73.26% vs. 52.62%, p < 0.001). With the help of AIAG, the endoscopists with intermediate experience (<8 years) demonstrated a relatively higher specificity (59.79% vs. 52.62%, p < 0.001). Experts (≥8 years) had similar results with or without AI assistance (with AI vs. without AI; sensitivities, 70.31% vs. 67.81%, p = 0.358; specificities, 83.85% vs. 85.88%, p = 0.116).

CONCLUSION: With the assistance of artificial intelligence (AI) systems, the ability of endoscopists with intermediate experience to diagnose gastric neoplasms is significantly improved, but AI systems have little effect on experts.

PMID:38348337 | PMC:PMC10859510 | DOI:10.3389/fmed.2024.1323516

Front Med (Lausanne). 2024 Jan 29;11:1290729. doi: 10.3389/fmed.2024.1290729. eCollection 2024.

ABSTRACT

BACKGROUND: Pneumoconiosis is the most important occupational disease all over the world, with high prevalence and mortality. At present, the monitoring of workers exposed to dust and the diagnosis of pneumoconiosis rely on manual interpretation of chest radiographs, which is subjective and low efficiency. With the development of artificial intelligence technology, a more objective and efficient computer aided system for pneumoconiosis diagnosis can be realized. Therefore, the present study reported a novel deep learning (DL) artificial intelligence (AI) system for detecting pneumoconiosis in digital frontal chest radiographs, based on which we aimed to provide references for radiologists.

METHODS: We annotated 49,872 chest radiographs from patients with pneumoconiosis and workers exposed to dust using a self-developed tool. Next, we used the labeled images to train a convolutional neural network (CNN) algorithm developed for pneumoconiosis screening. Finally, the performance of the trained pneumoconiosis screening model was validated using a validation set containing 495 chest radiographs.

RESULTS: Approximately, 51% (25,435/49,872) of the chest radiographs were labeled as normal. Pneumoconiosis was detected in 49% (24,437/49,872) of the labeled radiographs, among which category-1, category-2, and category-3 pneumoconiosis accounted for 53.1% (12,967/24,437), 20.4% (4,987/24,437), and 26.5% (6,483/24,437) of the patients, respectively. The CNN DL algorithm was trained using these data. The validation set of 495 digital radiography chest radiographs included 261 cases of pneumoconiosis and 234 cases of non-pneumoconiosis. As a result, the accuracy of the AI system for pneumoconiosis identification was 95%, the area under the curve was 94.7%, and the sensitivity was 100%.

CONCLUSION: DL algorithm based on CNN helped screen pneumoconiosis in the chest radiographs with high performance; thus, it could be suitable for diagnosing pneumoconiosis automatically and improve the efficiency of radiologists.

PMID:38348336 | PMC:PMC10859417 | DOI:10.3389/fmed.2024.1290729

Data Brief. 2024 Jan 24;53:110076. doi: 10.1016/j.dib.2024.110076. eCollection 2024 Apr.

ABSTRACT

The field of deep learning is rapidly advancing and impacting various industries, including banking. However, there are still challenges when it comes to accurately identifying the denomination of currencies, especially when dealing with issues like variation within the same class of currency and inconsistent lighting conditions. One notable problem is the lack of available data for Kazakhstan's currency. This research paper introduces the Kazakhstan Banknotes Dataset (KZ-BD), which is a unique collection of 4200 carefully annotated images covering 14 different categories. The dataset includes high-resolution images of authentic Kazakhstan Tenge in both coin and paper note forms, ranging from 1 to 20,000 tenge denominations. Each image has undergone strict de-identification and validation procedures, and the dataset is openly accessible to artificial intelligence researchers. This contribution addresses the data gap in deep learning research related to currency identification by offering a comprehensive dataset for Kazakhstan's currency, enabling better evaluation and fine-tuning of machine learning models with real-world data.

PMID:38348328 | PMC:PMC10859250 | DOI:10.1016/j.dib.2024.110076

Front Med (Lausanne). 2024 Jan 29;10:1349336. doi: 10.3389/fmed.2023.1349336. eCollection 2023.

ABSTRACT

INTRODUCTION: Oral Squamous Cell Carcinoma (OSCC) poses a significant challenge in oncology due to the absence of precise diagnostic tools, leading to delays in identifying the condition. Current diagnostic methods for OSCC have limitations in accuracy and efficiency, highlighting the need for more reliable approaches. This study aims to explore the discriminative potential of histopathological images of oral epithelium and OSCC. By utilizing a database containing 1224 images from 230 patients, captured at varying magnifications and publicly available, a customized deep learning model based on EfficientNetB3 was developed. The model's objective was to differentiate between normal epithelium and OSCC tissues by employing advanced techniques such as data augmentation, regularization, and optimization.

METHODS: The research utilized a histopathological imaging database for Oral Cancer analysis, incorporating 1224 images from 230 patients. These images, taken at various magnifications, formed the basis for training a specialized deep learning model built upon the EfficientNetB3 architecture. The model underwent training to distinguish between normal epithelium and OSCC tissues, employing sophisticated methodologies including data augmentation, regularization techniques, and optimization strategies.

RESULTS: The customized deep learning model achieved significant success, showcasing a remarkable 99% accuracy when tested on the dataset. This high accuracy underscores the model's efficacy in effectively discerning between normal epithelium and OSCC tissues. Furthermore, the model exhibited impressive precision, recall, and F1-score metrics, reinforcing its potential as a robust diagnostic tool for OSCC.

DISCUSSION: This research demonstrates the promising potential of employing deep learning models to address the diagnostic challenges associated with OSCC. The model's ability to achieve a 99% accuracy rate on the test dataset signifies a considerable leap forward in earlier and more accurate detection of OSCC. Leveraging advanced techniques in machine learning, such as data augmentation and optimization, has shown promising results in improving patient outcomes through timely and precise identification of OSCC.

PMID:38348235 | PMC:PMC10859441 | DOI:10.3389/fmed.2023.1349336

J Multidiscip Healthc. 2024 Feb 8;17:609-617. doi: 10.2147/JMDH.S442479. eCollection 2024.

ABSTRACT

PURPOSE: This study aimed to establish a stereoscopic neural learning network through deep learning and construct an artificial intelligence (AI) diagnosis system for the prediction of benign and malignant thyroid diseases, as well as repeatedly verified the diagnosis system and adjusted the data, in order to develop a type of AI-assisted thyroid diagnosis software with a low false negative rate and high sensitivity for clinical practice.

PATIENTS AND METHODS: From July 2020 to April 2023, A total of 36 patients with thyroid nodules in our hospital were selected for diagnosis of thyroid nodules based on the Expert Consensus on Thyroid Ultrasound; samples were taken by aspiration biopsy or surgically and sent for pathological diagnosis. The ultrasonic diagnosis results were compared with the pathological results, a database was established based on the ultrasonic diagnostic characteristics and was entered in the AI-assisted diagnosis software for judgment of benign and malignant conditions. The data in the software were corrected based on the conformity rate and the reasons for misjudgment, and the corrected software was used to evaluate the benign and malignant conditions of the 36 patients, until the conformity rate exceeded 90%.

RESULTS: The initial conformity rate of the AI software for identifying benign and malignant conditions was 88%, while that of the software utilizing the database was 94%.

CONCLUSION: We established a stereoscopic neural learning network and construct an AI diagnosis system for the prediction of benign and malignant thyroid diseases, with a low false negative rate and high sensitivity for clinical practice.

PMID:38348208 | PMC:PMC10860491 | DOI:10.2147/JMDH.S442479