Intelligence-based medicine最新文献_第10页

Exploring the intersection of cochlear implants and artificial intelligence: A mixed-method systematic and scoping review 探索人工耳蜗与人工智能的交叉：一种混合方法的系统和范围综述

Intelligence-based medicine

Pub Date : 2025-01-01 DOI: 10.1016/j.ibmed.2025.100296

Aurenzo Gonçalves Mocelin , Pedro Angelo Basei de Paula , Daniel Tiepolo Kochinski , Thayná Cristina Wiezbicki , Rogério de Azevedo Hamerschmidt , Mayara Risnei Watanabe , Rogério Hamerschmidt

Objective

This study systematically evaluates the role of artificial intelligence (AI) in cochlear implant (CI) technology, focusing on speech enhancement, automated fitting, AI-assisted surgery, predictive modeling, and rehabilitation. The review identifies key advancements, existing limitations, and areas for future development.

Methods

Following PRISMA guidelines, we conducted a systematic search across PubMed, IEEE Xplore, Scopus, ScienceDirect, and Embase. We included peer-reviewed primary data studies on AI applications in CIs. The selected studies were categorized into thematic subdomains, such as noise suppression, adaptive programming, AI-driven surgical planning, and telemedicine applications.

Results

From an initial pool of 743 records, 129 studies met the eligibility criteria and were included in the final analysis. These studies were categorized into eleven thematic subdomains. The review identified the main application areas and emerging research fronts at the intersection of artificial intelligence and cochlear implant technologies, including speech enhancement, automated fitting, predictive modeling, rehabilitation support, and AI-assisted surgery.

Discussion and conclusion

AI is transforming CI technology by improving speech perception, personalization, and surgical precision. However, challenges persist, including computational constraints, data heterogeneity, and the need for large-scale clinical validation. Future research should prioritize energy-efficient AI architectures, regulatory approval pathways, and ethical considerations in automated decision-making. Advancing AI-driven telemedicine solutions can expand CI accessibility, reducing the need for in-person programming. Addressing these challenges will accelerate the development of more adaptive and user-centered CI solutions, ultimately enhancing auditory rehabilitation and quality of life for CI users.

目的系统评估人工智能（AI）在人工耳蜗（CI）技术中的作用，重点关注语音增强、自动验配、人工智能辅助手术、预测建模和康复。该审查确定了主要进展、现有限制和未来发展的领域。方法遵循PRISMA指南，我们在PubMed、IEEE explore、Scopus、ScienceDirect和Embase中进行了系统搜索。我们纳入了人工智能在ci中的应用的同行评议的原始数据研究。选定的研究被分类为主题子领域，如噪声抑制、自适应编程、人工智能驱动的手术计划和远程医疗应用。结果从最初的743份记录中，有129项研究符合资格标准，并被纳入最终分析。这些研究分为11个主题子领域。该综述确定了人工智能和人工耳蜗技术交叉的主要应用领域和新兴研究前沿，包括语音增强、自动装配、预测建模、康复支持和人工智能辅助手术。人工智能正在通过提高语音感知、个性化和手术精度来改变CI技术。然而，挑战依然存在，包括计算限制、数据异质性和大规模临床验证的需要。未来的研究应优先考虑节能的人工智能架构、监管审批途径和自动化决策中的道德考虑。推进人工智能驱动的远程医疗解决方案可以扩大CI的可访问性，减少对亲自编程的需求。解决这些挑战将加速开发更具适应性和以用户为中心的CI解决方案，最终提高CI用户的听觉康复和生活质量。

{"title":"Exploring the intersection of cochlear implants and artificial intelligence: A mixed-method systematic and scoping review","authors":"Aurenzo Gonçalves Mocelin , Pedro Angelo Basei de Paula , Daniel Tiepolo Kochinski , Thayná Cristina Wiezbicki , Rogério de Azevedo Hamerschmidt , Mayara Risnei Watanabe , Rogério Hamerschmidt","doi":"10.1016/j.ibmed.2025.100296","DOIUrl":"10.1016/j.ibmed.2025.100296","url":null,"abstract":"<div><h3>Objective</h3><div>This study systematically evaluates the role of artificial intelligence (AI) in cochlear implant (CI) technology, focusing on speech enhancement, automated fitting, AI-assisted surgery, predictive modeling, and rehabilitation. The review identifies key advancements, existing limitations, and areas for future development.</div></div><div><h3>Methods</h3><div>Following PRISMA guidelines, we conducted a systematic search across PubMed, IEEE Xplore, Scopus, ScienceDirect, and Embase. We included peer-reviewed primary data studies on AI applications in CIs. The selected studies were categorized into thematic subdomains, such as noise suppression, adaptive programming, AI-driven surgical planning, and telemedicine applications.</div></div><div><h3>Results</h3><div>From an initial pool of 743 records, 129 studies met the eligibility criteria and were included in the final analysis. These studies were categorized into eleven thematic subdomains. The review identified the main application areas and emerging research fronts at the intersection of artificial intelligence and cochlear implant technologies, including speech enhancement, automated fitting, predictive modeling, rehabilitation support, and AI-assisted surgery.</div></div><div><h3>Discussion and conclusion</h3><div>AI is transforming CI technology by improving speech perception, personalization, and surgical precision. However, challenges persist, including computational constraints, data heterogeneity, and the need for large-scale clinical validation. Future research should prioritize energy-efficient AI architectures, regulatory approval pathways, and ethical considerations in automated decision-making. Advancing AI-driven telemedicine solutions can expand CI accessibility, reducing the need for in-person programming. Addressing these challenges will accelerate the development of more adaptive and user-centered CI solutions, ultimately enhancing auditory rehabilitation and quality of life for CI users.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"12 ","pages":"Article 100296"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145264862","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Expression of concern for DieT Transformer model with PCA-ADE integration for advanced multi-class brain tumor classification by Mohammad Amin, Khalid M.O. Nahar, et al. [Intell.-Base Med. 11, (2025), 100192, https://doi.org/10.1016/j.ibmed.2024.100192] Mohammad Amin, Khalid M.O. Nahar等人对PCA-ADE集成的DieT Transformer模型在晚期多级别脑肿瘤分类中的关注表达[intel]。-基础医学，（2025），100192,https://doi.org/10.1016/j.ibmed.2024.100192]

Intelligence-based medicine

Pub Date : 2025-01-01

引用次数: 0

Forecasting pediatric emergency department arrivals: Evaluating the role of exogenous variables using deep learning models 预测儿科急诊科到达：使用深度学习模型评估外生变量的作用

Intelligence-based medicine

Pub Date : 2025-01-01 DOI: 10.1016/j.ibmed.2025.100313

Egbe-Etu Etu , Jordan Larot , Kindness Etu , Joshua Emakhu , Sara Masoud , Imokhai Tenebe , Gaojian Huang , Satheesh Gunaga , Joseph Miller

Background

Forecasting pediatric emergency department (ED) demand remains a critical challenge in healthcare operations. This study aimed to identify exogenous variables influencing pediatric ED visits and evaluate the performance of different forecasting models.

Method

Using a retrospective observational design, we analyzed 192,347 pediatric ED visits across nine hospitals in Southeast Michigan between 2017 and 2019. Patient data were aggregated into daily arrival counts and enriched with exogenous variables such as weather, air quality, pollen, calendar, Google search trends, and chief complaints. Feature selection was performed using XGBoost and SHapley Additive exPlanations to identify the most influential predictors. Three forecasting models were developed: a Naïve baseline, Long Short-Term Memory (LSTM), and an attention-based neural network. The models were evaluated across 1-day, 7-day, and 14-day forecasting horizons using mean absolute percentage error (MAPE) and R² metrics.

Results

LSTM and attention-based model significantly outperformed the Naïve baseline across all horizons. The LSTM model incorporating calendar data achieved the best 1-day forecast (MAPE: 8.71 %, R²: 0.67). For 7-day forecasts, the attention-based model using chief complaint data performed best (MAPE: 9.18 %, R²: 0.57). At 14 days, the attention-based model without exogenous inputs outperformed most LSTM variants, reflecting superior performance in long-range forecasting. Among exogenous variables, calendar and chief complaint data added the most predictive value, while Google Trends and pollen data introduced noise and diminished model performance.

Conclusion

Combining deep learning architectures with selected external data improves pediatric ED arrival forecasting. From an operational perspective, such forecasts can support more efficient staffing, reduce wait times, and mitigate ED crowding.

背景预测儿科急诊科（ED）的需求仍然是医疗保健业务的关键挑战。本研究旨在确定影响儿科急诊科就诊的外生变量，并评估不同预测模型的性能。方法采用回顾性观察设计，分析2017年至2019年密歇根州东南部9家医院的192,347例儿科急诊科就诊情况。患者数据汇总为每日到达计数，并丰富了外生变量，如天气、空气质量、花粉、日历、谷歌搜索趋势和主诉。使用XGBoost和SHapley加性解释进行特征选择，以确定最具影响力的预测因子。开发了三种预测模型：Naïve基线，长短期记忆（LSTM）和基于注意的神经网络。使用平均绝对百分比误差（MAPE）和R2指标对模型进行1天、7天和14天的预测期评估。结果slstm和基于注意力的模型在所有视界上都显著优于Naïve基线。结合日历数据的LSTM模型获得了最好的1天预测（MAPE: 8.71%, R2: 0.67）。对于7天的预测，使用主诉数据的基于注意力的模型表现最好（MAPE: 9.18%, R2: 0.57）。在第14天，没有外源输入的基于注意力的模型优于大多数LSTM变体，反映出在长期预测方面的优越性能。在外源变量中，日历和主诉数据的预测价值最高，而谷歌趋势和花粉数据引入了噪声，降低了模型的性能。结论将深度学习架构与选定的外部数据相结合可以提高儿科急诊科的到来预测。从操作的角度来看，这样的预测可以支持更有效的人员配置，减少等待时间，并缓解急诊科拥挤。

{"title":"Forecasting pediatric emergency department arrivals: Evaluating the role of exogenous variables using deep learning models","authors":"Egbe-Etu Etu , Jordan Larot , Kindness Etu , Joshua Emakhu , Sara Masoud , Imokhai Tenebe , Gaojian Huang , Satheesh Gunaga , Joseph Miller","doi":"10.1016/j.ibmed.2025.100313","DOIUrl":"10.1016/j.ibmed.2025.100313","url":null,"abstract":"<div><h3>Background</h3><div>Forecasting pediatric emergency department (ED) demand remains a critical challenge in healthcare operations. This study aimed to identify exogenous variables influencing pediatric ED visits and evaluate the performance of different forecasting models.</div></div><div><h3>Method</h3><div>Using a retrospective observational design, we analyzed 192,347 pediatric ED visits across nine hospitals in Southeast Michigan between 2017 and 2019. Patient data were aggregated into daily arrival counts and enriched with exogenous variables such as weather, air quality, pollen, calendar, Google search trends, and chief complaints. Feature selection was performed using XGBoost and SHapley Additive exPlanations to identify the most influential predictors. Three forecasting models were developed: a Naïve baseline, Long Short-Term Memory (LSTM), and an attention-based neural network. The models were evaluated across 1-day, 7-day, and 14-day forecasting horizons using mean absolute percentage error (MAPE) and R<sup>2</sup> metrics.</div></div><div><h3>Results</h3><div>LSTM and attention-based model significantly outperformed the Naïve baseline across all horizons. The LSTM model incorporating calendar data achieved the best 1-day forecast (MAPE: 8.71 %, R<sup>2</sup>: 0.67). For 7-day forecasts, the attention-based model using chief complaint data performed best (MAPE: 9.18 %, R<sup>2</sup>: 0.57). At 14 days, the attention-based model without exogenous inputs outperformed most LSTM variants, reflecting superior performance in long-range forecasting. Among exogenous variables, calendar and chief complaint data added the most predictive value, while Google Trends and pollen data introduced noise and diminished model performance.</div></div><div><h3>Conclusion</h3><div>Combining deep learning architectures with selected external data improves pediatric ED arrival forecasting. From an operational perspective, such forecasts can support more efficient staffing, reduce wait times, and mitigate ED crowding.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"12 ","pages":"Article 100313"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145519506","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A novel pixel pair shuffling based image watermarking for tamper detection and self-recovery 一种基于像素对变换的篡改检测和自恢复图像水印

Intelligence-based medicine

Pub Date : 2025-01-01 DOI: 10.1016/j.ibmed.2025.100324

Radha Ramesh Murapaka , A.V.S. Pavan Kumar , Aditya Kumar Sahu

This work has introduced a novel image watermarking scheme leveraging a pixel pair-based shuffling (PPSh) technique for tamper detection and self-recovery. The proposed technique consists of five steps, initiating from secret bits generation, collectively known as watermark bits. Then, the next step is watermark embedding, after that, watermark extraction, tamper detection, and finally, dual self-recovery approaches have been implemented. For watermark bit generation, two prominent interpolation techniques, such as bipolar and bilinear, are applied to the cover image (CI) to obtain the compressed image. Later, Advanced Encryption Standard (AES) and Camellia with Cipher Block Chaining (CBC) mode of operation is utilized on the compressed image to generate watermark bits. Afterwards, a PPSh-based watermark embedding strategy has been utilized to achieve the watermarked image (WI) while maintaining a standard payload capacity. Further, a variety of image processing attacks is performed on the WI to check the imperceptibility and similarity of the proposed scheme. Consequently, tamper region detection is followed by the watermark extraction procedure. Therefore, to reconstruct the tampered pixels, inpainting based dual recovery approaches are presented, named as TELEA and Naiver-Stokes (NS). The robustness and imperceptibility of the proposed scheme is measured through peak-signal-to-noise ratio (PSNR), structural similarity index matrix (SSIM), and mean square error (MSE). The proposed technique has achieved an average PSNR and SSIM of 54.24 dB and 0.9983, respectively. With an increment of more than 2 dB in terms of PSNR the proposed technique outperforms the existing watermarking techniques. Additionally, the proposed technique obtains a recovery increment up to 5 dB in terms of PSNR for 10 %–50 % tampering rates against the existing methods.

这项工作引入了一种新的图像水印方案，利用基于像素对的洗牌（PPSh）技术进行篡改检测和自我恢复。该技术包括五个步骤，从生成秘密比特（统称为水印比特）开始。然后进行水印嵌入，再进行水印提取、篡改检测，最后实现双自恢复方法。对于水印位的生成，采用双极和双线性两种重要的插值技术对封面图像进行插值，得到压缩后的图像。随后，利用高级加密标准AES （Advanced Encryption Standard）和CBC （Cipher Block chains）操作模式的Camellia在压缩图像上生成水印位。然后，利用基于ppsh的水印嵌入策略，在保持标准载荷容量的情况下实现水印图像。此外，在WI上进行了各种图像处理攻击，以检查所提出方案的不可感知性和相似性。因此，篡改区域检测之后是水印提取程序。因此，为了重建被篡改的像素，提出了基于插值的双重恢复方法，称为TELEA和naver - stokes （NS）。通过峰值信噪比（PSNR）、结构相似指数矩阵（SSIM）和均方误差（MSE）来衡量该方案的鲁棒性和不可感知性。该技术的平均PSNR和SSIM分别为54.24 dB和0.9983。该方法的PSNR增量大于2 dB，优于现有的水印技术。此外，与现有方法相比，该技术在10% - 50%的篡改率下，可获得高达5 dB的PSNR恢复增量。

{"title":"A novel pixel pair shuffling based image watermarking for tamper detection and self-recovery","authors":"Radha Ramesh Murapaka , A.V.S. Pavan Kumar , Aditya Kumar Sahu","doi":"10.1016/j.ibmed.2025.100324","DOIUrl":"10.1016/j.ibmed.2025.100324","url":null,"abstract":"<div><div>This work has introduced a novel image watermarking scheme leveraging a pixel pair-based shuffling (PPSh) technique for tamper detection and self-recovery. The proposed technique consists of five steps, initiating from secret bits generation, collectively known as watermark bits. Then, the next step is watermark embedding, after that, watermark extraction, tamper detection, and finally, dual self-recovery approaches have been implemented. For watermark bit generation, two prominent interpolation techniques, such as bipolar and bilinear, are applied to the cover image (CI) to obtain the compressed image. Later, Advanced Encryption Standard (AES) and Camellia with Cipher Block Chaining (CBC) mode of operation is utilized on the compressed image to generate watermark bits. Afterwards, a PPSh-based watermark embedding strategy has been utilized to achieve the watermarked image (WI) while maintaining a standard payload capacity. Further, a variety of image processing attacks is performed on the WI to check the imperceptibility and similarity of the proposed scheme. Consequently, tamper region detection is followed by the watermark extraction procedure. Therefore, to reconstruct the tampered pixels, inpainting based dual recovery approaches are presented, named as TELEA and Naiver-Stokes (NS). The robustness and imperceptibility of the proposed scheme is measured through peak-signal-to-noise ratio (PSNR), structural similarity index matrix (SSIM), and mean square error (MSE). The proposed technique has achieved an average PSNR and SSIM of 54.24 dB and 0.9983, respectively. With an increment of more than 2 dB in terms of PSNR the proposed technique outperforms the existing watermarking techniques. Additionally, the proposed technique obtains a recovery increment up to 5 dB in terms of PSNR for 10 %–50 % tampering rates against the existing methods.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"12 ","pages":"Article 100324"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145683785","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Advancing drug discovery and development through GPT models: a review on challenges, innovations and future prospects 通过GPT模型推进药物发现和开发：挑战、创新和未来前景综述

Intelligence-based medicine

Pub Date : 2025-01-01 DOI: 10.1016/j.ibmed.2025.100233

Zhinya Kawa Othman , Mohamed Mustaf Ahmed , Olalekan John Okesanya , Adamu Muhammad Ibrahim , Shuaibu Saidu Musa , Bryar A. Hassan , Lanja Ibrahim Saeed , Don Eliseo Lucero-Prisno III

Advanced AI algorithms, notably generative pre-trained transformer (GPT) models, are revolutionizing healthcare and drug discovery and development by efficiently processing and interpreting large volumes of medical data. Specialized models, such as ProtGPT2 and BioGPT, extend their capabilities to protein engineering and biomedical text mining. Our study will contribute to ongoing discussions to revolutionize drug development, leading to a faster and more reliable validation of new therapeutic agents that are crucial for healthcare advancement and patient outcomes. GPT models, such as MTMol-GPT, are robust, generalizable, and provide important information for developing treatments for complicated disorders. SynerGPT utilizes a genetic algorithm to optimize prompts and select drug combinations for testing based on individual patient characteristics. Ligand generation for specific target proteins with potential drug activity is a significant stage in the drug design process, which enhances the quality of the synthesized compounds and augments the precision of capturing chemical structures and their activity correlations, highlighting the model's creativity and capability for innovative ligand design. Despite these advancements, there are still problems with the data volume, scalability, interpretability, and validation. Ethical considerations, robust methods, and omics data must be successfully integrated to develop AI for drug discovery and ensure successful deployment. In summary, these models significantly influence drug research and development, specifically in the earlier stages from initial target selection to post-marketing surveillance for medication safety monitoring.

先进的人工智能算法，特别是生成预训练变压器（GPT）模型，通过有效处理和解释大量医疗数据，正在彻底改变医疗保健和药物发现和开发。专门的模型，如ProtGPT2和BioGPT，将其功能扩展到蛋白质工程和生物医学文本挖掘。我们的研究将有助于正在进行的药物开发革命的讨论，从而更快、更可靠地验证对医疗保健进步和患者预后至关重要的新治疗药物。GPT模型，如MTMol-GPT，是鲁棒的，可推广的，并为开发治疗复杂疾病的重要信息。synergy pt利用遗传算法来优化提示和选择药物组合，以根据个体患者的特征进行测试。为具有潜在药物活性的特定靶蛋白生成配体是药物设计过程中的一个重要阶段，它提高了合成化合物的质量，提高了捕获化学结构及其活性相关性的精度，突出了模型的创造性和创新配体设计的能力。尽管取得了这些进步，但在数据量、可伸缩性、可解释性和验证方面仍然存在问题。伦理考虑、稳健的方法和组学数据必须成功整合，以开发用于药物发现的人工智能，并确保成功部署。综上所述，这些模型显著影响了药物研发，特别是在药物安全监测的早期阶段，从最初的目标选择到上市后监测。

{"title":"Advancing drug discovery and development through GPT models: a review on challenges, innovations and future prospects","authors":"Zhinya Kawa Othman , Mohamed Mustaf Ahmed , Olalekan John Okesanya , Adamu Muhammad Ibrahim , Shuaibu Saidu Musa , Bryar A. Hassan , Lanja Ibrahim Saeed , Don Eliseo Lucero-Prisno III","doi":"10.1016/j.ibmed.2025.100233","DOIUrl":"10.1016/j.ibmed.2025.100233","url":null,"abstract":"<div><div>Advanced AI algorithms, notably generative pre-trained transformer (GPT) models, are revolutionizing healthcare and drug discovery and development by efficiently processing and interpreting large volumes of medical data. Specialized models, such as ProtGPT2 and BioGPT, extend their capabilities to protein engineering and biomedical text mining. Our study will contribute to ongoing discussions to revolutionize drug development, leading to a faster and more reliable validation of new therapeutic agents that are crucial for healthcare advancement and patient outcomes. GPT models, such as MTMol-GPT, are robust, generalizable, and provide important information for developing treatments for complicated disorders. SynerGPT utilizes a genetic algorithm to optimize prompts and select drug combinations for testing based on individual patient characteristics. Ligand generation for specific target proteins with potential drug activity is a significant stage in the drug design process, which enhances the quality of the synthesized compounds and augments the precision of capturing chemical structures and their activity correlations, highlighting the model's creativity and capability for innovative ligand design. Despite these advancements, there are still problems with the data volume, scalability, interpretability, and validation. Ethical considerations, robust methods, and omics data must be successfully integrated to develop AI for drug discovery and ensure successful deployment. In summary, these models significantly influence drug research and development, specifically in the earlier stages from initial target selection to post-marketing surveillance for medication safety monitoring.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"11 ","pages":"Article 100233"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143548273","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Testing the real-world utility of Bayes theorem in artificial intelligence-enabled electrocardiogram algorithm for the detection of left ventricular systolic dysfunction 测试贝叶斯定理在人工智能心电图算法检测左心室收缩功能障碍中的实际效用

Intelligence-based medicine

Pub Date : 2025-01-01 DOI: 10.1016/j.ibmed.2025.100238

Betsy J. Medina-Inojosa , David M. Harmon , Jose R. Medina-Inojosa , Rickey E. Carter , Itzhak Zachi Attia , Paul A. Friedman , Francisco Lopez-Jimenez

Objective

To assess how the theoretical principles of Bayes' theorem hold true in a clinically impactful way when testing the diagnostic performance of an artificial intelligence (AI) tool, using the case of the AI-enabled electrocardiogram (AI-ECG) screening tool that detects left ventricular systolic dysfunction (LVSD) in a “real-world” setting.

Patient and methods

We analyzed data from 42,883 consecutive patients who underwent a clinically indicated ECG and an echocardiogram within two weeks at our center between January 1st and December 31st^, 2019. We then evaluated area under the curve (AUC) of the receiver operating characteristics, sensitivity, specificity, positive and negative predictive values (PPV and NPV) of the AI-ECG to detect LVSD (left ventricle ejection fraction of ≤40 %) across (i) cumulative risk factor prevalence (pre-test probabilities) (ii) different diagnostic thresholds, using paired ECG-echocardiogram data.

Results

Prevalence of LVSD was 1.9 %, 4.0 %, 7.0 % and 13.9 % for patients with 0, 1–2, 3–4 and ≥5 risk-factors for LVSD. The AUC of the AI-ECG for each group was 0.955, 0.933, 0.901 and 0.886, respectively (p for trend<0.001). Pre-test probabilities hardly influenced sensitivity but did impact specificity. PPV was affected more than NPV, which was modestly altered. Thresholds impacted diagnostic performance parameters, although their effect on NPV at low pre-test probability was negligible.

Conclusion

In real world, pre-test probabilities/cumulative risk-factors of disease do affect specificity. Using different diagnostic thresholds yields the highest impact on algorithm performance. A Bayesian approach may enhance individualized diagnostic performance when implementing AI algorithms.

目的评估贝叶斯定理的理论原理在测试人工智能（AI）工具的诊断性能时如何以临床有效的方式成立，使用在“现实世界”环境中检测左心室收缩功能障碍（LVSD）的人工智能启用心电图（AI- ecg）筛查工具。患者和方法我们分析了2019年1月1日至12月31日在我们中心连续两周内接受临床指示心电图和超声心动图检查的42,883例患者的数据。然后，我们使用配对的心电图超声心动图数据，评估受试者工作特征的曲线下面积（AUC）、敏感性、特异性、阳性和阴性预测值（PPV和NPV），以检测LVSD（左心室射血分数≤40%）(i)累积危险因素患病率（测试前概率）（ii）不同的诊断阈值。结果伴有0、1-2、3-4、≥5种LVSD危险因素的患者LVSD患病率分别为1.9%、4.0%、7.0%、13.9%。各组AI-ECG AUC分别为0.955、0.933、0.901、0.886 （p为趋势值<；0.001）。预测试概率几乎不影响敏感性，但影响特异性。PPV比NPV受影响更大，NPV略有改变。阈值影响诊断性能参数，尽管它们在低测试前概率下对NPV的影响可以忽略不计。结论在现实世界中，检测前概率/疾病累积风险因素确实影响特异性。使用不同的诊断阈值对算法性能的影响最大。在实现人工智能算法时，贝叶斯方法可以提高个性化诊断性能。

{"title":"Testing the real-world utility of Bayes theorem in artificial intelligence-enabled electrocardiogram algorithm for the detection of left ventricular systolic dysfunction","authors":"Betsy J. Medina-Inojosa , David M. Harmon , Jose R. Medina-Inojosa , Rickey E. Carter , Itzhak Zachi Attia , Paul A. Friedman , Francisco Lopez-Jimenez","doi":"10.1016/j.ibmed.2025.100238","DOIUrl":"10.1016/j.ibmed.2025.100238","url":null,"abstract":"<div><h3>Objective</h3><div>To assess how the theoretical principles of Bayes' theorem hold true in a clinically impactful way when testing the diagnostic performance of an artificial intelligence (AI) tool, using the case of the AI-enabled electrocardiogram (AI-ECG) screening tool that detects left ventricular systolic dysfunction (LVSD) in a “real-world” setting.</div></div><div><h3>Patient and methods</h3><div>We analyzed data from 42,883 consecutive patients who underwent a clinically indicated ECG and an echocardiogram within two weeks at our center between January 1st and December 31st<sup>,</sup> 2019. We then evaluated area under the curve (AUC) of the receiver operating characteristics, sensitivity, specificity, positive and negative predictive values (PPV and NPV) of the AI-ECG to detect LVSD (left ventricle ejection fraction of ≤40 %) across (i) cumulative risk factor prevalence (pre-test probabilities) (ii) different diagnostic thresholds, using paired ECG-echocardiogram data.</div></div><div><h3>Results</h3><div>Prevalence of LVSD was 1.9 %, 4.0 %, 7.0 % and 13.9 % for patients with 0, 1–2, 3–4 and ≥5 risk-factors for LVSD. The AUC of the AI-ECG for each group was 0.955, 0.933, 0.901 and 0.886, respectively (p for trend<0.001). Pre-test probabilities hardly influenced sensitivity but did impact specificity. PPV was affected more than NPV, which was modestly altered. Thresholds impacted diagnostic performance parameters, although their effect on NPV at low pre-test probability was negligible.</div></div><div><h3>Conclusion</h3><div>In real world, pre-test probabilities/cumulative risk-factors of disease do affect specificity. Using different diagnostic thresholds yields the highest impact on algorithm performance. A Bayesian approach may enhance individualized diagnostic performance when implementing AI algorithms.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"11 ","pages":"Article 100238"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143825511","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Advancing breast cancer detection in ultrasound images using a novel hybrid ensemble deep learning model 使用新型混合集成深度学习模型推进超声图像中的乳腺癌检测

Intelligence-based medicine

Pub Date : 2025-01-01 DOI: 10.1016/j.ibmed.2025.100222

Radwan Qasrawi , Omar Daraghmeh , Suliman Thwib , Ibrahem Qdaih , Ghada Issa , Stephanny Vicuna Polo , Haneen Owienah , Diala Abu Al-Halawa , Siham Atari

Breast cancer remains a leading cause of mortality among women globally, emphasizing the critical need for prompt and accurate detection to improve patient outcomes. This study introduces an innovative hybrid model combining ultrasound image enhancement techniques with advanced machine learning for rapid and more accurate breast cancer prognosis. The proposed model integrates Contrast Limited Adaptive Histogram Equalization (CLAHE) for image quality improvement with an Ensemble Deep Random Vector Functional Link Neural Network (edRVFL) for classification. Utilizing a dataset of 4103 high-resolution ultrasound images from the Dunya Women's Cancer Center in Palestine, categorized into normal, benign, and malignant groups, the model was trained and evaluated using a 25-fold cross-validation approach. Results demonstrate higher performance of the hybrid model compared to traditional machine learning algorithms, achieving accuracies of 96 % for benign and 98 % for malignant cases after CLAHE enhancement. To further improve lesion detection and segmentation, a new method combining YOLOv5 object detection with the MedSAM foundation model was developed, achieving a Dice Similarity Coefficient of 0.988 after CLAHE enhancement. Validation in a clinical setting on 850 cases showed promising results, with 91.4 % ± 0.021 accuracy for benign and 84 % ± 0.024 for malignant predictions compared to histopathology. The model's high accuracy and interpretability, supported by Grad-CAM analysis, demonstrate its potential for integration into clinical practice. This study advances the application of machine learning in breast cancer detection from ultrasound images, presenting a valuable tool for enabling early detection and improving prognosis for breast cancer patients.

乳腺癌仍然是全球妇女死亡的主要原因，因此迫切需要及时和准确地检测，以改善患者的预后。本研究介绍了一种创新的混合模型，将超声图像增强技术与先进的机器学习相结合，用于快速和更准确的乳腺癌预后。该模型将对比度有限自适应直方图均衡化（CLAHE）与集成深度随机向量功能链接神经网络（edRVFL）相结合，用于图像质量改善。利用来自巴勒斯坦Dunya妇女癌症中心的4103张高分辨率超声图像数据集，将其分为正常、良性和恶性三组，该模型使用25倍交叉验证方法进行训练和评估。结果表明，与传统的机器学习算法相比，混合模型的性能更高，在CLAHE增强后，良性病例的准确率达到96%，恶性病例的准确率达到98%。为了进一步提高病灶的检测和分割，我们开发了一种将YOLOv5目标检测与MedSAM基础模型相结合的新方法，CLAHE增强后的Dice Similarity Coefficient达到了0.988。在850例临床验证中显示出令人鼓舞的结果，与组织病理学相比，良性预测准确率为91.4%±0.021，恶性预测准确率为84%±0.024。该模型的高准确性和可解释性，在Grad-CAM分析的支持下，证明了其整合到临床实践中的潜力。本研究推进了机器学习在乳腺癌超声图像检测中的应用，为乳腺癌患者早期发现和改善预后提供了有价值的工具。

{"title":"Advancing breast cancer detection in ultrasound images using a novel hybrid ensemble deep learning model","authors":"Radwan Qasrawi , Omar Daraghmeh , Suliman Thwib , Ibrahem Qdaih , Ghada Issa , Stephanny Vicuna Polo , Haneen Owienah , Diala Abu Al-Halawa , Siham Atari","doi":"10.1016/j.ibmed.2025.100222","DOIUrl":"10.1016/j.ibmed.2025.100222","url":null,"abstract":"<div><div>Breast cancer remains a leading cause of mortality among women globally, emphasizing the critical need for prompt and accurate detection to improve patient outcomes. This study introduces an innovative hybrid model combining ultrasound image enhancement techniques with advanced machine learning for rapid and more accurate breast cancer prognosis. The proposed model integrates Contrast Limited Adaptive Histogram Equalization (CLAHE) for image quality improvement with an Ensemble Deep Random Vector Functional Link Neural Network (edRVFL) for classification. Utilizing a dataset of 4103 high-resolution ultrasound images from the Dunya Women's Cancer Center in Palestine, categorized into normal, benign, and malignant groups, the model was trained and evaluated using a 25-fold cross-validation approach. Results demonstrate higher performance of the hybrid model compared to traditional machine learning algorithms, achieving accuracies of 96 % for benign and 98 % for malignant cases after CLAHE enhancement. To further improve lesion detection and segmentation, a new method combining YOLOv5 object detection with the MedSAM foundation model was developed, achieving a Dice Similarity Coefficient of 0.988 after CLAHE enhancement. Validation in a clinical setting on 850 cases showed promising results, with 91.4 % ± 0.021 accuracy for benign and 84 % ± 0.024 for malignant predictions compared to histopathology. The model's high accuracy and interpretability, supported by Grad-CAM analysis, demonstrate its potential for integration into clinical practice. This study advances the application of machine learning in breast cancer detection from ultrasound images, presenting a valuable tool for enabling early detection and improving prognosis for breast cancer patients.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"11 ","pages":"Article 100222"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143403018","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Can artificial intelligence help physicians using diaphragmatic ultrasound? 人工智能可以帮助医生使用膈超声吗？

Intelligence-based medicine

Pub Date : 2025-01-01 DOI: 10.1016/j.ibmed.2025.100202

Tianjie Zhang , Changchun Li , Dongwei Xu , Yan Liu , Qi Zhang , Ye Song

Purpose

We investigated the role of artificially intelligent architecture based on deep learning radiomics (DLR) in analyzing M-mode and B-mode ultrasound videos of the diaphragm for diaphragmatic ultrasound.

Methods

A total of 196 subjects underwent pulmonary function and ultrasonic examination of the diaphragm. All diaphragmatic ultrasound videos were collected by experienced sonographers as the entire dataset used in this study. The experiment was partitioned into two parts. First, the diaphragm images (including M-mode and B-mode) of 157 subjects were input into the artificial intelligence architecture by the AI team. Second, the test set comprised 39 subjects, each equipped with three mobility images and three thickness images. We applied the proposed parameter calculation method to this set. The method entails segmenting the images, extracting the diaphragmatic motion and thickness variation curves from the segmentation results, and subsequently analyzing these curves to acquire the target parameters. Concurrently, we documented the time taken for each measurement. In parallel, three medical professionals performed analogue measurements. We analysed the accuracy and consistency of the artificial intelligence measurements.

Results

The study included a total of 196 subjects. The optimal segmentation model achieved dice scores of 73.51 % and 80.76 % on the test sets of mobility images and thickness images, respectively. Our method yielded results similar to those obtained by senior sonographers and demonstrated a high level of consistency with all three medical professionals, particularly the senior sonographer, in the measurements of diaphragm excursion (DE), diaphragm contraction duration (DCD), and diaphragmatic thickness at the end of inspiration (DTei). Meanwhile, our proposed method exhibited the highest level of time efficiency. The average duration for measuring the mobility images was 1.49s and for thickness images was 0.68s, compared to critical care physicians (8.23s, 15.89s), junior sonographers (6.14s, 9.69s), and senior sonographers (4.48s,6 0.77s).

Conclusions

Our study suggests that artificial intelligence can assist physicians in obtaining accurate diaphragmatic ultrasound data and reducing interobserver variability. Additionally, it could also improve time efficiency in this process.

目的探讨基于深度学习放射组学（deep learning radiomics， DLR）的人工智能架构在横膈膜m模和b模超声视频分析中的作用。方法对196例患者行肺功能及横膈膜超声检查。所有膈超声视频均由经验丰富的超声医师收集，作为本研究中使用的整个数据集。实验分为两部分。首先，由人工智能团队将157名受试者的光圈图像（包括m模式和b模式）输入到人工智能架构中。第二，测试集由39个被试组成，每个被试配备3个移动图像和3个厚度图像。我们将提出的参数计算方法应用于该集合。该方法对图像进行分割，从分割结果中提取膈肌运动和厚度变化曲线，然后对这些曲线进行分析，获得目标参数。同时，我们记录了每次测量所花费的时间。同时，三名医疗专业人员进行了模拟测量。我们分析了人工智能测量的准确性和一致性。结果本研究共纳入196名受试者。在移动图像和厚度图像的测试集上，最优分割模型的骰子得分分别为73.51%和80.76%。我们的方法得到的结果与高级超声医师得到的结果相似，并且在膈偏移（DE）、膈收缩持续时间（DCD）和吸气结束时膈厚度（DTei）的测量中，与所有三位医学专业人员，特别是高级超声医师的结果高度一致。同时，我们提出的方法具有最高的时间效率。测量流动图像的平均时间为1.49s，厚度图像的平均时间为0.68s，而重症医师（8.23s, 15.89s）、初级超声医师（6.14s, 9.69s）和高级超声医师（4.48s, 6.0.77 s）的平均时间为1.49s。结论人工智能可以帮助医生获得准确的膈超声数据，减少观察者之间的差异。此外，它还可以提高这个过程中的时间效率。

{"title":"Can artificial intelligence help physicians using diaphragmatic ultrasound?","authors":"Tianjie Zhang , Changchun Li , Dongwei Xu , Yan Liu , Qi Zhang , Ye Song","doi":"10.1016/j.ibmed.2025.100202","DOIUrl":"10.1016/j.ibmed.2025.100202","url":null,"abstract":"<div><h3>Purpose</h3><div>We investigated the role of artificially intelligent architecture based on deep learning radiomics (DLR) in analyzing M-mode and B-mode ultrasound videos of the diaphragm for diaphragmatic ultrasound.</div></div><div><h3>Methods</h3><div>A total of 196 subjects underwent pulmonary function and ultrasonic examination of the diaphragm. All diaphragmatic ultrasound videos were collected by experienced sonographers as the entire dataset used in this study. The experiment was partitioned into two parts. First, the diaphragm images (including M-mode and B-mode) of 157 subjects were input into the artificial intelligence architecture by the AI team. Second, the test set comprised 39 subjects, each equipped with three mobility images and three thickness images. We applied the proposed parameter calculation method to this set. The method entails segmenting the images, extracting the diaphragmatic motion and thickness variation curves from the segmentation results, and subsequently analyzing these curves to acquire the target parameters. Concurrently, we documented the time taken for each measurement. In parallel, three medical professionals performed analogue measurements. We analysed the accuracy and consistency of the artificial intelligence measurements.</div></div><div><h3>Results</h3><div>The study included a total of 196 subjects. The optimal segmentation model achieved dice scores of 73.51 % and 80.76 % on the test sets of mobility images and thickness images, respectively. Our method yielded results similar to those obtained by senior sonographers and demonstrated a high level of consistency with all three medical professionals, particularly the senior sonographer, in the measurements of diaphragm excursion (DE), diaphragm contraction duration (DCD), and diaphragmatic thickness at the end of inspiration (DTei). Meanwhile, our proposed method exhibited the highest level of time efficiency. The average duration for measuring the mobility images was 1.49s and for thickness images was 0.68s, compared to critical care physicians (8.23s, 15.89s), junior sonographers (6.14s, 9.69s), and senior sonographers (4.48s,6 0.77s).</div></div><div><h3>Conclusions</h3><div>Our study suggests that artificial intelligence can assist physicians in obtaining accurate diaphragmatic ultrasound data and reducing interobserver variability. Additionally, it could also improve time efficiency in this process.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"11 ","pages":"Article 100202"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143419553","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Artificial intelligence and patient care: Perspectives of audiologists and speech-language pathologists 人工智能和病人护理：听力学家和语言病理学家的观点

Intelligence-based medicine

Pub Date : 2025-01-01 DOI: 10.1016/j.ibmed.2025.100214

Komal Aggarwal , Rohit Ravi , Krishna Yerraguntla

Background

Artificial Intelligence has been implemented across various fields, including healthcare, where it has significantly advanced patient care in recent years. The present study aimed to explore the perspectives of audiologists and speech-language pathologists (ASLPs) toward AI in patient care.

Methods

The study employed a cross-sectional design with a convenience sampling method. The questionnaire included 27 questions consisting of demographic details and perspectives towards AI in audiology and speech language pathology services. Descriptive statistics were performed to analyze the data.

Results

Ninety-five ASLPs participated in the study, working across different work settings and with a mean age of 28.34 years, ranging between 18 and 47 years. Almost 50 % of participants reported AI tools can be helpful in diagnosis and planning the treatment. About One-fourth (25 %) believed that AI could help in rehabilitation. Few of participants (14.8 %) reported that AI may replace audiology and speech-language pathology services. ChatGPT was the most used platform by ASLPs in their practice. The ASLP clinicians believed AI would revolutionise ASLP practice without alarming effects on their employability.

Conclusion

The findings suggest that while AI has potential in ASLP practice, there is still a need for greater understanding and adoption of the technology.

人工智能已经在包括医疗保健在内的各个领域得到了应用，近年来，人工智能在医疗保健领域显著提高了患者护理水平。本研究旨在探讨听力学家和语言病理学家（aslp）对人工智能在患者护理中的观点。方法采用方便抽样的横断面设计。问卷包括27个问题，包括人口统计细节和对听力学和言语语言病理学服务中人工智能的看法。采用描述性统计方法对数据进行分析。结果95名aslp参与了研究，他们在不同的工作环境中工作，平均年龄28.34岁，年龄在18岁至47岁之间。近50%的参与者表示，人工智能工具可以帮助诊断和规划治疗。大约四分之一（25%）的人认为人工智能可以帮助康复。少数参与者（14.8%）报告说人工智能可能取代听力学和语言病理学服务。ChatGPT是aslp在实践中使用最多的平台。ASLP临床医生认为，人工智能将彻底改变ASLP实践，而不会对他们的就业能力产生惊人的影响。结论虽然人工智能在ASLP实践中具有潜力，但仍需要更多地了解和采用该技术。

{"title":"Artificial intelligence and patient care: Perspectives of audiologists and speech-language pathologists","authors":"Komal Aggarwal , Rohit Ravi , Krishna Yerraguntla","doi":"10.1016/j.ibmed.2025.100214","DOIUrl":"10.1016/j.ibmed.2025.100214","url":null,"abstract":"<div><h3>Background</h3><div>Artificial Intelligence has been implemented across various fields, including healthcare, where it has significantly advanced patient care in recent years. The present study aimed to explore the perspectives of audiologists and speech-language pathologists (ASLPs) toward AI in patient care.</div></div><div><h3>Methods</h3><div>The study employed a cross-sectional design with a convenience sampling method. The questionnaire included 27 questions consisting of demographic details and perspectives towards AI in audiology and speech language pathology services. Descriptive statistics were performed to analyze the data.</div></div><div><h3>Results</h3><div>Ninety-five ASLPs participated in the study, working across different work settings and with a mean age of 28.34 years, ranging between 18 and 47 years. Almost 50 % of participants reported AI tools can be helpful in diagnosis and planning the treatment. About One-fourth (25 %) believed that AI could help in rehabilitation. Few of participants (14.8 %) reported that AI may replace audiology and speech-language pathology services. ChatGPT was the most used platform by ASLPs in their practice. The ASLP clinicians believed AI would revolutionise ASLP practice without alarming effects on their employability.</div></div><div><h3>Conclusion</h3><div>The findings suggest that while AI has potential in ASLP practice, there is still a need for greater understanding and adoption of the technology.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"11 ","pages":"Article 100214"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143173632","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

V-NET-VGG16: Hybrid deep learning architecture for optimal segmentation and classification of multi-differentiated liver tumors V-NET-VGG16：用于多分化肝肿瘤最优分割和分类的混合深度学习架构

Intelligence-based medicine

Pub Date : 2025-01-01 DOI: 10.1016/j.ibmed.2025.100210

Amine Ben Slama , Hanene Sahli , Yessine Amri , Salam Labidi

Liver cancer is a leading cause of cancer-related mortality worldwide, underscoring the importance of early and accurate diagnosis. This study aims to develop an automatic system for liver tumor detection and classification using Computed Tomography (CT) images, addressing the critical challenge of accurately segmenting liver tumors and classifying them as benign, malignant, or normal tissues.

The proposed method combines two advanced deep learning models: V-Net for tumor segmentation and VGG16 for classification. A liver CT dataset augmented with various transformations, was used to enhance the model's robustness. The data was split into training (70 %) and testing (30 %) sets. The V-Net model performs the segmentation, isolating the liver and tumor regions from the CT images, while VGG16 is used for the classification of tumor types based on the segmented data.

The results demonstrate the effectiveness of this hybrid approach. The V-Net model achieved a Dice score of 97.34 % for accurate tumor segmentation, while the VGG16 model attained a classification accuracy of 96.52 % in differentiating between benign, malignant, and normal cases. These results surpass several existing state-of-the-art approaches in liver tumors analysis, demonstrating the potential of the proposed method for reliable and efficient medical image processing.

In conclusion, the hybrid V-Net and VGG16 architecture offers a powerful tool for the segmentation and classification of liver tumors, providing a significant improvement over manual segmentation methods that are prone to human error. This approach could aid clinicians in early diagnosis and treatment planning. Future work will focus on expanding the dataset and applying the method to other types of cancer to assess the model's generalizability and effectiveness in broader clinical settings.

肝癌是世界范围内癌症相关死亡的主要原因，强调了早期和准确诊断的重要性。本研究旨在开发一种基于计算机断层扫描（CT）图像的肝脏肿瘤自动检测和分类系统，以解决准确分割肝脏肿瘤并将其分类为良性、恶性或正常组织的关键挑战。该方法结合了两种先进的深度学习模型：用于肿瘤分割的V-Net和用于分类的VGG16。使用肝脏CT数据集增强各种转换，以增强模型的鲁棒性。数据分为训练集（70%）和测试集（30%）。V-Net模型进行分割，从CT图像中分离出肝脏和肿瘤区域，VGG16基于分割后的数据对肿瘤类型进行分类。结果证明了这种混合方法的有效性。V-Net模型在肿瘤准确分割方面的Dice评分为97.34%，而VGG16模型在良、恶性和正常病例区分方面的分类准确率为96.52%。这些结果超过了现有的几种最先进的肝脏肿瘤分析方法，证明了所提出的方法在可靠和有效的医学图像处理方面的潜力。综上所述，混合V-Net和VGG16架构为肝脏肿瘤的分割和分类提供了一个强大的工具，显著改善了容易出现人为错误的人工分割方法。这种方法可以帮助临床医生进行早期诊断和治疗计划。未来的工作将集中于扩展数据集，并将该方法应用于其他类型的癌症，以评估该模型在更广泛的临床环境中的普遍性和有效性。

{"title":"V-NET-VGG16: Hybrid deep learning architecture for optimal segmentation and classification of multi-differentiated liver tumors","authors":"Amine Ben Slama , Hanene Sahli , Yessine Amri , Salam Labidi","doi":"10.1016/j.ibmed.2025.100210","DOIUrl":"10.1016/j.ibmed.2025.100210","url":null,"abstract":"<div><div>Liver cancer is a leading cause of cancer-related mortality worldwide, underscoring the importance of early and accurate diagnosis. This study aims to develop an automatic system for liver tumor detection and classification using Computed Tomography (CT) images, addressing the critical challenge of accurately segmenting liver tumors and classifying them as benign, malignant, or normal tissues.</div><div>The proposed method combines two advanced deep learning models: V-Net for tumor segmentation and VGG16 for classification. A liver CT dataset augmented with various transformations, was used to enhance the model's robustness. The data was split into training (70 %) and testing (30 %) sets. The V-Net model performs the segmentation, isolating the liver and tumor regions from the CT images, while VGG16 is used for the classification of tumor types based on the segmented data.</div><div>The results demonstrate the effectiveness of this hybrid approach. The V-Net model achieved a Dice score of 97.34 % for accurate tumor segmentation, while the VGG16 model attained a classification accuracy of 96.52 % in differentiating between benign, malignant, and normal cases. These results surpass several existing state-of-the-art approaches in liver tumors analysis, demonstrating the potential of the proposed method for reliable and efficient medical image processing.</div><div>In conclusion, the hybrid V-Net and VGG16 architecture offers a powerful tool for the segmentation and classification of liver tumors, providing a significant improvement over manual segmentation methods that are prone to human error. This approach could aid clinicians in early diagnosis and treatment planning. Future work will focus on expanding the dataset and applying the method to other types of cancer to assess the model's generalizability and effectiveness in broader clinical settings.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"11 ","pages":"Article 100210"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143174327","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0