Intelligence-based medicine最新文献

{"title":"Development of an explainable machine learning model for predicting neurological deterioration in spontaneous intracerebral hemorrhage","authors":"Ming Jie, Jonathan Yeo ,&nbsp;Chun Peng Goh ,&nbsp;Christine Xia Wu ,&nbsp;Francis Phng ,&nbsp;Ping Yong ,&nbsp;Shiong Wen Low","doi":"10.1016/j.ibmed.2025.100237","DOIUrl":"10.1016/j.ibmed.2025.100237","url":null,"abstract":"<div><h3>Background</h3><div>Intracerebral hemorrhage (ICH) is a severe form of stroke associated with high morbidity and mortality. Early prediction of neurological deterioration (ND)—defined as a decline of at least 2 points on the Glasgow Coma Scale (GCS) within 48 h of admission or mortality at discharge—is essential for timely intervention and improved outcomes.</div></div><div><h3>Methods</h3><div>We developed an explainable machine learning model to predict ND using clinical, laboratory, and radiological data extracted from electronic medical records (EMR) of a retrospective cohort of 491 ICH patients, with ND observed in 52.3 % of cases. Multiple machine learning algorithms—including random forests, extra trees, and CatBoost—were trained, and model performance was evaluated using metrics such as the area under the receiver operating characteristic curve (AUC-ROC) and F1-score. Shapley Additive Explanations (SHAP) were employed to enhance interpretability.</div></div><div><h3>Results</h3><div>The final model, a blended ensemble, achieved an AUC-ROC of 0.8743, an F1-score of 0.8077, and a sensitivity of 0.8182 on the test set. Key predictors included initial GCS, hematoma volume, age, and the presence of intraventricular hemorrhage. SHAP analysis provided insights into the relative contributions of these predictors, reinforcing the model's clinical relevance.</div></div><div><h3>Conclusions</h3><div>Our model demonstrates promising predictive performance, suggesting its potential utility for early risk stratification and guiding interventions in ICH management. Further validation in diverse clinical settings is warranted.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"11 ","pages":"Article 100237"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143725919","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}

{"title":"“Evaluation of screening parameters and machine learning models for the prediction of neonatal sepsis: A systematic review.”","authors":"Peace Ezeobi Dennis ,&nbsp;Angella Musiimenta ,&nbsp;Wasswa William ,&nbsp;Stella Kyoyagala","doi":"10.1016/j.ibmed.2024.100195","DOIUrl":"10.1016/j.ibmed.2024.100195","url":null,"abstract":"<div><div>About 2.9 million neonates die every year worldwide, and most of these deaths occur in low resource settings where it causes about 30–50 % of the total neonatal deaths annually. Neonatal sepsis occurs when there is a bacterial invasion in the bloodstream; the immune system begins a systemic inflammatory response syndrome (SIRS) damaging to the body and can quickly advance to severe sepsis, multi-organ failure, and finally, death. Sepsis in neonates can progress more rapidly than in adults; therefore, timely diagnosis is critical. The gold standard test for diagnosing neonatal sepsis is blood culture, which takes at least 72 h. Hence, identifying key predictor variables and models that work best can help reduce neonatal morbidity and mortality.</div><div>Matching articles were identified by searching PubMed, IEEE, and Cochrane bibliography databases. Full-text articles with the following criteria were included for analysis based on 1) the subject population are neonates. 2) the study provided a clear definition of neonatal sepsis. 3) the study provides neonatal sepsis onset definition (i.e., time of onset). 4) the study clearly described the predictor variables used. 5) the study clearly described machine learning models used or evaluated any of the consolidated screening parameters. 6) the study must have provided diagnostic performance results. Thirty-one studies met full inclusion criteria. The duration of ROM was found to be more significant than other maternal risk factors. Heart rate and heart rate variability were found to be more significant than other neonatal clinical signs. C reactive protein and I/T ratio were found to be more significant than other laboratory tests.</div><div>A combination of predictor variables has shown to strengthen neonatal sepsis prediction, as shown by some of the reviewed studies. Predictive algorithms that combine multiple variables are urgently needed to improve models for early detection, prognosis, and treatment of neonatal sepsis.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"11 ","pages":"Article 100195"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143174756","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}

{"title":"Machine learning applications for predicting fracture of the adjacent vertebra after vertebroplasty","authors":"Maede Hasanpour ,&nbsp;Mohammadjavad (Matin) Einafshar ,&nbsp;Mohammad Haghpanahi ,&nbsp;Elie Massaad ,&nbsp;Ali Kiapour","doi":"10.1016/j.ibmed.2025.100205","DOIUrl":"10.1016/j.ibmed.2025.100205","url":null,"abstract":"<div><h3>Background</h3><div>Vertebroplasty, a minimally invasive procedure for treating vertebral compression fractures, has shown promising clinical outcomes due to its straightforward surgical technique, low complication rate, and rapid pain relief. However, a significant concern is the 25 % rate of subsequent vertebral fractures following treatment, with 50–67 % of these occurring in adjacent vertebrae that were previously augmented.</div></div><div><h3>Purpose</h3><div>To develop predictive models for fractures in vertebrae adjacent to those treated with vertebroplasty using machine learning techniques and a classification method based on pre-determined risk factors.</div></div><div><h3>Methods</h3><div>A retrospective study has been conducted to discover potential factors that could influence the effectiveness of vertebroplasty. Models were developed using data from 84 patients with osteoporotic vertebral compression fractures (OVCF) who underwent vertebroplasty. K-nearest neighbors (KNN), decision tree (DT), support vector machine (SVM), and logistic regression (LR) algorithms were used to predict fractures at the adjacent level of the augmented vertebra after vertebroplasty. The accuracies of the models were also reported.</div></div><div><h3>Results</h3><div>The DT and LR models achieved an accuracy of 0.94, while KNN and SVM models had an accuracy of 0.88. The DT identified bone mineral density (BMD), cement volume, and cement stiffness as key predictive factors. In contrast, the LR determined BMD, cement volume, and cement location to be the most essential features. Furthermore, the DT and LR models demonstrated the highest macro-average and weighted average metrics, calculated as 0.92 and 0.95, respectively.</div></div><div><h3>Conclusion</h3><div>The high accuracies achieved by the machine learning models confirm their effectiveness in predicting subsequent adjacent vertebral fractures (SAVF) following vertebroplasty. Utilizing these predictive models in clinical practice may enable the successful identification of patients at high risk for SAVF, potentially contributing to preventing these complications through personalized treatment planning and follow-up care.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"11 ","pages":"Article 100205"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143377346","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}

{"title":"Feasibility study: Detection of developmental dysplasia of the hip using ultrasound performed by a novice user","authors":"Fleur L.E. Kersten ,&nbsp;Chris L. de Korte ,&nbsp;Maartje M.R. Verhoeven ,&nbsp;Willemijn M. Klein ,&nbsp;Thomas L.A. van den Heuvel","doi":"10.1016/j.ibmed.2025.100228","DOIUrl":"10.1016/j.ibmed.2025.100228","url":null,"abstract":"<div><div>Developmental hip dysplasia (DDH) affects 1 %–4 % of infants globally, and ultrasound is the standard method to diagnose using Graf's method. However, ultrasound is not commonly used in primary care tool due to the required extensive training. Instead, physicians rely on physical examinations and risk stratification, resulting in a significant number of infants without DDH being referred. This study aimed to investigate if a novice user could be trained within 1 h to use an AI-assisted handheld ultrasound device to diagnose DDH.</div><div>The novice user conducted hip ultrasounds on 31 infants at the Radboud UMC. Trained radiologists performed ultrasounds on the same infants, serving as the ground truth. The ultrasound acquisitions by the novice user were evaluated by a pediatric radiologist to determine if they adhered to the standard plane of Graf. When deemed sufficient, the pediatric radiologist assessed if DDH could be excluded, and subsequently, it was compared to the ground truth diagnosis.</div><div>The ground truth identified 28 infants as no DDH, of which 23 (82 %) had AI-assisted ultrasounds in line with the standard plane of Graf, so DDH could be excluded. Additionally, all three infants with DDH (100 %) were correctly identified by the AI-assisted ultrasounds as ‘not excluding DDH’. This study demonstrates that it is possible for a novice user to acquire ultrasound images that satisfy the Graf criteria in 82 % of infants with 1 h of training. Such an approach could reduce the barrier of introducing ultrasound in the first-line of care and decrease the number of referred infants without DDH.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"11 ","pages":"Article 100228"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143479403","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}

{"title":"Optimizing hyperparameters for dual-attention network in lung segmentation","authors":"Rima Tri Wahyuningrum ,&nbsp;Rizki Abdil Fadillah ,&nbsp;Indah Yunita ,&nbsp;Budi Dwi Satoto ,&nbsp;Arif Muntasa ,&nbsp;Amillia Kartika Sari ,&nbsp;Paulus Rahardjo ,&nbsp;Deshinta Arrova Dewi ,&nbsp;Achmad Bauravindah","doi":"10.1016/j.ibmed.2025.100221","DOIUrl":"10.1016/j.ibmed.2025.100221","url":null,"abstract":"<div><div>Medical imaging, particularly chest X-rays (CXR), is a cornerstone in the diagnosis of lung diseases, such as pneumonia, tuberculosis and COVID-19, owing to its accessibility and effectiveness. However, the sheer volume of CXR images, especially during pandemics, combined with the complexity of subtle abnormalities, poses significant challenges for manual analysis. Lung segmentation plays a pivotal role in artificial intelligence-driven CXR analysis by isolating lung fields, which facilitates the detection of disease-affected regions. Recent advances in deep learning, particularly with attention mechanisms, have improved segmentation accuracy, but the performance of these models heavily depends on the selection of appropriate hyperparameters. This study investigates the impact of key hyperparameters—learning rate and number of epochs—on the performance of the dual-attention network (DANet) in lung segmentation tasks. DANet was tested on a CXR dataset from Qatar University and evaluated under four different hyperparameter configurations: 20 epochs with a learning rate of 0.001, 10 epochs with a learning rate of 0.001, 10 epochs with a learning rate of 0.0001 and 20 epochs with a learning rate of 0.0001. The model's performance was assessed using two widely recognised segmentation metrics: the Dice coefficient and Intersection over Union (IoU). The results indicated that higher learning rates and greater numbers of epochs lead to improved segmentation performance. Specifically, the DANet model achieved a Dice coefficient of 97.29 % and an IoU value of 94.74 %, demonstrating its effectiveness compared to other models. These findings highlight the importance of hyperparameter tuning in achieving high segmentation accuracy and demonstrate the potential of the DANet model to improve diagnostic workflows for CXR analysis.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"11 ","pages":"Article 100221"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143480059","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}

{"title":"Enhancing Alzheimer's disease detection: An explainable machine learning approach with ensemble techniques","authors":"Eram Mahamud ,&nbsp;Md Assaduzzaman ,&nbsp;Jahirul Islam ,&nbsp;Nafiz Fahad ,&nbsp;Md Jakir Hossen ,&nbsp;Thirumalaimuthu Thirumalaiappan Ramanathan","doi":"10.1016/j.ibmed.2025.100240","DOIUrl":"10.1016/j.ibmed.2025.100240","url":null,"abstract":"<div><div>Alzheimer's Disease (AD) is a progressive neurodegenerative disorder that necessitates early and accurate diagnosis for effective intervention. This study presents a novel machine learning (ML)-driven predictive framework for AD diagnosis, integrating Explainable Artificial Intelligence (XAI) methodologies to enhance interpretability. The dataset, sourced from Kaggle, comprises 2149 patient records with 34 distinct attributes, representing a comprehensive range of demographic, clinical, and lifestyle-related factors. To improve model robustness, rigorous data preprocessing techniques were employed, including mean/mode imputation for missing values, feature scaling using min-max normalization, and class balancing via SMOTE, SMOTEENN, and ADASYN. Feature selection technique was performed using Chi-Square and Recursive Feature Elimination (RFE) to retain the most relevant predictors. Various ML models—including Naïve Bayes, Decision Tree, Random Forest, Logistic Regression, AdaBoost, XGBoost, K-Nearest Neighbors (KNN), and Gradient Boosting—were assessed using accuracy, precision, recall, F1-score, and AUC (Area Under the Curve). The proposed ensemble model, combining LightGBM (LGBM) and Random Forest (RF) with Chi-Square feature selection and utilizing soft voting, achieved the highest test accuracy of 96.35 %, surpassing existing models. Additionally, SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) were utilized to interpret the model's decision-making process, identifying key risk factors and improving transparency for clinical applications. These findings highlight the potential of ML and XAI in advancing AD diagnosis, with future work aiming to validate the model on larger, more diverse datasets and integrate it into real-world clinical workflows.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"11 ","pages":"Article 100240"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143820707","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}

{"title":"Artificial intelligence in child development monitoring: A systematic review on usage, outcomes and acceptance","authors":"Lisa Reinhart, A. C. Bischops, Janna-Lina Kerth, Maurus Hagemeister, Bert Heinrichs, Simon Eickhoff, Juergen Dukart, Kerstin Konrad, Ertan Mayatepek, Thomas Meissner","doi":"10.1016/j.ibmed.2024.100134","DOIUrl":"https://doi.org/10.1016/j.ibmed.2024.100134","url":null,"abstract":"","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"40 3","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139817559","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}

{"title":"Rapid-Motion-Track: Markerless tracking of fast human motion with deep learning","authors":"Renjie Li ,&nbsp;Chun-yu Lau ,&nbsp;Rebecca J. St George ,&nbsp;Katherine Lawler ,&nbsp;Saurabh Garg ,&nbsp;Son N. Tran ,&nbsp;Quan Bai ,&nbsp;Jane Alty","doi":"10.1016/j.ibmed.2024.100162","DOIUrl":"10.1016/j.ibmed.2024.100162","url":null,"abstract":"<div><p>Human movement patterns reflect central nervous system function. Small deficits in repetitive fast movements, such as slightly slowed finger-tapping or mildly irregular rhythm of stepping, are often an early sign of a neurological disorder. Accessible tools that precisely measure the individual components of fast movements would thus enhance disease detection, monitoring and research. Deep learning-based computer vision methods applied to digital video-recordings hold promise but current state-of-the-art tools, including DeepLabCut (DLC) and other advanced models, fail to accurately track the fastest range of human movements, primarily due to image blur. To solve this, we developed a new end-to-end, Rapid-Motion-Track (RMT) computer vision tool. This study aimed to evaluate the accuracy of RMT compared to DLC and other advanced computer vision tools. 220 finger-tapping tests were performed at frequencies between 0.5Hz and 6Hz and recorded simultaneously with a standard 30 frames/sec 2D laptop camera and a high-speed 250 frames/sec 3D motion tracking system (ground-truth). Bland-Altman plots and paired Welch's <em>t</em>-test were used to quantify the validity of movement features extracted by computer vision methods with the ground-truth. The movement features extracted by RMT (e.g. frequency, speed, variance) exhibited high concurrent validity across all tapping-frequencies. RMT outperformed other computer vision methods for very fast movements &gt;4Hz. RMT also robustly tracked other fast motions including sit-to-stand, head-turning, foot-tapping, and leg agility. This new tool provides an accurate method to precisely and automatically measure even the fastest and finest human movements. It holds potential of wide reach as digital cameras are ubiquitous in homes, clinics and research centres.</p></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"10 ","pages":"Article 100162"},"PeriodicalIF":0.0,"publicationDate":"2024-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S2666521224000292/pdfft?md5=528a7c60c9ba2b2a41fea45266e09369&pid=1-s2.0-S2666521224000292-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141979664","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}

{"title":"Outcome prediction for adult mechanically ventilated patients using machine learning models and comparison with conventional statistical methods: A single-centre retrospective study","authors":"Wei Jun Dan Ong ,&nbsp;Chun Hung How ,&nbsp;Woon Hean Keenan Chong ,&nbsp;Faheem Ahmed Khan ,&nbsp;Kee Yuan Ngiam ,&nbsp;Amit Kansal","doi":"10.1016/j.ibmed.2024.100165","DOIUrl":"10.1016/j.ibmed.2024.100165","url":null,"abstract":"<div><p>In this retrospective single-centre study spanning five years (2016–2021) and involving 2368 adult Intensive Care Unit (ICU) patients requiring over 4 h of mechanical ventilation (MV) in a tertiary care hospital, we investigated the feasibility and accuracy of using machine learning (ML) models in predicting outcomes post-ICU discharge compared to conventional statistical methods (CSM). The study aimed to identify associated risk factors impacting these outcomes. Poor outcomes, defined as ICU readmission, mortality, and prolonged hospital stays, affected 40.2 % of the discharged MV patients. The Extreme Gradient Boost (XGBoost) ML model showed superior performance compared to CSM (Area under the receiver operating characteristic curve: 0.693 vs. 0.667; p-value = 0.03). At 95 % specificity, XGBoost displayed enhanced sensitivity (30.6 % vs. 23.8 %) compared to CSM. Risk factors such as Glasgow Coma Score (GCS) and GCS best motor score at ICU discharge, MV duration, ICU length of stay, and Charlson Comorbidity Index were identified. While both ML and CSM exhibited moderate accuracy, the study suggests ML algorithms have the potential for better predictive capabilities and individual risk factor identification, potentially aiding in the improvement of patient outcomes by identifying high-risk patients requiring closer monitoring. Further validation in larger studies is necessary, but the study underscores the potential for real-time application of ML algorithms developed from the increasing availability of electronic medical records (EMR).</p></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"10 ","pages":"Article 100165"},"PeriodicalIF":0.0,"publicationDate":"2024-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S2666521224000322/pdfft?md5=9b5e16fa3de6867cc99501f81f07c14a&pid=1-s2.0-S2666521224000322-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142011163","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}

{"title":"Oral cancer segmentation and identification system based on histopathological images using MaskMeanShiftCNN and SV-OnionNet","authors":"R. Dharani ,&nbsp;K. Danesh","doi":"10.1016/j.ibmed.2024.100185","DOIUrl":"10.1016/j.ibmed.2024.100185","url":null,"abstract":"<div><h3>Background</h3><div>Oral squamous cell carcinoma (OSCC) is the most common type of oral cancer and a significant threat to public health because of its high mortality rate. Early detection of OSCC is crucial for successful treatment and improved survival rates, but traditional diagnostic methods, such as biopsy, are time-consuming and require expert analysis. Deep learning algorithms have shown promise in detecting various cancers, including OSCC. However, accurately detecting OSCC on histopathological images remains challenging because of tumor heterogeneity.</div></div><div><h3>Methods</h3><div>This study proposes two new deep learning approaches, MaskMeanShiftCNN and SV-OnionNet, for segmenting and identifying OSCC. MaskMeanShiftCNN uses color, texture, and shape features to segment OSCC regions from input images, while SV-OnionNet is suitable for identifying OSCC at an early stage from histopathological images.</div></div><div><h3>Results</h3><div>The proposed approaches outperformed existing methods for OSCC detection, achieving a classification accuracy of 98.94 %, sensitivity of 98.96 %, specificity of 97.18 %, and error rate of 1.05 %. These results demonstrate the effectiveness of the proposed approaches in accurately detecting OSCC and potentially improving the efficiency of OSCC diagnosis.</div></div><div><h3>Conclusion</h3><div>The proposed deep learning approaches, MaskMeanShiftCNN and SV-OnionNet accurately detected OSCC in input and histopathological images. These approaches can improve the efficiency and accuracy of OSCC diagnosis, ultimately improving patient outcomes.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"10 ","pages":"Article 100185"},"PeriodicalIF":0.0,"publicationDate":"2024-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142554775","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}