Surgical oncology · 2026
MEMS-based near-infrared spectroscopy with AI for real-time breast cancer margin assessment.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Ibrahim Hossam Mostafa, Aref Mohamed Hisham Fouad, Elhadad Mohamed K, Hussein Abdallah Abdelkader, Sharawi Amr, Saleh Neven
Аннотация
BACKGROUND: Achieving clear surgical margins is essential for successful breast-conserving surgery and reducing re-excision rates. Conventional intraoperative approaches such as Frozen Section Analysis and Imprint Cytology are accurate but time-consuming and resource-dependent, underscoring the need for rapid, label-free, and objective diagnostic alternatives.
OBJECTIVE: This study presents a clinically oriented framework that integrates diffuse reflectance near-infrared (DR-NIR) spectroscopy with machine-learning (ML) and deep-learning (DL) algorithms for real-time breast-cancer diagnosis and intraoperative margin assessment.
METHODS: A total of 68 tissue specimens from 34 patients were analyzed using a MEMS-based DR-NIR sensor (NeoSpectra-Micro). Spectral data were processed through linear, ensemble, and convolutional models to classify malignant and normal tissues. A novel evaluation metric-the Metric for Classification Assessment and Success (MCAS)-was introduced to capture clinically weighted trade-offs between sensitivity and specificity. A web-based interface built in Gradio enables real-time diagnostic use by surgeons without programming expertise.
RESULTS: Several classifiers, including LinearSVC, VotingClassifier, MLPClassifier, and a convolutional neural network (CNN), achieved perfect discrimination (accuracy and AUROC = 1.000), confirming strong spectral separability between tissue classes. The MCAS metric provided additional insight into class-specific reliability beyond conventional performance indices.
CONCLUSION: The proposed AI-enhanced DR-NIR spectroscopy system delivers rapid, accurate, and interpretable intraoperative feedback on tumor margins. By combining portable MEMS sensing, data-driven modeling, and interactive clinical deployment, this study advances toward a practical, real-time optical alternative for reliable breast-cancer margin assessment.
OBJECTIVE: This study presents a clinically oriented framework that integrates diffuse reflectance near-infrared (DR-NIR) spectroscopy with machine-learning (ML) and deep-learning (DL) algorithms for real-time breast-cancer diagnosis and intraoperative margin assessment.
METHODS: A total of 68 tissue specimens from 34 patients were analyzed using a MEMS-based DR-NIR sensor (NeoSpectra-Micro). Spectral data were processed through linear, ensemble, and convolutional models to classify malignant and normal tissues. A novel evaluation metric-the Metric for Classification Assessment and Success (MCAS)-was introduced to capture clinically weighted trade-offs between sensitivity and specificity. A web-based interface built in Gradio enables real-time diagnostic use by surgeons without programming expertise.
RESULTS: Several classifiers, including LinearSVC, VotingClassifier, MLPClassifier, and a convolutional neural network (CNN), achieved perfect discrimination (accuracy and AUROC = 1.000), confirming strong spectral separability between tissue classes. The MCAS metric provided additional insight into class-specific reliability beyond conventional performance indices.
CONCLUSION: The proposed AI-enhanced DR-NIR spectroscopy system delivers rapid, accurate, and interpretable intraoperative feedback on tumor margins. By combining portable MEMS sensing, data-driven modeling, and interactive clinical deployment, this study advances toward a practical, real-time optical alternative for reliable breast-cancer margin assessment.