سامانه همایش های دانشگاه آزاد اسلامی واحد نجف آباد

صفحه اصلی / اولین همایش بین المللی تحول خدمات سلامت: تلفیق مراقبت، درمان، بازتوانی و امیدبخشی

Genetic Algorithm-Driven Independent RNN for Accurate Segmentation and Classification of Lung Nodules in CT Imaging

نویسندگان :

Alireza Zarei¹ Ebrahim Daneshifar²

1- دانشگاه بین المللی امام رضا علیه السلام 2- دانشگاه بین المللی امام رضا علیه السلام

کلمات کلیدی :

Lung Cancer،Computed Tomography،IndRNN،Genetic Algorithm

چکیده :

ackground: Lung carcinoma remains a leading cause of cancer-related mortality globally, rendering early diagnosis and precise intervention imperative for improving patient survival rates [31]. Given the complexity of medical imaging, developing intelligent Computer-Aided Diagnosis (CAD) systems is essential to support specialists in accurate decision-making [32, 34]. This paper introduces a novel hybrid framework for pulmonary nodule detection in Computed Tomography (CT) images, integrating deep learning with evolutionary optimization. Method: The proposed methodology is structured into three distinct phases: pre-processing for noise reduction using advanced filtering techniques, core segmentation utilizing Independent Recurrent Neural Networks (IndRNN) [30], and post-processing optimization. The IndRNN architecture is specifically employed to mitigate the vanishing and exploding gradient problems inherent in traditional recurrent networks, thereby capturing long-term spatial dependencies in CT data effectively. Furthermore, a Genetic Algorithm (GA) is applied to fine-tune the network parameters and optimize the segmentation boundaries, ensuring high precision [35]. Finding: Experimental evaluations demonstrate that this genetic-optimized IndRNN model delivers robust performance, accurately delineating nodule regions and significantly enhancing diagnostic reliability compared to conventional methods. (Note: Achieving an accuracy of 97.61%, sensitivity of 98.00%, specificity of 97.27%, and a DSC of 0.9751). Conclusion: By effectively overcoming the limitations of traditional recurrent networks and leveraging evolutionary optimization, the proposed framework offers a highly precise and automated solution. This intelligent CAD system demonstrates significant potential to assist medical specialists in the early screening and accurate diagnosis of pulmonary nodules.