Deep Learning for Real-Time Data Analysis from Sensors
C. Sagar, Harshit Bhardwaj, Anupama Bhan · 2024
Changes in the cervix are the root cause of the common disease known as cervical cancer, which primarily affects women. Due to the potential for cancer to metastasize or spread to other organs, such as the bladder, liver, lungs, and rectum, early detection, screening, and preventive measures are crucial in improving the chances of a complete recovery. Pap smear tests and coloscopy are the two available ways to examine cells, with the latter being preferable because it is less expensive, less uncomfortable, and more readily available. This research provides a machine learning classification approach that leverages Support Vector Machines (SVMs) on the Herlev Pap smear image dataset. Good cell classification needs good cell image capture and segmentation. Using the dice index, expert cytologists manually annotated photos to determine how accurate the segmented images were. With a match rate of 92%, the outcomes demonstrated a high level of agreement. Active contour models with Gaussian fitting energy were used to achieve this segmentation. The pictures were then characterized utilizing Polynomial SVMs, which created a grouping exactness of 95%, the most elevated among the different SVM models tried. The exhibition highlights the measures taken for the identification framework of cervical disease. The suggested system utilizes Internet of Things (IoT) technology to demonstrate enhanced performance compared to traditional approaches for cervical cancer detection and segmentation. The dataset comprising images related to cervical cancer can be obtained and accessed through the image acquisition toolbox available in MATLAB. Furthermore, an Arduino microcontroller is utilized for the purpose of interfacing, and the gathered data is transmitted via a cloud-based web page utilizing an IoT module. Cervical cancer poses a rapidly escalating global health challenge, particularly impacting women in developing nations and resulting in elevated mortality rates. The early identification and categorization of cells in Pap smear samples hold immense significance in facilitating effective disease diagnosis and timely intervention. This study introduces a novel framework founded upon the Internet of Healthcare Things (IoHT) and employs deep learning techniques, particularly transfer learning, to accomplish the task of detecting and categorizing cervical cancer within Pap smear images. The innovative strategy aims to improve the accuracy and effectiveness of cervical cancer diagnosis by utilizing the transfer learning concept within the IoHT domain.