Density based Multiclass Support Vector Machine using IoT driven Service Oriented Architecture for Predicting Cervical Cancer

다국어 초록 (Multilingual Abstract)

Cervical Cancer stands out among those deadliest diseases, which threatens women in an alarming rate causing approximately 2, 66,000 mortalities per annum worldwide. This cancer can be diagnosed early enough through Pap smear test; a cervical cancer screening program. Finding out the true positive rates of the Cervical Cancer cells with precision is more complex when identifying the same categories of the cancer disease. Various researchers have proposed many approaches over the past four decades and the solutions are pertinent to cervical cancer; however, the challenge remains partially unresolved. The significant contribution of this paper is in two folds, firstly discuss a cloud ready Adapter Driven Service Oriented Architecture (RESPRO 3.0), developed by us for automated screening of Pap Smear can be extended to any International Classification of Diseases (ICD). Secondly, present an Internet of Things (IoT) driven Cervical Cancer prediction adapter built for RESPRO 3.0 based on Density based Multiclass Support Vector Machines (MCSVM) in combination with Polynomial Kernel Trick. The density parameters provide unique space in identifying cervical cancer cell categories compared to exising researches. This cloud solution’s results are bench marked and verified against cyto technician’s ground truth results, found to be highly satisfactory with respect to 93% Sensitivity and 99% Specificity while minimizing test repeatability ratio for the supervised training set of images.

목차 (Table of Contents)

• Abstract
• 1. Introduction
• 2. Related Work
• 3.2. Bethesda System
• 4. IoT Driven Clinical Analysis Architecture (IDCAA)
• 4.1. Service Oriented Architecture
• 4.2. Architecture
• 4.3. Automation Stages and Ranking Features
• 5. Support Vector Machines (SVMs)
• 5.1. Binary Classification Using SVM
• 5.2. Multiclass Support Vector Machine (MSVM)
• 5.3. Proposed Density based Multi Class Support Vector Machine(DMCSVM)
• 6. Experimental Classification Results and Analysis
• 6.1. Image Pre Processing and Preparation for Training Set
• 6.2. Pap Smear Image Processing through Multi Class Support Vector Machine
• 6.3. Performance Assessment
• 7. Conclusion and Suggestions
• References