Particle swarm optimization based segmentation of Cancer in multi-parametric prostate MRI

• Published in: Multimedia tools and applications Vol. 80; no. 20; pp. 30557 - 30580
• Main Authors: Garg, Gaurav, Juneja, Mamta
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.08.2021; Springer Nature B.V
• Subjects: 1155T: Advanced machine learning algorithms for biomedical data and imaging; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Diagnosis; Diffusion coefficient; Image contrast; Image enhancement; Image segmentation; Inhomogeneity; Magnetic resonance imaging; Medical imaging; Methodology; Multimedia Information Systems; Particle swarm optimization; Prostate; Prostate cancer; Special Purpose and Application-Based Systems; Time dependence; Multi-parametric; Segmentation; MRI; Prostate; Optimization; Cancer
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-021-11133-2

Prostate Cancer (PCa) is one among the prominent causes of mortality in men, which can only be reduced by early diagnosis. Multi-parametric Magnetic Resonance Imaging (mp-MRI) is increasingly utilized by clinicians for performing diagnostics tasks because it possesses functional and morphological competencies. Although, manual segmentation of PCa on MRI is a tedious, operator-dependent and time consuming task. Therefore, Computer Aided Diagnosis (CAD) of PCa using mp-MRI images is highly desirable by employing computer-assisted segmentation approaches. In this paper, a method is proposed for segmentation of PCa based on level set with Particle Swarm Optimization (PSO) technique to address the limitations of existing techniques as PSO does not require any cost or objective function to be differentiable and it is easy to implement. The energy function is optimized with PSO based technique. The proposed approach is tested over three different mp-MRI modalities i.e., T2 weighted (T2w), Dynamic Contrast Enhanced (DCE) images and Apparent Diffusion Coefficient (ADC) Maps derived from Diffusion Weighted Images (DWI). The accuracy achieved by PSO based methodology is 7.6% greater than without PSO integration i.e., using Gradient descent with added computational overhead of 0.03 s. The experimental outcomes reveal that the proposed methodology shows better results in terms of considered evaluation metrics when compared with the existing techniques on the I2CVB dataset. The impact of the proposed methodology is that it has the ability for precise segmentation even with intensity inhomogeneity, which validates its applications in clinical treatments. Additionally, the proposed technique reduces the manual interference, which in turn minimizes the execution time.