Analysis of quantum noise-reducing filters on chest X-ray images: A review. (1st March 2020)

• Record Type: Journal Article
• Authors: Chandra, Tej Bahadur; Verma, Kesari
• Is Part Of: Measurement. Volume 153(2020)
• Publication Date: 2020-03-01
• Subjects: Quantum noise -- Poisson noise -- De-noising -- Image filtering -- X-ray images -- Noise filters; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2019.107426
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: Effects of Quantum noise on CXR images and performance of CAD systems. Review and analyzed various benchmark filters for reducing noise in CXR images. Investigated the tradeoff between de-noising and texture preserving performance. Guided filter outperformed in terms of quantitative, and subjective measures. The statistical evaluation proved the significance of the obtained results. Abstract: Radiography is one of the important clinical adjuncts for preliminary disease investigation. The X-ray images are corrupted with inherent quantum noise affecting the performance of computer-aided diagnosis systems. This paper presents an extensive experimental review and impact of six benchmark filters for reducing noise and disease classification on chest X-ray images. The tradeoff between de-noising and texture preserving performance is investigated through classification performances using the state-of-the-art machine learning methods – Support Vector Machine and Artificial Neural Network. Moreover, the qualitative, subjective, and statistical evaluation is performed by using the image quality metrics, expert radiologist opinion, and statistical test, respectively. The experimental results confirm the significant improvement in classification performance using Guided filtered images. Furthermore, the results of qualitative measures and subjective analysis demonstrate that the guided filter and anisotropic diffusion filter both performed significantly better. Finally, a non-parametric statistical test is used to validate statistical significance of the obtained results.