CT brain image advancement for ICH diagnosis.

Nor Shahirah Shaik Amir,Law Zhe Kang,Shahizon Azura Mukari,Ramesh Sahathevan,Kalaivani Chellappan

doi:10.1049/htl.2018.5003

Abstract

A critical step in detection of primary intracerebral haemorrhage (ICH) is an accurate assessment of computed tomography (CT) brain images. The correct diagnosis relies on imaging modality and quality of acquired images. The authors present an enhancement algorithm which can improve the clarity of edges on CT images. About 40 samples of CT brain images with final diagnosis of primary ICH were obtained from the UKM Medical Centre in Digital Imaging and Communication in Medicine format. The images resized from 512 × 512 to 256 × 256 pixel resolution to reduce processing time. This Letter comprises of two main sections; the first is denoising using Wiener filter, non-local means and wavelet; the second section focuses on image enhancement using a modified unsharp masking (UM) algorithm to improve the visualisation of ICH. The combined approach of Wiener filter and modified UM algorithm outperforms other combinations with average values of mean square error, peak signal-to-noise ratio, variance and structural similarity index of 2.89, 31.72, 0.12 and 0.98, respectively. The reliability of proposed algorithm was evaluated by three blinded assessors which achieved a median score of 65%. This approach provides reliable validation for the proposed algorithm which has potential in improving image analysis.

Highlights

Intracerebral haemorrhage (ICH), an example of haemorrhagic stroke, is characterised by bleeding within the brain
Haemorrhage appears as a high-attenuation mass on Computed tomography (CT) scan but one of the limitations in ICH extraction is that the bleed may be poorly visualised on CT due to the presence of noise during the CT acquisition process [2]
We exclude any secondary ICH in this Letter consisting of trauma, subarachnoid haemorrhage, primary or secondary tumour as the mechanism, location and appearance of haemorrhage in these conditions is less uniform, and they may overlap the primary ICH (p-ICH) bleeds which may appear as microbleed on CT scan

Summary

Introduction

Intracerebral haemorrhage (ICH), an example of haemorrhagic stroke, is characterised by bleeding within the brain. Computed tomography (CT) is the main imaging modality used in diagnosing ICH during the initial evaluation of acute stroke [1]. The ability to detect haemorrhage is dependent on the quality of the CT brain images. Haemorrhage appears as a high-attenuation mass on CT scan but one of the limitations in ICH extraction is that the bleed may be poorly visualised on CT due to the presence of noise during the CT acquisition process [2]. To ensure that the performance of ICH detection will be robust, irrespective of the quality of input CT images, it is essential to incorporate a CT enhancement algorithm in the haemorrhage extraction module.

Methods

Results

Discussion

Conclusion