Cloud-Enabled Preprocessing and Feature Extraction Framework for Hydrocephalus Detection from MRI Scans

Hydrocephalus is a brain disorder characterized by
an abnormal accumulation of the cerebrospinal fluid (CSF) in
the fluid-filled spaces of the brain (ventricles), which may cause increased pressure on the brain, mental impediment, and lifethreatening complications if not properly treated. Traditional diagnostic practices are based on manual interpretation of MRI scans, which is both time consuming, subjective and resource consuming - particularly inaccessible in rural or under-resourced regions of the world. Conventional diagnostic workflows lack the elements of real-time and are standardized, preventing early diagnosis and intervention. There is a crucial need for a cloud-integrated and standardized preprocessing
system for automating the feature extraction process for
classification following this step. In this work the technical principles of Phase 1 of a more comprehensive cloud-based framework for hydrocephalus detection are posted and the secure data acquisition and preprocessing and feature extraction from MRI scans are addressed. A dataset containing 1,000 MRI images (500 hydrocephalus and 500 healthy) will be curated from images with different demography. Preprocessing Malaysia included image cropping, normalized audio intensity and its segmentation to pick apart ventricular regions. Quantitative features including ventricle volume, shape morphology and the distribution pattern of CSF will be extracted using advanced medical imaging libraries. The preprocessed data is stored and managed by a scalable cloud infrastructure, so that it is available in real time and protects the privacy of the data in order to provide access to machine learning modules. The processed dataset ensures structural consistency and quality of data, which is a good input for classification models. Integration with cloud storage makes it more accessible and it aids in the development of diagnostic tools in a fast manner. This preprocessing step is essential to create the platform to efficiently perform hydrocephalus detection using a large amount of physiological parameters in real-time and with a low latency.