Research frameworks for image analysis and to design and develop a machine learning model that would classify brain tumour MRI images.
Medical image processing is one of the most difficult and time-consuming tasks in the healthcare sector. Despite rigorous trials and tests, brain tumours remain one of the most lethal types of cancer. Glioblastoma affects one-third of adults and kills patients within two years of diagnosis. The project acknowledges the advancements in technology and utilised machine learning to develop a model which classifies brain tumour MRI images. The aim being to successfully analysis and identify tumours within MRI images and categorizing them into images with tumours and images without tumours. The chosen methodology for the project was waterfall with the development of the machine learning model being lean-agile. The project also included an experiment which provided insight into the computational performance of three different GPU’s and their impact on model execution times. The brain tumour model was developed using the Kaggle open-source dataset which include 3,000 MRI images. The model achieved 96.22% with a loss of 11.98%, however the model’s validation loss rate was higher than the training loss rate which signifies performance loss through finetuning both the model and dataset this could be improved.
This project provides an insight into medical image processing and incorporates advancements in technology such as machine learning to effectively classify medical MRI images. Reducing the strain on the healthcare sector while mitigating risks such as human error.