Work in Progress seminar by Mohammad Ahmed, discussing Deep Learning for Cancer Recognition from Ultrasound Images
8 May 2019
Mohammed began his discussion with us on 3rd May and began by introducing his project, ‘Towards Deep Learning for Cancer Recognition from Ultrasound Images’. In the introduction, he explains what defines as cancer, showing examples of cancerous cells and their types while also indicating the global impact through percentages. Following the examples shown, Mohammed relays the prior information to statistic information with 55,000 of women diagnosed with breast cancer globally each year.
Mohammed then uses medical imaging systems to detect the types of cancers he is researching on. Using a process involving ultrasound images > image processing > feature extraction and then classification on regions of interest. For machine learning, a similar process is involved, by using supervised learning a repeated process using input image > pre-processing stage > segmentation stage > feature extraction followed by extraction to determine the class of the selected area.
He then explains the value of deep learning, its role in machine learning and the influence convolutional neural network (CNN). By using the CNN and supervised learning to determine the classification of image input through feature extraction, this will create efficient support for the results. This helps the machine learn and understand its purpose for the ultrasound image data with similar areas of interest.
After explaining this process, he begins to show examples of recently published articles testing on their own CNN architecture which includes the classification accuracy, dataset specifications and types of cancers used. He plans to implement sections that they’ve used into his own network.
Mohammad understands that after he gains an improved understanding of CNN, there will be challenges such as lack of CNN-interpretability, model overfitting and Datasets (volume & quality), some that he currently faces.
Our work in progress speaker, Mohammed Ahmed, gave us a better understanding of his research and progress towards understanding and creating processes for deep learning of cancer recognised ultrasound images. This seminar also gave students and staff the opportunity to benefit from the discussion of ‘Towards Deep Learning for Cancer Recognition from Ultrasound Images’, possibly incorporating the new knowledge into their own research. For individuals who may be considering a course in computing or for those who may have a general interest, these discussions provide an insight into current students’ research and where it will lead them.
Find out more about our courses in computing.