Skip to main content
Data-Centric Engineering
  1. Data-Centric Engineering
  2. Research
  3. Research Projects on Offer
  4. Medicine & Dentistry

Medicine & Dentistry

Below you will find Data-Centric Engineering projects offered by supervisors within the School of Medicine & Dentistry

Please note that these are examples and the list will be confirmed in January 2022.

This is not an exhaustive list. If you have your own research idea, or if you are a prospective PDS candidate, please return to the main DCE Research page for further guidance, or contact us at dce-cdt@qmul.ac.uk 

Application of novel computer vision machine learning to diagnose thyroid cancers

This proposed project aims to use thyroid histopathology specimens to develop state-of-the art artificial intelligence algorithms for computer-aided diagnoses. Currently, following a scan, thyroid nodules suspected by the personal judgement of a radiologist to be malignant are further examined through a technique known as fine-needle aspiration (FNA). This can lead to variabillities between radiologists. FNA is an invasive and costly test with risks for the patient. Obtaining an accurate diagnosis based on the histopathology sample can avoid recalling the patient for this invasive procedure or performing unnecessary surgery, if no solid diagnosis was reached.

This approach will be used to determine whether the thyroid nodule sample is malignant or benign. Even though many cases end up being benign, some are diagnosed as malignant or “intermediate”, i.e. unclear diagnosis after performing FNA. Patients with malignant and intermediate diagnoses will undergo diagnostic surgery, but only around 30% are found to be malignant post-operation. This imperfect system leads to a higher burden of unnecessary thyroid scans, FNAs, and surgical operations in the NHS with financial implications extending to post-surgery care, hospital expenses and anaesthesia services.

We will develop automatic interpretation of thyroid biopsy specimen slides for automated diagnosis using machine and deep learning approaches. Computer programs can then be devised for feature extraction from image data and feeding of this information to classifiers for automated labelling of thyroid nodules as benign or malignant. Such approaches, can prove to be more objective, fast and accurate. This open source classification algorithm will be of use across the NHS helping histopathologists deliver diagnosis with accuracy, and supporting trainee healthcare professionals. 

Supervisor: Dr Eirini Marouli, William Harvey Research Institute