The collection explores image similarity comparison of oesophageal cancer video frames to find cancer lesions to train the ML model .
The oesophageal cancer, which is detected by endoscopy when it has reached an advanced stage and treatment is less effective and patient prognosis is poor. Currently, the miss rate for early oesophageal cancer during endoscopy is 25% – the cancer lesions are very hard to spot and can be life threatening if overlooked. Detecting early cancer, on the other hand, offers a significantly higher chance of cure, as the tumour can be easily removed during an endoscopic examination.
We are speeding up the training of machine learning models using anonymised video footage from endoscopy. Our aim is to develop a real-time computer system that can run during the procedure (using an endoscope with inbuilt algorithm in it), highlighting areas of concern by overlaying markers to guide biopsy taking. Early detection is paramount because the cancer is curable as long as it is defined to the mucosa whilst the 5-year survival is less than 20% in more advanced stages.
This mammogram dataset consists of 3486 DICOM (Digital Imaging and Communications in Medicine) images. It contains normal, benign, and malignant cases with verified pathology information.
The Muscle Atlas created for the Institure of Myology, Sorbonne University, contains over a thousand of muscle biopsy images from people of different ages as well as animals.