MRI radiomics in ovarian masses
Research type
Research Study
Full title
The role of radiomics and deep learning in identifying malignant ovarian masses on magnetic resonance imaging
IRAS ID
270809
Contact name
Evis Sala
Contact email
Sponsor organisation
Cambridge University Hospitals NHS Foundation Trust and The University of Cambridge
Duration of Study in the UK
2 years, 11 months, 28 days
Research summary
Ovarian cancer is the most malignant gynaecological disease in women which is usually disseminated to the other parts of the body at the time of diagnosis. In a doubt of a pelvic mass, doctors often take pictures of them in order to measure the size and identify features of the mass such as septa and solid parts.
Adnexa is the region in pelvis adjacent to the uterus that includes the ovary, fallopian tube, and associated structures. It is well known that ultrasound is the first line investigation for assessment of a possible adnexal mass (ovarian, tubal or uterine origin) but a significant proportion (approximately 20%) cannot be diagnosed on ultrasound. For these patients, MRI is used to further assess tissue characterization.
With great soft tissue resolution and no radiation, MRI is widely used as a problem solving modality in clinical units to identify the aetiology of adnexal lesions which cannot be determined by ultrasound or computed tomography. However, identifying malignancy in ovarian masses by using MRI depends on the experience of the radiologist. Currently, we do not have a way of computer-based diagnostic system for differential diagnosis of ovarian masses in routine clinical practice.
Radiomics is defined as a new computer based approach for extracting large sets of innumerable quantitative features from cross-sectional images, which cannot be determined by naked eye of the doctor. With sophisticated image processing methods, computers can create new image features, which can be named as radiomics signatures. Quantitative parameters derived from MRI radiomics could be used as imaging biomarkers to assess the presence of malignancy in an ovarian mass and predict an ovarian cancer subtypes.
Deep learning is a machine learning technique that teaches computers to learn by examples. In deep learning, a computer model learns to perform classification tasks directly from images, text or sound. In radiology, models are trained by using large set of labelled data. Computer model learns deep features about anatomy and morphology of organs, how do they look like (e.g. signal intensities of pelvic organs, muscles and bones in different MRI sequences), and features of pelvic masses. The combination of radiomics with deep learning approaches might further increase the ability to differentiate between benign and malignant adnexal and ovarian masses.
The purpose of this research is to evaluate the diagnostic performance of the MRI radiomics features and deep learning algorithms in discriminating benign ovarian tumours from malignancies in patients with adnexal masses.REC name
East of England - Cambridge Central Research Ethics Committee
REC reference
19/EE/0347
Date of REC Opinion
27 Feb 2020
REC opinion
Further Information Favourable Opinion