Condition category
Date applied
Date assigned
Last edited
Prospectively registered
Overall trial status
Recruitment status

Plain English Summary

Background and study aims
Magnetic resonance imaging (MRI) is a type of scan that produces detailed images of the inside of the body. Whole body magnetic resonance imaging (MRI) is being increasingly used to assess the amount of tumours in patients with metastatic cancer (cancer that has spread). However, whole body MRI has limitations in its diagnostic performance. Moreover, it is time-consuming to report a whole body scan, even for experienced readers. Machine learning is the process of training a computer to make decisions, based on already existing data. The aim of this study is to develop and test machine learning methods, applied to whole body MRI, to improve the diagnostic performance of whole body MRI and reduce the amount of time it takes for a radiologist to read a whole body MRI scan.

Who can participate?
This is a study limited to working with data from completed or currently recruiting studies. The study is divided into three Phases. In Phase 1 whole body MRI scans from healthy adult volunteers will be used. In Phases 2 and 3 data from child and adult patients with metastatic cancer will be used.

What does the study involve?
The MRI scans’ diagnostic performance and radiologists’ reading time will be compared with and without the assistance of the developed machine learning methods.

What are the possible benefits and risks of participating?
There are no perceived risks or direct benefits to the participants, as the application of the machine learning methods will take place after their clinical imaging has taken place. Future potential benefits include improvements in the diagnostic accuracy of whole body MRI scans, which could improve the initial cancer staging process, by improving the accurate detection of metastatic disease and reducing staging errors that could lead to unnecessary tests. Undergoing a single highly effective test for staging would also avoid the need for multiple patient visits to hospital.

Where is the study run from?
The study is run and led by the Imperial College Comprehensive Cancer Imaging Centre (C.C.I.C.) and Department of Computing. Participating centres (providing imaging data and radiological reading support) include the following: University College London (UCL) and UCL Hospital (UCLH), King’s College London and the Royal Marsden Hospital (RMH).

When is the study starting and how long is it expected to run for?
From February 2015 to July 2018.

Who is funding the study?
National Institute of Health Research (UK).

Who is the main contact?
Prof Andrea G. Rockall

Trial website

Contact information



Primary contact

Prof Andrea Rockall


Contact details

Department of Radiology
Imperial College NHS Healthcare Trust
Hammersmith Hospital Campus
W12 0NN
United Kingdom
+44 (0)786 658 5476

Additional identifiers

EudraCT number number

Protocol/serial number


Study information

Scientific title

Development and evaluation of machine learning methods in whole body MR with diffusion weighted imaging for staging of patients with cancer



Study hypothesis

The use of machine learning can improve the diagnostic performance of whole body MRI.

Ethics approval

IC REC for Phases 2 and 3: 15IC2647

Study design

Observational study (study limited to working with data)

Primary study design


Secondary study design

Study limited to working with data

Trial setting


Trial type


Patient information sheet




Research will be carried out at Imperial College London in collaboration with the teams of main studies (NIHR STREAMLINE (colon & lung cancer patients) and CRUK MELT (lymphoma patients)) who have recruited all the patients for diffusion weighted imaging (WB-DW-MR) datasets. The proposed study has three stages: firstly, WB-DW-MR from 50 healthy volunteers will be used to develop the machine learning (ML) method for automatic recognition of normal appearances. Secondly, the ML method will be tested on 150 WB-MR scans from the main studies, in whom the different sites of disease have already been confirmed. The ML method will be refined by radiologists who will identify the correct sites of disease and find ML errors. This will be used to improve the ML method. Thirdly, the refined ML method will be tested in a second group of 169 patients from the main studies to see if the technique can improve radiology reporting by improving diagnostic accuracy (DA) and reading time (RT).

Intervention type



Drug names

Primary outcome measures

Per site sensitivity and specificity of MRI for nodal and extra-nodal sites and concordance in final disease stage with the multi-modality reference standard (at staging). The reference standard for the MELT study is contemporaneous MDT with all other staging eg PET CT and CT at the time of diagnosis and initial staging,

Secondary outcome measures

1. Inter-observer agreement for MR radiologists
2. Evaluation of different MRI sequences on diagnostic accuracy
3. Simulated effect of MRI on clinical management

Overall trial start date


Overall trial end date


Reason abandoned


Participant inclusion criteria

As per each source (contributing) study (,,

Participant type


Age group




Target number of participants


Participant exclusion criteria

As per each source (contributing) study (,,

Recruitment start date


Recruitment end date



Countries of recruitment

United Kingdom

Trial participating centre

Imperial College, London - Hammersmith Campus
W12 0NN
United Kingdom

Sponsor information


Imperial College, London - Joint Research Compliance Office

Sponsor details

5th Floor Lab Block
Charing Cross Hospital
Fulham Palace Road
W6 8RF
United Kingdom
+44 (0)20 3311 0205

Sponsor type




Funder type


Funder name

National Institute for Health Research

Alternative name(s)


Funding Body Type

government organisation

Funding Body Subtype

Federal/National Government


United Kingdom

Results and Publications

Publication and dissemination plan

We plan to disseminate our study findings in the appropriate conventional, peer-reviewed conference presentations, including high impact radiology, computing sciences, MR physics and clinical conferences (in abstract form as scientific presentations at radiology conferences, as well as computing sciences conferences) and as full original scientific peer-reviewed publications in high impact factor scientific journals. The co-applicants have significant experience in presenting and publishing in their own fields.The study findings may have an important impact on the use of WB-DW-MRI in cancer and the subsequent patient management and therefore we intend to present this information to the appropriate forum at the Royal Colleges of Radiologists, International Cancer Imaging Society and MIUA (Medical Image Understanding and Analysis). We will present the findings at national cancer meetings, including the NCRI conference. We hope to have scientific abstracts accepted for publication at the MICCAI (Medical Image Computing and Computer Assisted Intervention) and the clinical imaging conferences ISMRM (International Society of Magnetic Resonance in Medicine) and RSNA (Radiological Society of North America). Dissemination of the project outcomes will be vital to the NHS and public organisations who may be able to benefit. We will use link with national cancer networks and disseminate findings through any relevant NHS clinical care groupings.

Intention to publish date

Participant level data

Stored in repository

Results - basic reporting

Publication summary

Publication citations

Additional files

Editorial Notes