MICCAI 2015 Guess the Primary Challenge
2015 MICCAI Guess the Primary Challenge (CBTC Challenge 3, brain metastases MRI)
Brain MRI of metastases from lung cancer or melanoma, released for a challenge on telling the primary cancer from the imaging of the brain lesion. Contrast-enhanced T1, some with T2 and FLAIR, plus tumor masks for the training image sets.
Overview
The Guess the Primary Challenge was the third challenge of the Computational Brain Tumor Cluster of Events (CBTC), a satellite event of MICCAI 2015 in Munich. It asked whether the primary cancer behind a brain metastasis can be inferred from MRI alone: participants classified each image set as coming from a lung cancer or a melanoma primary. The data were contributed by Northwestern University and distributed through the Synapse project syn4112414, in the folder "Challenge Data" (syn4841968).
Composition
The challenge pages describe 87 brain MRI image sets from patients with brain metastases of lung cancer or melanoma. Each image set is one MRI study, and some are repeat scans of the same patient on a later date, so the number of patients is lower than 87 and is not stated. 57 image sets formed the training set, released on 12 August 2015 with a key giving the primary cancer, and 30 formed the test set, released on 15 September 2015. The split was pseudo-random with both primaries in both parts. The counts of lung and melanoma cases are not published on the open pages.
Acquisition
Every image set has a post-contrast (gadolinium) T1-weighted scan, and some also have T2-weighted or FLAIR scans. Files are gzip-compressed NIfTI, one folder per image set and sequence. Scanner, field strength and protocol are not described.
Annotations
Training image sets come with binary tumor masks drawn by trained raters and spot-checked by a US board-certified radiologist. The primary cancer label (lung or melanoma) is the classification target.
Known limitations
Repeat scans of one patient can sit in both splits, since the split was by image set. The organizers note that some masks may need a left-right flip to line up with the image in some software, and twelve image sets had corrupted files that were replaced in September 2015. Two teams submitted results, with 48% and 71% correct. Under the terms of use the data were for use within the challenge only and had to be destroyed by 31 December 2015 unless the data steward, Dr. Kristin Swanson, agreed to further use.
Cohort
Aggregate numbers from the sources below. Bars are relative to the largest value.
Split
Studies- Training 57 66%
- Test 30 34%
Step 2 (Data) in 2015 MICCAI Guess the Primary Challenge, How to Participate page
License and access
Our reading of the license, not legal advice. Before you use the data, read the original license and confirm that your use is allowed. We take no responsibility for how you use a dataset. Full disclaimer
Sign a data use agreement, often reviewed by the provider
2015 MICCAI Guess the Primary Challenge Terms of Use and Rules
Participants register on Synapse and return a signed copy of the Terms of Use. The data may be used only within the MICCAI 2015 challenge, any publication or other study needs permission from the data steward, and the data had to be destroyed by 31 December 2015 unless further use was authorized.
The signed Terms of Use PDF (syn4843138) can only be downloaded by logged-in Synapse users and was not read. The rule answers rest on the challenge Rules page.
What you can do
- No
- Not stated
- Conditional
- Create derived data Not stated
- Conditional
What you can share
- Not stated
- Share derived data Not stated
- Share trained models Not stated
What you must do
- Cite or credit Not stated
- Not stated
- Yes
- Not stated
- Conditional
- Not stated
- Yes
- Yes
Limits
- No re-identification Not stated
- Not stated
Commercial license: Not stated
Sources
Every number on this page comes from one of these documents. Each chart names the table or page it is taken from. The raw numbers are in stats.csv.