NLST (Learn2Reg 2022-23), validation

Baseline CT (fixed) and follow-up CT (moving) of the lungs of one NLST participant, a validation pair of Learn2Reg 2022/23, with the challenge's public corrField keypoints.

10 test pairslungsame patientfixed CT, moving CTdelivered aligned

Look at the images

Read before comparing (2)

Truth. The landmark error is the board's TRE(KP), on the public corrField keypoints: automatic correspondences, not manual landmarks. The board's TRE(LM) uses 100 landmarks a case that are not public.
Overlap. The scans are those of Lung250M's pairs 114 to 123, which carry manual landmarks: the lung track ranks on Lung250M, and this dataset stays for the comparison with the challenge's board.

No method has a result on every pair of this dataset yet.

		Accuracy			Smoothness		Cost
#	Method	TRE mm	Change	rTRE	SDlogJ	Folds %	Time s	RAM GB	GPU GB	Images
	No registration	9.70 ±1.66		0.0175 ±0.0030	0.000	0.00

TRE: distance between matching landmarks. rTRE: the TRE over the length of the image's diagonal, to compare datasets of other sizes and units. SDlogJ: how irregular the deformation is, 0 is perfectly smooth. Folds: share of the image where the deformation folds over itself.

Scored the way the challenge scores

The same results with the NLST (Learn2Reg 2022-23) challenge's own measures, on its validation pairs: mean ± standard deviation.

Method	TRE	SDlogJ
No registration	9.703 ± 2.202	0.000 ± 0.000

What the challenge published

On the challenge's test set: other pairs than the ones above, and entries trained or tuned for the task. It shows the level of the challenge; these are not results of this benchmark. The challenge measures its landmark error on manual landmarks that are not public; the lines of this board above are on the public keypoints, which are automatic. The same ten validation scans carry manual landmarks in Lung250M (pairs 114 to 123): read that board against these rows.

Method	TRE	SDlogJ
No registration	10.40
LapIRNv2	1.44	0.042
corrfield	1.64	0.038
Birmingham	1.43	0.045
UKE	2.44	0.062
MONAI	2.04	0.039
UCLA	2.93	0.071

Source: Learn2Reg 2024: New Benchmark Datasets Driving Progress on New Challenges, MELBA 2025, Table 1 (NLST 2023).

Pair by pair

Measure

TRE mm	1	2	3	4	5	6	7	8	9	10
No registration	8.56	8.53	8.86	7.51	10.84	14.93	7.51	9.49	12.10	8.68

Click a score to look at that result: the fixed image, and over it the moving image after registration, with the annotations. Underlined scores have their image; for the others you see the pair without the result. A green cell gained against no registration on that pair, a red one lost.

About the data

The first test caseClick to look at every case.

Test pairs: The challenge's 10 official validation pairs are the test cases, with 10 development pairs.
Truth: landmarks by a program the challenge's public points
Source: learn2reg.grand-challenge.org
github.com/MDL-UzL/L2R
Licence: CC BY 4.0 (TCIA NLST)

10 pairs for tuning, never tested

0039_0000_0039_0001, 0256_0000_0256_0001, 0298_0000_0298_0001, 0226_0000_0226_0001, 0263_0000_0263_0001, 0202_0000_0202_0001, 0014_0000_0014_0001, 0264_0000_0264_0001, 0214_0000_0214_0001, 0097_0000_0097_0001