fidopt Interactive view
CT to PET, synthesis

RIDER Lung PET-CT, CT to FDG-PET

CT (source) and FDG-PET in body-weight SUV (target) of one lung cancer patient's PET/CT session on a GE Discovery STE, the CT brought onto the PET grid through their shared scanner frame (RIDER Lung PET-CT collection).

10 test caseswhole bodysource CT, target PET
Look at the images
Read before comparing (8)
  • Pairs. A case is a patient's first PET/CT study on the GE Discovery STE that holds a 'PET FDG SUV' series: that attenuation-corrected FDG PET, and the CT of its frame of reference reconstructed on an even 2.5 mm grid ('CT 2.5MM STD', or 'Recon 2: CTAC 2.5 THICK' on three 2005 studies), 120 or 140 kV, no contrast agent in its header. prepare.py checks the shared FrameOfReferenceUID and pins both series. Nothing is registered: the CT is resampled onto the PET grid of 5.47 by 5.47 by 3.27 mm with a cubic B-spline, air outside it, and rounded to whole Hounsfield units.
  • Field. The CT's field in the plane is 321 to 537 mm, the PET's 700: arms and the flanks of large patients lie outside the CT on most cases, and outside the mask. Of the voxels over 0.3 SUV, 7 to 40% fall outside the mask.
  • Target. Body-weight SUV from the PET's rescale values and its DICOM dose, weight and decay information (AutoPET_CTPET_Syn.suv_factor), 46 to 77 minutes after injection. The mask is the body in the shared field, lungs included (AutoPET_CTPET_Syn.body), except where the lungs meet the edge of the CT's field: it leaves out three quarters of the lungs on RIDER-1172912404 and RIDER-1298598021 and half on RIDER-2310941115.
  • Motion and uptake. Patients breathed freely during both scans; the lung bases and the heart disagree by an amount nobody measured. Every patient has lung cancer; tracer uptake is not set by the CT alone. Some PETs are noisy, with radial streaks outside the body.
  • Range. A dynamic range of 20 SUV puts PSNR and SSIM on the scale of the other FDG-PET synthesis datasets; it clips neither the target nor its errors.
  • Baseline. On the 10 test cases a constant at the target's median inside the mask scores, as the mean over the cases and its range, an MAE of 0.59 (0.52 to 0.72) SUV, a PSNR of 22.2 (16.8 to 24.9) dB and an SSIM of 0.62 (0.45 to 0.71).
  • Repeats. The collection's test-retest design scanned many patients twice; only the first study is used, and phantom studies and research reconstructions without SUV information are not.
  • Overlap. AorticTreeCTA holds CTs of 18 RIDER Lung PET-CT patients (19 names): they are outside the pool. TCIA holds four examinations twice under two names, matched by date, sex, age, weight, dose and injection time; that match finds AorticTreeCTA's known pair (R18, under RIDER-2112049538 and RIDER-2766484014). RIDER-2871069045, a second name of RIDER-2069446030's examination, and RIDER-2617411955, whose study holds RIDER-2277125393's examination, are held out. RIDER-1827278221 holds RIDER-1343746903's PET (same date, series time, weight, dose and injection time) and no CT: it is outside the pool. RIDER-2900155912 is held out: its PET in Bq/mL reads about 25 times under what its dose gives (a body median of 0.03 SUV). No other descriptor names this collection; RIDERLungCT is another collection.

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

Case by case

Click a score to look at that result: the target image, and over it the method's synthetic image. Underlined scores have their image; for the others you see the case without the result.

About the data

Test cases
10 test patients drawn with seed 2026, then 10 development patients, one session each, from one scanner. The pool is the 38 patients whose first PET/CT study holds a 'PET FDG SUV' series with the dose, injection time and weight SUV needs and a CT of its frame; draw.held_out lists two patients whose examination TCIA also holds under another patient's name and one whose PET reads 25 times under its dose.
Truth
target image as released the attenuation-corrected FDG-PET of the same PET/CT session in body-weight SUV, aligned by the scanner frame
Licence
CC BY 3.0
10 cases for tuning, never tested

RIDER-9640765801, RIDER-2310941115, RIDER-2883392803, RIDER-6525572907, RIDER-8054797922, RIDER-1887858289, RIDER-2852173628, RIDER-2388849043, RIDER-2020571155, RIDER-3089263741