Photon-counting Detector CT with Deep Learning Noise Reduction to Detect Multiple Myeloma
Francis I Baffour1, Nathan R Huber1, Andrea Ferrero1
1From the Department of Radiology (F.I.B., N.R.H., A.F., K.R., K.N.G., S.L., C.H.M., J.G.F.), Division of Biomedical Statistics and Informatics, Department of Quantitative Health Sciences (N.B.L.), and Division of Hematology, Department of Medicine (S.K., J.M.C.), Mayo Clinic, 200 First St SW, Rochester, MN 55905; and Siemens Medical Solutions USA, Malvern, Pa (E.R.S.).
Photon-counting detector (PCD) CT with deep learning offers improved spatial resolution for detecting multiple myeloma lesions compared to conventional energy-integrating detector (EID) CT. This advancement enhances visualization of bone lesions and other abnormalities at low radiation doses.
Area of Science:
- Radiology and Imaging Science
- Medical Physics
- Oncology Imaging
Background:
- Photon-counting detector (PCD) CT and deep learning noise reduction show potential for enhanced spatial resolution at reduced radiation doses compared to energy-integrating detector (EID) CT.
- Improved imaging techniques are crucial for early and accurate detection of multiple myeloma, a condition often assessed using whole-body low-dose CT scans.
Purpose of the Study:
- To evaluate the diagnostic impact of enhanced spatial resolution using PCD CT with deep learning denoising for whole-body low-dose CT in visualizing multiple myeloma.
- To compare the performance of PCD CT against conventional EID CT in detecting various pathological findings associated with multiple myeloma.
Main Methods:
- Prospective enrollment of adult participants undergoing whole-body EID CT, followed by scanning with a PCD CT system in ultra-high-resolution mode at a matched radiation dose (8 mSv).
- Image reconstruction involved Br44 and Br64 kernels at 2-mm thickness for both EID and PCD CT, and Br44 and Br76 kernels at 0.6-mm thickness for PCD CT, with deep learning denoising applied to thinner slices.
- Objective image quality assessment in phantoms and subjective assessment by two blinded radiologists using a five-point Likert scale to score the detection of multiple myeloma findings.
Main Results:
- Blinded assessment of 2-mm images revealed significant improvements in visualizing lytic lesions, intramedullary lesions, fatty metamorphosis, and pathologic fractures with PCD CT compared to EID CT (P < .05).
- The 0.6-mm PCD CT images, enhanced with convolutional neural network denoising, further improved the detection of all four pathological abnormalities.
- PCD CT detected one or more lytic lesions in 21 of 27 participants, a significant improvement over 2-mm EID CT (P < .001).
Conclusions:
- Ultra-high-resolution PCD CT significantly enhances the visibility of multiple myeloma lesions compared to conventional EID CT.
- The combination of PCD CT technology and deep learning denoising offers a promising advancement for low-dose, whole-body CT imaging in multiple myeloma detection and characterization.
