The Radiation Oncology Feed For radiation oncologists · physicists · RTTs · dosimetrists
technical-relevance lung treatment-planning

CT–MRI fusion improved contrast-free lung perfusion mapping for radiotherapy planning

Multimodal CT–MRI perfusion maps matched SPECT better than single-modality methods and reduced high-function lung dose in exploratory planning.

RadOncToday · ·1 min read

KEY POINTS

  • This prospective single-institution study included 21 patients with lung cancer who underwent pulmonary perfusion SPECT, planning CT, and free-breathing non-contrast proton MRI within one week. The primary analysis used seven-fold cross-validation with 18 training and three test cases per fold.
  • The MCF-Q framework combined a CT branch with an MRI branch using perfusion maps reconstructed from free-breathing MRI through the PREFUL method. Multilevel features were fused in a dual-branch three-dimensional U-Net, with SPECT perfusion serving as the reference.
  • MCF-Q achieved the strongest voxel-wise agreement with SPECT, with a mean Spearman correlation of 0.7831 ± 0.0821. Corresponding correlations were 0.7374 for CT alone, 0.7496 for refined MRI perfusion, and 0.5542 for conventional PREFUL reconstruction.
  • Spatial overlap with SPECT-defined functional regions was also highest with MCF-Q. Mean Dice scores were 0.8396 ± 0.0379 for high-function lung and 0.7680 ± 0.0555 for low-function lung, compared with 0.8288 and 0.7495 for CT-based mapping.
  • Patient-level leave-one-out validation produced similar results, with an average correlation of 0.7878, high-function Dice of 0.8431, and low-function Dice of 0.7731. This supported consistency across data splits, although all patients still came from the same institution.
  • In exploratory functional-avoidance planning, MCF-Q guidance reduced mean dose to high-function lung from 7.817 to 6.856 Gy and V20 from 13.820% to 11.464% compared with conventional planning. Whole-lung, heart, oesophageal, spinal cord, and target-coverage metrics remained broadly comparable.
  • Performance generally improved as the training cohort increased from three to 21 patients and was explored in two later independent cases. However, the study did not test clinical pneumonitis, pulmonary-function decline, reproducibility on external scanners, or whether MCF-Q-guided plans improve patient outcomes.

CLINICAL TAKEAWAY

Combining CT with non-contrast MRI-derived perfusion information may provide more reliable functional lung maps than either modality alone and could support functional-avoidance radiotherapy without additional tracers. The evidence remains technical and preliminary because the dataset was small, single-centre, and limited to dosimetric rather than clinical validation.

SOURCE

International Journal of Radiation Oncology, Biology, Physics