Clinical feasibility of 4D single-source dual-energy CT for proton therapy of lung-cancer patients

Clinical feasibility of 4D single-source dual-energy CT for proton therapy of lung-cancer patients

Wohlfahrt, P.; Hofmann, C.; Troost, E. G. C.; Richter, C.; Jakobi, A.


Dual-energy CT (DECT) provides additional patient information to potentially improve delineation and range accuracy in proton therapy. Motion during sequentially acquired DECT scans might hamper its reliability. Here, we analysed the clinical feasibility of sequential 4D DECT scans (2 consecutive respiratory correlated 4D CT scans) for non-small cell lung cancer (NSCLC) patients and its applicability for proton dose calculation.


For 3 advanced stage NSCLC patients with maximal tumor motion of 1mm in cranio-caudal direction, 4D DECT scans were sequentially acquired during the course of treatment with a Siemens single-source DECT scanner. 80/140kVp average CT datasets and 4 breathing phases (relative amplitude sorting) were reconstructed and compared visually. These DECT datasets were further processed in syngo.via (Siemens Healthineers) to calculate 79keV pseudo-monoenergetic CT (MonoCTs) and stopping-power- ratio datasets (SPR, derived from electron density and photon cross section), Fig.1a. Passively scattered proton treatment plans were recalculated on MonoCT and 140kVp datasets using the clinical heuristic CT-number-to-SPR conversion (HLUT). Furthermore, worst-case scenarios using a single proton beam covering artificial target volumes encompassing the diaphragm (13-23mm motion) were generated. Dose distributions derived from MonoCT and 140kVp datasets were compared with 2D gamma analyses (0.1% dose and 1mm geometrical difference) to validate DECT image post processing. Finally, a patient- specific DECT-based SPR prediction was applied on 4D DECT datasets and followed by dose calculation to assess proton range shifts compared to the MonoCT-based HLUT approach.


Visually, no differences between the two sequential 4D DECT scans were found. Breathing patterns did not change more between the 2 scans than within a single scan. Clinical dose distributions on MonoCT and 140kVp datasets were similar with an average gamma passing rate of 99.9% (99.2%-100%). The maximal dose difference was 0.8%, Fig.1b. The worst-case scenario plans had a minimal passing rate of 92.4% (average 99.3%) with maximal dose difference of 3.3%. Using the MonoCT dataset with clinical HLUT instead of the DECT-based SPR dataset for dose calculation led to clinically relevant mean range shifts (±1SD) of 2.3(±0.8)%, Fig.2.


For this challenging patient cohort, sequentially acquired 4D DECT scans showed similar patient anatomy and stable breathing pattern allowing a consistent generation of DECT-based 79keV MonoCT datasets applicable for proton dose calculation. Patient-specific DECT-based SPR prediction on average CT datasets and breathing phases performed appropriately and can potentially reduce current range uncertainty in proton therapy. Even if large motion differences occur during the 2 sequential 4D DECT scans, dose distributions can still be reliably calculated using only the 140kVp dataset and beyond that important information on motion variability and robustness is gathered.

Keywords: dual-energy CT; proton range; range uncertainty

Publ.-Id: 26232