Image-guided adaptive radiation therapy (IGART), on the other hand, employs daily cone-beam CT (CBCT) imaging to track volumetric deformation, determine its dosimetric implications and adapt the treatment plan for optimal delivery. Researchers at the Henry Ford Health System (Detroit, MI) have developed such an IGART framework and validated their approach in phantoms and five early-stage prostate-cancer patient cases (Phys. Med. Biol. 57 5361).

"In prostate-cancer treatment, bladder and rectum filling has always been a challenge and the actual dose delivered to critical organs is quite patient dependent and higher than planned," said Ning Wen, physicist at the Henry Ford Department of Radiation Oncology. "IGART is particularly suited to prostate-cancer radiotherapy to understand and correct the dose discrepancy between planning and treatment."

Phantom tests

The IGART framework developed by Wen and colleagues incorporates deformable image registration (DIR) of the pre-treatment CBCT to the simulation-CT (SIM-CT) and DIR validation, as well as dose reconstruction, calculation of accumulated dose and treatment-plan evaluation. "IGART consolidates image guidance, DIR, dose mapping and re-optimization to fully elucidate the dosimetric impact and make corresponding corrections during the treatment course," explained Wen.

The team performed a series of phantom experiments to evaluate their IGART approach. They used a Catphan 500 phantom containing high-contrast targets to evaluate the Hounsfield Unit (HU) to electron-density relationship. The phantom was scanned using CBCT and multi-slice helical CT (the SIM-CT). On average, the HU for the CBCT was 10.1% lower than the HU for the SIM-CT, with the largest discrepancy observed for the densest target.

They also evaluated CBCT dose-calculation accuracy by scanning a Rando Pelvic phantom and calculating dose discrepancies between SIM-CT and CBCT for two treatment plans: a four-field box plan and a nine-field intensity-modulated radiation therapy (IMRT) plan. For the box plan with inhomogeneity correction, only slight deviations (below 2%) in dose distributions to the prostate and organs-at-risk were observed between CBCT and SIM-CT. For IMRT, the prostate dose distribution was very similar between SIM-CT and CBCT. The dose to the rectum showed greater discrepancy, though the absolute dose difference (0.3 Gy) was small.

Patient plans

Having confirmed the accuracy of the dose calculation in the phantom studies, the researchers applied IGART to five patient plans. Daily CBCTs were acquired before radiotherapy and patients were repositioned based on rigid registration. DIR was performed offline to determine the actual delivered dose.

The integrity of the IGART framework is heavily reliant on DIR performance; thus the researchers evaluated the registration accuracy in depth. First, they visually compared SIM-CT and deformed CBCT images for 38 fractions of one patient's treatment. The images matched well in the prostate region, although the deformed CBCT showed distortion in the superior/inferior borders.

To quantitatively evaluate DIR performance, they compared the offsets of three implanted markers following rigid registration and DIR. Seed positions correlated well between the SIM-CT and the deformed CBCT after DIR, with a mean marker offset of 0.0 mm in all directions and standard deviations of 0.5, 0.7, and 0.7 mm. Displacements were larger for the rigid registration.

The researchers also used finite element modelling (FEM) to compute the unbalanced energy (UE), which validates registration accuracy at a voxel level. Unbalanced energy is based on the concept of unbalanced forces. If a registration is theoretically perfect, the unbalanced forces are zero. However, since registrations are imperfect, one can compute the unbalanced forces and hence the unbalanced energy. Compared with the SIM-CT image, the deformed CBCT image was distorted in the regions of seminal vesicles, bladder and rectum (these regions exhibited high UE values). In contrast, UE values were small in the prostate region, indicating high DIR accuracy.

In the final part of this work, the dose on each CBCT image set was reconstructed and accumulated over all fractions to reflect the actual dose delivered to the patient. The team then compared the delivered plans to the original (static) plans to evaluate dose discrepancies for the tumour and organs-at-risk.

Isodose distributions were similar for the static and delivered cases, with clinical target volume (CTV) and planning target volume (PTV) well covered by the prescription dose. For standard clinical planning margins (10 mm surrounding the prostate uniformly except at the prostate/rectal wall interface (6 mm)), dose–volume histograms revealed that CTV coverage was similar between static and delivered doses, whereas dose degradation was apparent in the delivered PTV minimum doses for some patients.

However, for all patients, the maximum and mean PTV doses were still within 2%. Adaptive planning and dose accumulation showed much larger dosimetric differences to the normal tissues, rectum and bladder, likely due to daily physiological changes in these organs and associated deformation.

Examining the change in normal-tissue complication probability (NTCP) between static and delivered plans revealed a mean change in NTCP for grade 2/3 rectal bleeding of 4.6%. NTCP for grade 3 bleeding increased by 4.5%. NCTP for Grade 1 or greater late genitourinary toxicity decreased by 7.3%, since the complications were directly related to maximum bladder dose, which decreased in the delivered dose estimations.

The authors conclude that these results support adaptive planning and its use to determine dosimetric impact. Next, they will retrospectively analyse a large population of prostate-cancer patients to compare two IGART approaches to determine the one with the highest therapeutic ratio. "The first generates the optimized dose using standard clinical margins by implementing an offline replanning-based IGART framework," explained Wen. "The second involves calculating the actual dose delivered to patients at two smaller margin settings without plan adaptation."