Session Item

Consistent deformable image registration (DIR) is the first step for reliable dose deformations. The current study aimed to test whether a commercially available DIR algorithm can reproducibly and consistently deform structures within the same patient between free breathing (FB) and deep inspiration breath hold (DIBH) breast cancer patients datasets.

The study includes 73 patients diagnosed with left sided breast cancer with CTs in FB and DIBH.  Breast clinical target volumes (CTV) according to ESTRO, the heart, the left anterior descending artery (LAD), both lungs were contoured in FB and DIBH in the RayStation (RaySearch Laboratories, Stockholm, Sweden). DIR was performed taking the structure set of the FB to be deformed on the DIBH (FB→DIBH) and vice versa (DIBH→FB). Further, DIR was performed focusing on the whole image (“no focus”- DIR_none), the CTV (DIR_CTV), the heart (DIR_heart) or on the surgical clip in the tumor bed (DIR_clip), respectively.

DSI= 2xCV / (mROI_V+dirROI_V);

CV- common volume of mROI_V and dirROI_V;

mROI_V - manually contoured region of interest volume;

dirROI_V - by DIR automatically generated volume.

Generalized estimating equations were used to assess the statistical effect of the different DIR focuses on the DSI.

Table 1 depicts the volume differences (∆V in %) between manually and automatically contoured structures exemplary for the CTV breast and the heart.

The highest DSI for the CTV was achieved by DIR_CTV (p<0.05) regardless of DIR direction (FB→DIBH or DIBH→FB); there was no statistical difference between DIR_clip and DIR_none for the DSI CTV. The DSI of the heart was the highest when focusing on the heart. There are no significant effects of breast size on DSI CTV_breast, ∆V breast left or ∆V heart. Patients with smaller breasts achieved higher DSI of the heart compared to those with larger breasts. By every inhaled 100cm³ the DSI CTV_breast would increase by 0.05 (p<0.05). In contrast the DSI heart would decrease by 0.002 (p<0.05) for every inhaled 100cm³.

The individual anatomy of the patient - such as breast volume or amount of deep inspiration – have an impact on DIR accuracy. DIR of the whole image is not enough for precise volumetry of organs and implicitly dose deformations between FB and DIBH datasets. Performing focused DIR is an option.