Deformable Registration of CT and Ultrasound Images for Radiation Treatment of Skin Cancer
V Hart1*, T Liu2, K Fishman3, J Wilson1, X Li1, (1) Medical College of Wisconsin, Milwaukee, WI, (2) Emory Univ, Atlanta, GA, (3) Sensus Healthcare, Boca Raton, FLSU-E-U-11 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
Purpose: High-frequency ultrasound (HFUS) images capable of revealing accurate spatial information in skin lesions may be used to plan and guide radiation therapy (RT) for skin cancer. In this work, we develop a method for deformable image registration of ultrasound and CT images for skin cancer RT.
Methods: Skin lesion images acquired with HFUS and the corresponding CTs were registered in the regions of interest. A correspondence function was used to map attenuation values to tumor elasticity and a histogram matching was implemented. A novel DIR technique, based on the symmetric force Demons algorithm, was used in which the size of the Gaussian smoothing kernel was adaptively adjusted. The Pearson correlation coefficient (PCC) was used to assess the registration accuracy.
Results: Observed PCC values of 0.9794 and 0.9815 (enhanced contrast) indicated excellent agreement between the dynamic (HFUS) and static (CT) images. The technique exhibited an ability to process large deformations as ROI displacement exceeding tumor thickness was observed near the edges of the lesion. The gradual decrease of kernel dimension was found to prevent non-physical warping, thereby improving registration accuracy. This robustness is critical to skin cancer RT as tumors may shrink significantly during treatment. Run times averaged 3.26 seconds per fraction.
Conclusion: Combined with image processing techniques, the proposed method is sufficiently accurate for deformable registration of HFUS and CT images for skin cancer RT. The ability to perform image-guided treatment for skin cancer could reduce the number of biopsies needed during tumor excision and could spare healthy tissue. The presented technique is accurate and robust enough to process large deformations, making it a promising tool for RT planning and delivery guidance for skin cancer.