Abstract : Cryotherapy is a rapidly growing minimally invasive technique for the treatment of different kinds of tumors, such as breast cancer, renal and prostate cancer. Several hollow needles are percutaneously inserted in the target area under image guidance and a gas (usually argon) is then decompressed inside the needles. Based on the Thompson-Joule principle, the temperature drops drown and a ball of ice crystals forms around the tip of each needle. Radiologists rely on the geometry of this iceball (273K), visible on computer tomographic (CT) or magnetic resonance (MR) images, to assess the status of the ablation. However, cellular death only occurs when the temperature falls below 233K. The complexity of the procedure therefore resides in planning the optimal number, position and orientation of the needles required to treat the tumor, while avoiding any damage to the surrounding healthy tissues.
This planning is currently done qualitatively, based on experience, and can take several hours, with a result that is often different from the expected one. To solve this important limitation of cryotherapy, a few planning systems have been proposed in the literature. Currently, commercial systems are nearly non existent, and emerging tools are limited to a visualization of the isotherms obtained for each needle in ideal conditions (usually in a gel). They do not account for any influence of the soft tissue properties, the presence of blood vessels, or the combined effect of multiple needles. As a consequence, large safety margins over 5mm are defined.
To address this challenge, our method extracts information from medical images (CT or MR) and allows to assess different strategies with an augmented visualization of the resulting iceball and the associated isotherms.