A novel approach to the assessment of response to chemotherapy in sarcoma imaged with PET-FDG

Human sarcoma is a class of malignant tumors affecting soft tissue, cartilage and bone. Standard therapy typically involves resection with neo-adjuvant chemotherapy being employed to shrink the tumor mass prior to surgery. The effectiveness of chemotherapy is variable and there is on-going interest in being able to rapidly determining this so that alternative therapies might be considered, when indicated. We are exploring ways in which PET imaging with FDG can be used for the assessment of chemotherapy response with sarcoma. Quantitative assessment requires an alignment or co-registration of pre- and post-therapy scans data. Obviously PET attenuation information is key but this can be augmented to ensure that the uptake weighted spatial mean and covariance of the tumor pre- and post-therapy are also aligned. This is achieved by means of a Procrustes rotation procedure applied to the pre- and post-covariance matrices. Once the pre- and post-therapy scans are aligned, a statistical summary of voxel-by-voxel changes in uptake can be used to assess the chemotherapy response. An advantage of this assessment is that it is necessarily sensitive to the changes in the spatial pattern of uptake within the tumor. A regression analysis is proposed in which the average change from the post- to the pre-therapy values is evaluated in terms of a standardized coefficient that has the form of a z-test statistic. The statistic provides a more meaningful scale in which to evaluate chemotherapy effects. We illustrate the use of these procedures is a series of data from a set of 48 sarcoma patients. The relation between the proposed chemotherapy response assessment and patient outcome shows promise.