Rapid three-dimensional seed reconstruction from incomplete data sets for image guided prostate brachytherapy

Prostate cancer is one of the most common forms of cancer among men. Treatment of prostate cancer involves implanting radioactive seeds into the prostate, a procedure known as prostate brachytherapy. Brachytherapy involves implanting seeds precisely into locations as defined by the plan. Any deviation from the plan could lead to lower cure rates. Dosimetry calculation involves computing the 3D locations of the seeds. Two major challenges for performing dosimetry in the operating room (intraoperative dosimetry) are (a) speed of computation and (b) undetected seeds on any of the three 2D images. We propose an epipolar geometry based reconstruction algorithm that is computationally efficient and can handle incomplete data sets. The algorithm divides the images into bands for speed of matching and reconstruction and uses a duplicate and match methodology incase any of the bands have undetected seeds. We test the algorithm on 205 patient cases.; In order to obtain a complete picture of the doses the prostate boundary needs to be visualized along with the seeds. Prostate boundary is not visualized on fluoroscopy, the imaging modality used for seed reconstruction. Ultrasound is a modality that has a good prostate visualization but poor seed visualization. The two imaging modalities can be used for their strengths and data fusion can be performed to get the complete picture. We provide three fusion techniques (a) dummy probe based fusion; (b) embedded marker based fusion and (c) fusion using seeds themselves. The algorithm was tested using a few phantoms and also on patient cases.