Intraoperative segmentation of iodine and palladium radioactive sources in C-arm images

Abstract

An automated segmentation method was developed that is able to perform the identification and annotation processes of seeds on par with a human expert. This method was shown to be robust and suitable for integration in the dynamic dosimetry workflow of prostate brachytherapy interventions.

Segmentation, Separation and Pose Estimation of Prostate Brachytherapy Seeds in CT Images

“…Overall of 1063 seeds implanted in 14 patient cases in Table IV, the false positive (FD2 + FD3) percentage of our approach is 1.03%. This value is lower compared to the state-of-the-art methods reported in the Table II of the publication of San Filippo et al [24], where [17] was 2.2%, [22], [24] were 1.7%. These results open the door to accurate dose calculation and procedure quality assessment.…”

“…Numerous D'Amico et al [10] 2000 MRI 3D IO (1) N/G (2) X I d C Su et al [11] 2004 X-ray n2D (5) IO or IPO N/G X X N/G C+P Wei et al [12] 2006 US 3D IO X L X X D P Singh et al [13] 2007 X-ray n2D N/G N/G X N/G C+P Fallavollita et al [14] 2010 CT 3D IO N/G X N/G C+P Kuo et al [15] 2010 MRI 3D N/G N/G X D P Lee et al [16] 2011 X-Ray 2D IO N/G X Pd C Moult et al [17] 2012 X-Ray 2D IO N/G X X Id C+P Moult et al [18] 2012 X-Ray 2D IO N/G X X N/G C Defghan et al [19] 2012 studies [9]- [25] have been published based on different prostate image modalities, including magnetic resonance (MR), US, Xray and CT images. As previously mentioned, determining seed positions is useful at different stages of the clinical protocol: During the implantation US and/or X-ray images are necessary for simply monitoring seed deposition or for intraoperative dynamic dosimetry [12], [22]- [24] and iterative correction in case of inaccurate delivery; X-ray images are also used immediately after the implantation for recording purpose; finally, CT data are most often used after one month for dose evaluation. MRI can also be used postoperatively alone or in combination with CT.…”

“…For instance, a series of methods [14], [19], [22], [24], [25] propose a volume-to-volume and pointto-volume registration scheme of US images with the implants reconstructed from fluoroscopy. Recently, Hu et al [23] considered the prior knowledge of the US-to-CT transformation via registration and the planning seed position to define an atlas of regions of interest for seed detection.…”

Information processing in computer-assisted interventions: 4th international conference, 2013

“…The accepted papers represent a small selection of the contributions presented at the 4th IPCAI conference, held in Heidelberg, Germany, in 2013. They do, however, reflect the breadth of activity presented at the conference, in IJCARS, and the field in general, covering the key topics of 3D tracking for instrument navigation [2], intraoperative image segmentation for online radiation dosimetry [3], image-based anatomical models for catheter placement [4], and task and crisis analysis during surgical training through the combination of virtual reality and monitoring human multisensory responses to provide an immersive environment [5].…”

“…In [3], the authors describe an automatic method for detecting and localising brachytherapy seeds implanted into the prostate from C-arm X-ray images. This technique overcomes the issues of extended procedure times and operator error associated with manual identification of seeds, which limit the practical feasibility of online, dynamic radiation therapy planning in this application.…”