Automatic quantification of multi-modal rigid registration accuracy using feature detectors.

In radiotherapy, the use of multi-modal images can improve tumor and target volume delineation. Images acquired at different times by different modalities need to be aligned into a single coordinate system by 3D/3D registration. State of the art methods for validation of registration are visual inspection by experts and fiducial-based evaluation. Visual inspection is a qualitative, subjective measure, while fiducial markers sometimes suffer from limited clinical acceptance. In this paper we present an automatic, non-invasive method for assessing the quality of intensity-based multi-modal rigid registration using feature detectors. After registration, interest points are identified on both image data sets using either speeded-up robust features or Harris feature detectors. The quality of the registration is defined by the mean Euclidean distance between matching interest point pairs. The method was evaluated on three multi-modal datasets: an ex vivo porcine skull (CT, CBCT, MR), seven in vivo brain cases (CT, MR) and 25 in vivo lung cases (CT, CBCT). Both a qualitative (visual inspection by radiation oncologist) and a quantitative (mean target registration error-mTRE-based on selected markers) method were employed. In the porcine skull dataset, the manual and Harris detectors give comparable results but both overestimated the gold standard mTRE based on fiducial markers. For instance, for CT-MR-T1 registration, the mTREman (based on manually annotated landmarks) was 2.2 mm whereas mTREHarris (based on landmarks found by the Harris detector) was 4.1 mm, and mTRESURF (based on landmarks found by the SURF detector) was 8 mm. In lung cases, the difference between mTREman and mTREHarris was less than 1 mm, while the difference between mTREman and mTRESURF was up to 3 mm. The Harris detector performed better than the SURF detector with a resulting estimated registration error close to the gold standard. Therefore the Harris detector was shown to be the more suitable method to automatically quantify the geometric accuracy of multimodal rigid registration.

Performance validation of deformable image registration in the pelvic region.

Patients undergoing radiotherapy will inevitably show anatomical changes during the course of treatment. These can be weight loss, tumour shrinkage, and organ motion or filling changes. For advanced and adaptive radiotherapy (ART) information about anatomical changes must be extracted from repeated images in order to be able to evaluate and manage these changes. Deformable image registration (DIR) is a tool that can be used to efficiently gather information about anatomical changes. The aim of the present study was to evaluate the performance of two DIR methods for automatic organ at risk (OAR) contour propagation. Datasets from ten gynaecological patients having repeated computed tomography (CT) and cone beam computed tomography (CBCT) scans were collected. Contours were delineated on the planning CT and on every repeated scan by an expert clinician. DIR using our in-house developed featurelet-based method and the iPlan(®) BrainLab treatment planning system software was performed with the planning CT as reference and a selection of repeated scans as the target dataset. The planning CT contours were deformed using the resulting deformation fields and compared to the manually defined contours. Dice's similarity coefficients (DSCs) were calculated for each fractional patient scan structure, comparing the volume overlap using DIR with that using rigid registration only. No significant improvement in volume overlap was found after DIR as compared with rigid registration, independent of which image modality or DIR method was used. DIR needs to be further improved in order to facilitate contour propagation in the pelvic region in ART approaches.

SU-E-J-90: 2D/3D Registration Using KV-MV Image Pairs for Higher Accuracy Image Guided Radiotherapy.

PURPOSE: In this work, we investigate the impact of using paired portal mega-voltage (MV) and kilo-voltage (kV) images, on 2D/3D registration accuracy with the purpose of improving tumor motion tracking during radiotherapy. Tumor motion tracking is important as motion remains one of the biggest sources of uncertainty in dose application. 2D/3D registration is successfully used in online tumor motion tracking, nevertheless, one limitation of this technique is the inability to resolve movement along the imaging beam axis using only one projection image. METHODS: Our evaluation consisted in comparing the accuracy of registration using different 2D image combinations: only one 2D image (1-kV), one kV and one MV image (1kV-1MV) and two kV images (2-kV). For each of the image combinations we evaluated the registration results using 250 starting points as initial displacements from the gold standard. We measured the final mean target registration error (mTRE) and the success rate for each registration. Each of the combinations was evaluated using four different merit functions. RESULTS: When using the MI merit function (a popular choice for this application) the RMS mTRE drops from 6.4 mm when using only one image to 2.1 mm when using image pairs. The success rate increases from 62% to 99.6%. A similar trend was observed for all four merit functions. Typically, the results are slightly better with 2-kV images than with 1kV-1MV. CONCLUSIONS: We evaluated the impact of using different image combinations on accuracy of 2D/3D registration for tumor motion monitoring. Our results show that using a kV-MV image pair, leads to improved results as motion can be accurately resolved in six degrees of freedom. Given the possibility to acquire these two images simultaneously, this is not only very workflow efficient but is also shown to be a good approach to improve registration accuracy.