Paper Submission & Registration
8th Dutch Bio-Medical Engineering Conference
16:30   Oncology - I
Chair: Peter van Ooijen
15 mins
Automatic trajectory planning for IRE treatment in liver tumours A numerical study
Girindra Wardhana, Adriana Leticia Vera Tizatl, Momen Abayazid, Jurgen J Fütterer
Abstract: Background: Irreversible electroporation (IRE) had shown promising results for treating tumors by avoiding heat/cold sink effect and preserving vital structure from thermal injury. Treatment of tumor demands maximization of ablation coverage by finding optimum electrode positioning. However, there has been little attention to the position feasibility during needle insertion. Purpose: This study attempts to combine IRE simulation with needle trajectory planning in order to estimate the actual volume of the ablation region during IRE therapy. Material and Methods: The trajectory of multiple needles is planned automatically using 3D segmentation of critical structures. Selection of possible insertion zone is considered as a multi-objective optimization problem which should optimize clinical requirements, such as distance to critical structure, needle length, and insertion angle. The final insertion points are chosen using Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) [1] and Faire Un Choix Adéquat (FUCA) [2]. Insertion points are evaluated by finite element modeling in 3D realistic models of target tissues (liver tumors and surrounding non-cancerous tissue considered as a 5 mm safety margin) to determine the maximum target volumes exposed to IRE (Electric Field ≥ 800 V/cm) based on the location of the needles and pulse amplitude tuning. Results: Seven tumors (n=3; diameter ≤ 1cm, n=3; 1cm
15 mins
Phantom-based investigation of the relationship between microvascular architecture and ultrasound-contrast-agent kinetics
Peiran Chen, Simona Turco, Ruud van Sloun, Andreas Pollet, Jaap den Toonder, Hessel Wijkstra, Massimo Mischi
Abstract: Introduction As a recognized hallmark of cancer, tumor-driven angiogenic microvasculature is characterized by increased microvascular density (MVD), smaller vessel diameter (dv), and higher vessel tortuosity, leading to complex blood flow patterns [1, 2]. Analysis of the dispersion kinetics of an ultrasound contrast agent (UCA) in the tumor vasculature has shown promise for prostate cancer (PCa) diagnostics [1, 2], but the link between UCA kinetics and underlying microvascular architecture is still under investigation. In this work, modelling the microvasculature as a porous medium, we developed a set-up including dedicated porous phantoms to investigate this relationship between convective-dispersion parameters and microvascular architecture with dynamic contrast-enhanced ultrasound (DCE-US). Moreover, we developed a simulation framework to further advance our understanding of the UCA kinetics through porous media. Methods The porous phantoms were built by packing 3% alginate beads in a polyurethane tube. Variable MVD and dv were realized by packing beads with diameters of 3.1, 2.5, and 1.6 mm, respectively. DCE-US was performed to record the UCA flow through the realized phantoms using a Verasonics ultrasound platform. The velocity and dispersion coefficient were estimated by a model-based deconvolution method. The in-silico model of a porous phantom was realized using a 3D sphere-packing algorithm. Thousands of mono-sized spheres were uniformly distributed inside a cylindrical space. Spheres were then separated iteratively until no overlapping was present. The architecture parameters including porosity, pore size and pore density were analysed. Dynamic simulations of bubble transport flowing through the porous phantoms are currently performed using COMSOL. Results and Conclusions The pore size and density distribution demonstrate that the pore size decreases while the density increases as the bead size decreases, representing smaller dv and higher MVD. This result confirms our hypothesis that porous phantoms with smaller beads can mimic angiogenic microvasculature. Moreover, the estimated velocity increases and the dispersion coefficient decreases with decreasing bead size, which is in line with previous in-vivo finding [1, 2]. This investigation represents a first step in understanding the relationship between microvascular architecture and UCA dynamics. [1] Kuenen et al., IEEE TMI, 2011. [2] van Sloun et al., Med. Image Anal, 2016.
15 mins
Anton Nikolaev, Leon de Jong, Gert Wijers Wijers, Vincent Groenhuis, Françoise Siepel, Stefano Stramigioli, Hendrik Hansen, Chris de Korte
Abstract: Volumetric ultrasound breast imaging, combined with MRI can improve lesion detection rate, reduce examination time, and improve lesion diagnosis. However, to our knowledge, there are no 3D US breast imaging systems available that facilitate 3D US-MRI image fusion. In this work, we introduce a novel Automated Cone-based Breast Ultrasound System (ACBUS) for breast screening and biopsy. The system facilitates volumetric ultrasound acquisition of the breast in a prone position without deforming it by US transducer. To our conclusion, ACBUS might be a suitable candidate for a second-look breast US exam, patient follow-up, and US-guided biopsy planning.
15 mins
Deep synthesis from MR to CT using U-Net
Bingjiang Qiu
Abstract: Purpose: Three dimensional virtual surgical planning (3D VSP) and guided surgery techniques are widely applied in the treatment of oral cancer with tumor removal. In the absence of CT images, MRI images can not be used alone for tumor resection and image guidance in 3D VSP. Moreover, the workflow requires image registration in MRI and CT modalities. Considering the ionizing radiation of CT scanning, synthetic CT based on MRI data is increasingly focused on 3D VSP. This work utilizes a U-Net approach to generate synthetic CT images using MRI data that provides a strategy of MRI-only treatment planning into the 3D VSP. Methods and Materials: The experiment is performed using head and neck (H&N) dataset of 28 CT and MRI scans. Each MR or CT volume contains 100-142 axial 2D image slices. These were resampled into 256×256 pixel images with 256 gray values to achieve uniform distribution. The dataset was randomly split into the training, validation and test subsets, each of which contained 13, 1 and 14 cases, respectively. We adopt a U-Net architecture that combines L1 loss and structural similarity index (SSIM) loss. Results: The performance of this method was evaluated by comparing with the reference CT images. The average peak signal to noise ratio (PSNR) of 25.3493 ± 0.6042, the normalized mean square error (NMSE) of 0.0321 ± 0.0042 and the SSIM of 0.7598 ± 0.0317 were obtained. The experimental results demonstrate that the synthesized images are closer to the reference CT images. Conclusion: This method provides accurate synthetic CT performance from MRI for H&N scans. The results show that the proposed method has the potential to generate synthetic CT images to support MRI-only treatment planning and image-guided workflow in 3D VSP.

end %-->