Biomedical subjects

Adrian Chung

Publications and source records attributed to Adrian Chung.

3 recordsLinked to original sources

PubMed2005

Photo-realistic tissue reflectance modelling for minimally invasive surgical simulation.

Computer-based simulation is an important tool for surgical skills training and assessment. In general, the degree of realism experienced by the trainees is determined by the visual and biomechanical fidelity of the simulator. In minimally invasive surgery, specular reflections provide an important visual cue for tissue deformation, depth and orientation. This paper describes a novel image-based lighting technique that is particularly suitable for modeling mucous-covered tissue surfaces. We describe how noise functions can be used to control the shape of the specular highlights, and how texture noise is generated and encoded in image-based structure at a pre-processing stage. The proposed technique can be implemented at run-time by using the graphics processor to efficiently attain pixel-level control and photo-realism. The practical value of the technique is assessed with detailed visual scoring and cross comparison experiments by two groups of observers.

Bronchoscopy↗

PubMed2005

Predictive camera tracking for bronchoscope simulation with CONDensation.

This paper exploits the use of temporal information to minimize the ambiguity of camera motion tracking in bronchoscope simulation. The condensation algorithm (Sequential Monte Carlo) has been used to propagate the probability distribution of the state space. For motion prediction, a second-order auto-regressive model has been used to characterize camera motion in a bounded lumen as encountered in bronchoscope examination. The method caters for multimodal probability distributions, and experimental results from both phantom and patient data demonstrate a significant improvement in tracking accuracy especially in cases where there is airway deformation and image artefacts.

Algorithms↗

PubMed2004

Patient-specific bronchoscope simulation with pq-space-based 2D/3D registration.

OBJECTIVE: The use of patient-specific models for surgical simulation requires photorealistic rendering of 3D structure and surface properties. For bronchoscope simulation, this requires augmenting virtual bronchoscope views generated from 3D tomographic data with patient-specific bronchoscope videos. To facilitate matching of video images to the geometry extracted from 3D tomographic data, this paper presents a new pq-space-based 2D/3D registration method for camera pose estimation in bronchoscope tracking. METHODS: The proposed technique involves the extraction of surface normals for each pixel of the video images by using a linear local shape-from-shading algorithm derived from the unique camera/lighting constraints of the endoscopes. The resultant pq-vectors are then matched to those of the 3D model by differentiation of the z-buffer. A similarity measure based on angular deviations of the pq-vectors is used to provide a robust 2D/3D registration framework. Localization of tissue deformation is considered by assessing the temporal variation of the pq-vectors between subsequent frames. RESULTS: The accuracy of the proposed method was assessed by using an electromagnetic tracker and a specially constructed airway phantom. Preliminary in vivo validation of the proposed method was performed on a matched patient bronchoscope video sequence and 3D CT data. Comparison to existing intensity-based techniques was also made. CONCLUSION: The proposed method does not involve explicit feature extraction and is relatively immune to illumination changes. The temporal variation of the pq distribution also permits the identification of localized deformation, which offers an effective way of excluding such areas from the registration process.

Algorithms↗