Regions of interest are routinely marked on x-ray CT and MR images by clinicians and radi--olo@ s... more Regions of interest are routinely marked on x-ray CT and MR images by clinicians and radi--olo@ sts and there is a need to compare variations in the definition o€ regions of interest amongst individuals and relate this back to the image'evidence'. Simple indices such as maximum projection, area, volume, and centre of mass are all excellent and robust ways of summarising variations. Radial signatures have been used to good effect for boundary comparison but are insensitive to errors due to rotation and can be unstable. Thus all the ...
2006 IEEE International Conference on Acoustics Speed and Signal Processing Proceedings
Wireless capsule endoscopy is a colour imaging technology that enables close examination of the i... more Wireless capsule endoscopy is a colour imaging technology that enables close examination of the interior of the entire small intestine. The wireless capsule endoscope (WCE) operates for ~ 8 hours and captures ~ 40,000 useful images. The images are viewed by a clinician as a video sequence, generally taking over an hour to analyse. In this paper we present a
Abstract. In this paper we present an approach for speeding-up the generation of Digitally Recons... more Abstract. In this paper we present an approach for speeding-up the generation of Digitally Reconstructed Radiographs (DRRs). DRRs are needed to confirm patient setup before preplanned clinical procedures such as robotic surgery or radiation therapy in a process known as 2D/3D medical image ...
A Fast and Automatic Approach for Removing Artefacts due to Immobilisation Masks in X-ray CT, 2017
Immobilisation masks are fixation devices that are used when administering radiotherapy treatment... more Immobilisation masks are fixation devices that are used when administering radiotherapy treatment to patients with tumours affecting the head and neck. Radiotherapy planning X-ray Computer Tomography (CT) data sets for these patients are captured with the immobilisation mask fitted and manually editing the X-ray CT images to remove artefacts due to the mask is time consuming and error prone. This paper represents the first study that employs a fast and automatic approach to remove image artefacts due to masks in X-ray CT images without affecting pixel values representing tissue. Our algorithm uses a fractional order Darwinian particle swarm optimisation of Otsu's method combined with morphological post-processing to classify pixels belonging to the mask. The proposed approach is tested on five X-ray CT data sets and achieves an average specificity of 92.01% and sensitivity of 99.39%. We also present results demonstrating the comparative speed-up obtained by fractional order Darwinian particle swarm optimisation.
Conformal therapy attempts to produce accurate medical treatment that will produce a uniform dose... more Conformal therapy attempts to produce accurate medical treatment that will produce a uniform dose over cancerous regions whilst at the same time sparing healthy tissues, especially the organs at risk. The constrained optimisation problem, that consists of working back from a given dose specification to elemental beam weight intensities is referred as the so called inverse problem. Due to the nature of the problem and in particular to its conflicting objectives, it is believed that heuristic techniques may have advantages over direct methods to solve for the beam intensities.
This paper explores the use of self-ensembling with random image augmentation [11] – a technique ... more This paper explores the use of self-ensembling with random image augmentation [11] – a technique that has achieved impressive results in the area of semi-supervised learning – for visual domain adaptation problems. We modify the approach of Laine et al.to improve stability and ease of use. Our approach demonstrates state of the art results when performing adaptation between the following pairs of datasets: MNIST and USPS, CIFAR-10 and STL, SVHN and MNIST, Syn-Digits to SVHN and SynSigns to GTSRB. We also explore the use of richer data augmentation to solve the challenging MNIST to SVHN adaptation path.
2018 14th IEEE International Conference on Signal Processing (ICSP)
JellyMonitor is an self-contained automated system that detects jellyfish blooms and reports thei... more JellyMonitor is an self-contained automated system that detects jellyfish blooms and reports their presence. It uses an embedded platform to analyse sonar imagery captured by a sonar imaging device. The software utilises a combination of classic computer vision techniques and deep neural networks to detect and classify objects captured by the sonar imaging device. We report on the development of this system and present results obtained from deploying a prototype.
Regions of interest are routinely marked on x-ray CT and MR images by clinicians and radi--olo@ s... more Regions of interest are routinely marked on x-ray CT and MR images by clinicians and radi--olo@ sts and there is a need to compare variations in the definition o€ regions of interest amongst individuals and relate this back to the image'evidence'. Simple indices such as maximum projection, area, volume, and centre of mass are all excellent and robust ways of summarising variations. Radial signatures have been used to good effect for boundary comparison but are insensitive to errors due to rotation and can be unstable. Thus all the ...
2006 IEEE International Conference on Acoustics Speed and Signal Processing Proceedings
Wireless capsule endoscopy is a colour imaging technology that enables close examination of the i... more Wireless capsule endoscopy is a colour imaging technology that enables close examination of the interior of the entire small intestine. The wireless capsule endoscope (WCE) operates for ~ 8 hours and captures ~ 40,000 useful images. The images are viewed by a clinician as a video sequence, generally taking over an hour to analyse. In this paper we present a
Abstract. In this paper we present an approach for speeding-up the generation of Digitally Recons... more Abstract. In this paper we present an approach for speeding-up the generation of Digitally Reconstructed Radiographs (DRRs). DRRs are needed to confirm patient setup before preplanned clinical procedures such as robotic surgery or radiation therapy in a process known as 2D/3D medical image ...
A Fast and Automatic Approach for Removing Artefacts due to Immobilisation Masks in X-ray CT, 2017
Immobilisation masks are fixation devices that are used when administering radiotherapy treatment... more Immobilisation masks are fixation devices that are used when administering radiotherapy treatment to patients with tumours affecting the head and neck. Radiotherapy planning X-ray Computer Tomography (CT) data sets for these patients are captured with the immobilisation mask fitted and manually editing the X-ray CT images to remove artefacts due to the mask is time consuming and error prone. This paper represents the first study that employs a fast and automatic approach to remove image artefacts due to masks in X-ray CT images without affecting pixel values representing tissue. Our algorithm uses a fractional order Darwinian particle swarm optimisation of Otsu's method combined with morphological post-processing to classify pixels belonging to the mask. The proposed approach is tested on five X-ray CT data sets and achieves an average specificity of 92.01% and sensitivity of 99.39%. We also present results demonstrating the comparative speed-up obtained by fractional order Darwinian particle swarm optimisation.
Conformal therapy attempts to produce accurate medical treatment that will produce a uniform dose... more Conformal therapy attempts to produce accurate medical treatment that will produce a uniform dose over cancerous regions whilst at the same time sparing healthy tissues, especially the organs at risk. The constrained optimisation problem, that consists of working back from a given dose specification to elemental beam weight intensities is referred as the so called inverse problem. Due to the nature of the problem and in particular to its conflicting objectives, it is believed that heuristic techniques may have advantages over direct methods to solve for the beam intensities.
This paper explores the use of self-ensembling with random image augmentation [11] – a technique ... more This paper explores the use of self-ensembling with random image augmentation [11] – a technique that has achieved impressive results in the area of semi-supervised learning – for visual domain adaptation problems. We modify the approach of Laine et al.to improve stability and ease of use. Our approach demonstrates state of the art results when performing adaptation between the following pairs of datasets: MNIST and USPS, CIFAR-10 and STL, SVHN and MNIST, Syn-Digits to SVHN and SynSigns to GTSRB. We also explore the use of richer data augmentation to solve the challenging MNIST to SVHN adaptation path.
2018 14th IEEE International Conference on Signal Processing (ICSP)
JellyMonitor is an self-contained automated system that detects jellyfish blooms and reports thei... more JellyMonitor is an self-contained automated system that detects jellyfish blooms and reports their presence. It uses an embedded platform to analyse sonar imagery captured by a sonar imaging device. The software utilises a combination of classic computer vision techniques and deep neural networks to detect and classify objects captured by the sonar imaging device. We report on the development of this system and present results obtained from deploying a prototype.
Uploads
Papers by Mark Fisher