Nuwan D. Nanayakkara

Prostate segmentation by feature enhancement using domain knowledge and adaptive region based operations

Estimation of prostate location and volume is essential in determining a dose plan for ultrasound-guided brachytherapy, a common prostate cancer treatment. However, manual segmentation is difficult, time consuming and prone to variability. In this paper, we present a semi-automatic discrete dynamic contour (DDC) model based image segmentation algorithm, which effectively combines a multi-resolution model refinement procedure together with the domain knowledge of the image class. The segmentation begins on a low-resolution image by defining a closed DDC model by the user. This contour model is then deformed progressively towards higher resolution images. We use a combination of a domain knowledge based fuzzy inference system (FIS) and a set of adaptive region based operators to enhance the edges of interest and to govern the model refinement using a DDC model. The automatic vertex relocation process, embedded into the algorithm, relocates deviated contour points back onto the actual prostate boundary, eliminating the need of user interaction after initialization. The accuracy of the prostate boundary produced by the proposed algorithm was evaluated by comparing it with a manually outlined contour by an expert observer. We used this algorithm to segment the prostate boundary in 114 2D transrectal ultrasound (TRUS) images of six patients scheduled for brachytherapy. The mean distance between the contours produced by the proposed algorithm and the manual outlines was 2.70 +/- 0.51 pixels (0.54 +/- 0.10 mm). We also showed that the algorithm is insensitive to variations of the initial model and parameter values, thus increasing the accuracy and reproducibility of the resulting boundaries in the presence of noise and artefacts.

Mobile based wound measurement

This paper proposes a portable wound area measurement method based on the segmentation of digital images. Its objective is to provide a practical, fast and non-invasive technique for medical staff to monitor the healing process of chronic wounds. Segmentation is based on active contour models which identifies the wound border irrespective of coloration and shape. The initial segmentation can also be modified by the user, providing higher control and accuracy. Area measurements are further normalized to remove effects of camera distance and angle. The application has been implemented for the Android platform version 2.2 with a prototype model running on Samsung Galaxy Tab. The results to evaluate the efficacy of the application have been encouraging with an accuracy level of 90%.

Automatic prostate boundary detection in ultrasound images using multiresolution deformable models and fuzzy logic

In this paper, we propose an automatic model based image segmentation system, where the instantiated model is refined incrementally using the domain knowledge combined by Fuzzy Logic. The Fuzzy Inference System (FIS) combines several different image features, which are used by experts to detect prostates in noisy ultrasound images. We use the Discrete Dynamic Contour (DDC) model because of its favorable performances in both open and closed contour models. The FIS governs the automatic open DDC model initialization and the following incremental growing process on a low-resolution image. At this stage, the initial open contour model grows by tracking the coarse edge details until it closes. The resulting closed contour model is then refined incrementally up to the original image resolution, incorporating finer edge details on to the model. The algorithm developed here is a general tool for object detection in an image analysis system, which employs a flexible framework designed to support multiple decision tools to collaborate in forming a solution. The FIS in our tool retrieves the domain knowledge it needs from the framework, to govern the model refinement process. The proposed algorithm can be used to detect the boundary of any object on an image, if the knowledge of the dominant image features is stored in the system. We have included results of the algorithm successfully applied to several ultrasound images to define the boundary of the prostate.

Unsupervised model based image segmentation using domain knowledge based fuzzy logic and edge enhancement

In this paper, we present an automatic model based image segmentation system, which combines a multi-resolution discrete dynamic contour (DDC) model refinement procedure and the domain knowledge of the image class. The segmentation begins on a low-resolution image by defining an open DDC model, followed by a contour growing process generates the closed DDC model, which deforms progressively towards higher resolution images. A combination of knowledge based fuzzy inference system (FIS) and a set of adaptive region based operators is used to enhance the edges of interest and to govern the DDC model deformation. With the above process we were able to greatly reduce the sensitivity to the initial model, thus paving the way for automatic segmentation on noisy images. Domain knowledge of a particular class of images is encapsulated within the FIS such that it can be easily changed for different image classes. We applied this algorithm successfully to detect the organ boundary in ultra-sound images of prostates and examples are shown in order to illustrate the advantages of the proposed method.

Automated diagnosis of cardiac abnormalities using heart sounds

Listening to the heart sounds is a common practice in identifying cardiac malfunctions. Since this method has many limitations, tools that aid physicians in their diagnosis of heart diseases are very useful. This paper presents a software tool to predict cardiac abnormalities which can be identified using heart sounds. Both heart sound information and symptoms are used in disease prediction. First audio inputs at four clinically important locations on the chest are acquired using an electronic stethoscope and entered to a database with symptoms for each patient. After de-noising, prominent features and statistical parameters needed for disease detection are extracted from the heart sound samples using several algorithms. Then the disease classification is performed to find out possible disease and murmur types. The software tool reported in this paper is capable of identifying normal heart sounds and abnormal heart sounds with possible kind of disease and murmurs presented there. Hence, it helps doctors to detect diseases early and can be integrated as a standard module of electronic stethoscope software.

A surface-based metric for registration error quantification

In medical image registration, the quantification of registration errors is important in deciding the capabilities of a registration technique for a given problem, and/or for a given pair of images. The most common approach is the geometrical registration error called Target Registration Error (TRE) that measures the distance between corresponding landmarks in the target and registered images. However, finding sufficient number of corresponding landmarks is not always possible in medical images, and therefore, other measures such as, image similarity measures and surface-based error metrics have been used in quantification of registration errors. Surface-based error quantification is more appropriate than intensity-based methods, but the widely used surface-based Closest Point Registration Error (CPRE) is known for under-estimating registration errors. In this paper, we present a surface-based method for quantification of registration errors using Matched Points Registration Error (MPRE) by computing distances between “matched-points” on segmented object surfaces in target and registered images. We compared small rigid registration errors of tube-shaped and closed surface objects quantified using MPRE with TRE and CPRE, and showed that MPRE did not show a significant difference from TRE and that CPRE was significantly lower than both MPRE and TRE.

Morphological variations of the human ejaculatory ducts in relation to the prostatic urethra: Morphological Variations of the Human Ejaculatory Ducts

Loss of ejaculation can follow transurethral resection of the prostate (TURP). Periverumontanal prostate tissue is preserved in ejaculation‐preserving TURP (ep‐TURP). Knowledge of ejaculatory duct anatomy in relation to the prostatic urethra can help in ep‐TURP. This was evaluated in cross‐sections of the prostate using a 3 D model to determine a safe zone for resecting the prostate in ep‐TURP. A 3 D reconstruction of the ejaculatory ducts was developed on the basis of six prostate gland cross‐sections. The measurements obtained from the 3 D model were standardized according to the maximum width of the prostate. Simple linear regressions were used to predict the relationships of the ejaculatory ducts. The maximum widths of the prostates ranged from 22.60 to 52.10 mm. The ejaculatory ducts entered the prostate with a concavity directed posterolaterally. They then proceeded toward the seminal colliculus in a fairly straight course, and from that point they angulated anteromedially. As they opened into the prostatic urethra they diverged. Significant regression models predicted the relationships of the ejaculatory ducts to the prostatic urethra based on the sizes of the prostates. The 3 D anatomy of ejaculatory ducts can be predicted on the basis of prostate width. The ejaculatory ducts can be preserved with 95% accuracy if a block of tissue 7.5 mm from the midline on either side of the seminal colliculus is preserved, up to 10 mm proximal to the level of the seminal colliculus, during TURP. Clin. Anat. 31:456–461, 2018.

Evaluating effects of haptic feedback in Virtual Reality simulators for laparoscopic skill development

Virtual Reality (VR) simulators are currently accepted as a good way of training of laparoscopic surgeries. Even with several commercially available VR simulators, trainees are still unable to obtain a proper psychomotor abilities and skills needed for MIS due to lack effectiveness in existing simulators. Realistic organ-force model is a key requirement of a VR simulator to experience real time interaction forces. This is critical in Minimally Invasive Surgeries (MIS) due to complex behavior of biological tissues and anatomical variability. We previously presented a novel method to integrate soft, firm and hard tissue properties into abdomen organ models by changing the stiffness properties of organ models. Our system was developed using the software libraries of Open Haptic Toolkit from SensAble Technology incorporated with the graphic libraries in Open GL and a Phantom Omni Haptic device with 6 Degrees of Freedom (DoF) position sensing and 3 DoF of force feedback. The simulated haptic models were evaluated with experienced surgeons in the field. They were also used to evaluate the effectiveness of force feedback for laparoscopic surgical skill development for surgical interns. In this paper, we present experimental results obtained from experts and trainee surgeons.

Stitching 3D ultrasound head images of neonates to monitor changes in ventricular volume

Ultrasound (US) images are taken in 3-dimensional (3D) space to monitor the changes in the ventricular volume of neonates. Due to the small field of view of US imaging, the head image is taken as two parts. To properly monitor and analyze the images, the two parts need to be stitched together in 3D space. In this paper, we propose a technique based on rigid registration for this purpose. User provided corresponding landmarks of the two images are used for coarse alignment. Then mutual information based rigid transformation is performed for fine alignment of moving image on to the fixed image. Finally, the fixed and moving images are stitched together to produce the final output. Furthermore, we have analyzed the optimal parameter settings for the registration procedure to minimize the registration error. We achieved a mean registration accuracy of 2.71mm for four data sets of neonatal 3D US head images using the proposed method.

Toolkit for extracting electrocardiogram signals from scanned trace reports

Cardiovascular disease (CVD) is the leading cause of death throughout the world. Since electrocardiogram-reports (ECG) have a great CVD predicting potential, the demand for their real-time analysis is high. Although algorithms are present to perform analysis, most countries still use analogue acquisition systems that can only output a printed trace. It is necessary to extract the signal from these printouts to perform analysis. With time, as the reports pile up and the trace fades from the printout, the task becomes increasingly difficult. The method presented specifically focuses on extracting signals from faded traces. Due to the large variability of scans, it is difficult to automate this task completely. In this paper, we propose several tools for ECG extraction while maintaining a minimum user involvement requirement. The proposed method was tested on a dataset of 550 trace snippets and comparative analysis shows an average accuracy of 96%.