Bahbibi Rahmatullah

Evaluation and Comparison of Current Fetal Ultrasound Image Segmentation Methods for Biometric Measurements: A Grand Challenge

This paper presents the evaluation results of the methods submitted to Challenge US: Biometric Measurements from Fetal Ultrasound Images, a segmentation challenge held at the IEEE International Symposium on Biomedical Imaging 2012. The challenge was set to compare and evaluate current fetal ultrasound image segmentation methods. It consisted of automatically segmenting fetal anatomical structures to measure standard obstetric biometric parameters, from 2D fetal ultrasound images taken on fetuses at different gestational ages (21 weeks, 28 weeks, and 33 weeks) and with varying image quality to reflect data encountered in real clinical environments. Four independent sub-challenges were proposed, according to the objects of interest measured in clinical practice: abdomen, head, femur, and whole fetus. Five teams participated in the head sub-challenge and two teams in the femur sub-challenge, including one team who tackled both. Nobody attempted the abdomen and whole fetus sub-challenges. The challenge goals were two-fold and the participants were asked to submit the segmentation results as well as the measurements derived from the segmented objects. Extensive quantitative (region-based, distance-based, and Bland-Altman measurements) and qualitative evaluation was performed to compare the results from a representative selection of current methods submitted to the challenge. Several experts (three for the head sub-challenge and two for the femur sub-challenge), with different degrees of expertise, manually delineated the objects of interest to define the ground truth used within the evaluation framework. For the head sub-challenge, several groups produced results that could be potentially used in clinical settings, with comparable performance to manual delineations. The femur sub-challenge had inferior performance to the head sub-challenge due to the fact that it is a harder segmentation problem and that the techniques presented relied more on the femurs appearance.

Integration of Local and Global Features for Anatomical Object Detection in Ultrasound

The use of classifier-based object detection has found to be a promising approach in medical anatomy detection. In ultrasound images, the detection task is very challenging due to speckle, shadows and low contrast characteristic features. Typical detection algorithms that use purely intensity-based image features with an exhaustive scan of the image (sliding window approach) tend not to perform very well and incur a very high computational cost. The proposed approach in this paper achieves a significant improvement in detection rates while avoiding exhaustive scanning, thereby gaining a large increase in speed. Our approach uses the combination of local features from an intensity image and global features derived from a local phase-based image known as feature symmetry. The proposed approach has been applied to 2384 two-dimensional (2D) fetal ultrasound abdominal images for the detection of the stomach and the umbilical vein. The results presented show that it outperforms prior related work that uses only local or only global features.

Towards on Develop a Framework for the Evaluation and Benchmarking of Skin Detectors Based on Artificial Intelligent Models using Multi-Criteria Decision-Making Techniques

Evaluation and benchmarking of skin detectors are challenging tasks because of multiple evaluation attributes and conflicting criteria. Although several evaluating and benchmarking techniques have been proposed, these approaches have many limitations. Fixing several attributes based on multi-attribute benchmarking approaches is particularly limited to reliable skin detection. Thus, this study aims to develop a new framework for evaluating and benchmarking skin detection on the basis of artificial intelligent models using multi-criteria analysis. For this purpose, two experiments are conducted. The first experiment consists of two stages: (1) discussing the development of a skin detector using multi-agent learning based on different color spaces to create a dataset of various color space samples for benchmarking and (2) discussing the evaluation and testing the developed skin detector according to multi-evaluation criteria (i.e. reliability, time complexity, and error rate within dataset) to create a decision matrix. The second experiment applies different decision-making techniques (AHP/SAW, AHP/MEW, AHP/HAW, AHP/TOPSIS, AHP/WSM, and AHP/WPM) to benchmark the results of the first experiment (i.e. the developed skin detector). Then, we discuss the use of the mean, standard deviation, and paired sample t-test to measure the correlations among the different techniques based on ranking results.

Quality control of fetal ultrasound images: Detection of abdomen anatomical landmarks using AdaBoost

A fetal ultrasound (US) biometry plane can be identified from the presence and absence of landmarks in the image. We propose an automated method of detecting two important anatomical landmarks (stomach bubble and umbilical vein) from the fetal ultrasound abdomen scan for the purpose of scoring the image quality. The implementation is based on the AdaBoost learning algorithm with an execution time less than 6 seconds. Evaluation performed on 2384 images shows detection of the stomach is more accurate compared to the umbilical vein and the approach worth further investigation for quality assurance framework.

Automated selection of standardized planes from ultrasound volume

The search for the standardized planes in a 3D ultrasound volume is a hard and time consuming process even for expert physicians. A scheme for finding the standardized planes would be beneficial in advancing the use of volumetric ultrasound for clinical diagnosis. In this paper, we propose a new method to automatically select the standard plane from the fetal ultrasound volume for the application of fetal biometry measurement. To our knowledge, this is the first study in the fetal ultrasound domain. The method is based on the AdaBoost learning algorithm and has been evaluated on a set of 30 volumes. The experimental results are promising with a recall rate of 91.29%. We believe this will increase the accuracy and efficiency in patient monitoring and care management in obstetrics, specifically in detecting growth restricted fetuses.

Development of an Off-Axis Digital Holographic Microscope for Large Scale Measurement in Fluid Mechanics

Holographic particle image velocimetry is a promising technique to probe and characterize complex flow dynamics since it is a truly three-dimensional (3D) three-component measurement technique. The technique simply records the coherent light scattered by small seeding particles that are assumed to faithfully follow the flow and uses it to reconstruct the event afterward. Reconstruction of the event is usually performed using a digital video microscope mounted on a 3D translation stage. The microscope records the intensity only which consequently results in loss of phase information. The objective of this paper is to develop and apply digital holographic microscopy with the aim to recover the phase information. Digital holographic microscopy has immense potentials in microscale solid and fluid measurements as it offers the possibility of digital wavefront processing by manipulating amplitude and phase of the recorded holograms. In this paper, we have developed an off-axis digital holographic microscope to capture both amplitude and phase of the reconstructed object simultaneously. This inherently solves twin image problem in the recorded digital holograms. The microscope was integrated into the reconstruction system and was successfully used to digitize holographic images of 10 μm polystyrene spheres and 1 μm olive oil droplets. The spatial resolution of the system is 0.63 μm, and the field of view is 1250 × 625 μm2. A 3D holographic reconstruction using a k-space analysis (wave-vector) of the optical field is applied to numerically refocus the images. Another potential application includes digital wavefront processing to compensate for aberration in the images.

Image Analysis Using Machine Learning: Anatomical Landmarks Detection in Fetal Ultrasound Images

Accurate and robust image analysis software is crucial for assessing the quality of ultrasound images of fetal biometry. In this work, we present the result of our automated image analysis method based on a machine learning algorithm in detecting important anatomical landmarks employed in manual scoring of ultrasound images of the fetal abdomen. Experimental results on 2384 images are promising and the clinical validation using 300 images demonstrates a high level agreement between the automated method and experts.

Astigmatism compensation in digital holographic microscopy using complex-amplitude correlation

Digital holographic microscopy (DHM) is a promising tool for a three-dimensional imaging of microscopic particles. It offers the possibility of wavefront processing by manipulating amplitude and phase of the recorded digital holograms. With a view to compensate for aberration in the reconstructed particle images, this paper discusses a new approach of aberration compensation based on complex amplitude correlation and the use of a priori information. The approach is applied to holograms of microscopic particles flowing inside a cylindrical micro-channel recorded using an off-axis digital holographic microscope. The approach results in improvements in the image and signal qualities.

Aberration compensation of holographic particle images using digital holographic microscopy

Characterisation of small and large-scale vortices in turbulent flows demands a system with high spatial resolution. The measurement of high spatial resolution, three-dimensional vector displacements in fluid mechanics using holography, is usually hampered by aberration. Aberration poses some problems in particle image identification due to low fidelity of real image reconstruction. Phase mismatch between the recording and the reconstruction waves was identified as the main source of aberration in this study. This paper demonstrates how aberration compensation can be achieved by cross-correlating the complex amplitude of an aberrated reconstructed object with the phase conjugate of a known reference object in the plane of the hologram (frequency space). Results favourably show significant increase in Strehl ratio and suppression of background noise that are more pronounced for particle images of 10 and 5 microns. It is clear from the work conducted that wavefront aberration measurement and compensation of holographic microscopic objects are now possible with the use of a variant digital holographic microscope.

Self-adapting approach in parameter tuning for differential evolution

Higher expectation has been requested from artificial intelligence (AI) owing to its success in various applications and domains. The use of AI is no longer limited to solve static optimization problems, but to perform well in dynamic optimization problems as well. The performance of AI in problem solving depends greatly on its own control parameters. The set of parameters which has been tuned to solve current optimization problem may not lead to the same performance if there is a shift or change in the optimization problem. To ensure its functionality in such condition, a machine learning needs to be able to self-determine its own control parameters. In short, a machine learning needs to be adaptive. Evolutionary algorithms (EAs) associated with adaptive ability turn out to be a potential solution under this condition. Therefore, our research focuses on the use of self-adaptive approach in parameter tuning in EAs, specifically in differential evolution (DE). Given that our proposed DE is no longer depending on a user to determine its control parameters, we are interested to know whether the self-adapting parameters will ensure good performance from DE or not. Two versions of DEs with the ability to self-adapt their parameters are developed. Most of DE related studies have suggested certain ranges of parameters to ensure appropriate operation of standard DE. In this research, we take the opportunity to confirm whether the ranges of self-adapting parameters fall within the suggested ranges or not. The experimental results have shown that both self-adapting DEs perform adequately well in 20 different benchmark problems without depending on user to determine the parameters explicitly. Besides that, it is interesting to find out the control parameters of the self-adapting DEs are not necessarily within the suggested ranges and they are still performing adequately well.