Objective: Segmentation of skin lesions is an important step in the automated computer aided diagnosis melanoma. However, existing segmentation methods have a tendency to over- or under-segment and perform poorly when fuzzy boundaries, low contrast with background, inhomogeneous textures, contain artifacts. Furthermore, performance these are heavily reliant on appropriate tuning large number parameters as well use effective preprocessing techniques, such illumination correction hair removal....

Malignant melanoma has one of the most rapidly increasing incidences in world and a considerable mortality rate. Early diagnosis is particularly important since can be cured with prompt excision. Dermoscopy images play an role non-invasive early detection [1]. However, using human vision alone subjective, inaccurate poorly reproducible even among experienced dermatologists. This attributed to challenges interpreting diverse characteristics including lesions varying sizes shapes, that may...

The segmentation of skin lesions in dermoscopic images is a fundamental step automated computer-aided diagnosis melanoma. Conventional methods, however, have difficulties when the lesion borders are indistinct and contrast between surrounding low. They also perform poorly there heterogeneous background or that touches image boundaries; this then results underand oversegmentation lesion. We suggest saliency detection using reconstruction errors derived from sparse representation model coupled...

Segmentation of skin lesions is considered as an important step in computer aided diagnosis (CAD) for automated melanoma diagnosis. Existing methods however have problems with over-or under-segmentation and do not perform well when a lesion partially connected to the background or image contrast low. To overcome these limitations, we propose new segmentation method via image-wise supervised learning (ISL) multi-scale superpixel based cellular automata (MSCA). We using ISL derive...

Dermoscopy image as a non-invasive diagnosis technique plays an important role for early of malignant melanoma. Even experienced dermatologists, however, by human vision can be subjective, inaccurate and non-reproducible. This is attributed to the challenging characteristics including varying lesion sizes their shapes, fuzzy boundaries, different skin color types presence hair. To aid in interpretation, automatic classification dermoscopy images have been shown valuable clinical decision...

The accuracy and robustness of image classification with supervised deep learning are dependent on the availability large-scale labelled training data. In medical imaging, these large datasets sparse, mainly related to complexity in manual annotation. Deep convolutional neural networks (CNNs), transferable knowledge, have been employed as a solution limited annotated data through: 1) fine-tuning generic knowledge relatively smaller amount imaging data, 2) representation that is invariant...

Recent advances in supervised deep learning methods are enabling remote measurements of photoplethysmography-based physiological signals using facial videos. The performance these methods, however, dependent on the availability large labelled data. Contrastive as a self-supervised method has recently achieved state-of-the-art performances representative data features by maximising mutual information between different augmented views. However, existing augmentation techniques for contrastive...

The segmentation of skin lesions in dermoscopic images is considered as one the most important steps computer-aided diagnosis (CAD) for automated melanoma diagnosis. Existing methods, however, have problems with over-segmentation and do not perform well when contrast between lesion its surrounding low. Hence, this study, we propose a new saliency-based (SSLS) that designed to exploit inherent properties images, which focal central region subtle discrimination regions. proposed method was...

The accuracy and robustness of image classification with supervised deep learning are dependent on the availability large-scale, annotated training data. However, there is a paucity data available due to complexity manual annotation. To overcome this problem, popular approach use transferable knowledge across different domains by: 1) using generic feature extractor that has been pre-trained large-scale general images (i.e., transfer-learned) but which not suited capture characteristics from...

Modern whole slide imaging technique together with supervised deep learning approaches have been advancing the field of histopathology, enabling accurate analysis tissues. These use images (WSIs) at various resolutions, utilising low-resolution WSIs to identify regions interest in tissue and high-resolution for detailed cellular structures. Due labour-intensive process annotating gigapixels WSIs, remains challenging approaches. Self-supervised (SSL) has emerged as an approach build efficient...

The classification of medical images is a critical step for imaging-based clinical decision support systems. Existing methods X-ray images, however, generally represent the image using only local texture or generic features (e.g. color shape) derived from predefined feature spaces. This limits ability to quantify characteristics general data-derived learned datasets. In this study we present new algorithm improve performance classification, where propose late-fusion domain transferred...

Segmentation of skin lesions is considered as an important step in computer aided diagnosis (CAD) for melanoma diagnosis. There have many attempts to segment a semi- or fully-automated manner. Existing methods, however, problems with over- under-segmentation and do not perform well challenging such when lesion partially connected the background image contrast low. To overcome these limitations, we propose new semi-automated segmentation method that incorporates fully convolutional networks...

High-resolution (HR) magnetic resonance imaging is essential in aiding doctors their diagnoses and image-guided treatments. However, acquiring HR images can be time-consuming costly. Consequently, deep learning-based super-resolution reconstruction (SRR) has emerged as a promising solution for generating (SR) from low-resolution (LR) images. Unfortunately, training such neural networks requires aligned authentic LR image pairs, which are challenging to obtain due patient movements during...

Dynamic medical images are often limited in its application due to the large radiation doses and longer image scanning reconstruction times. Existing methods attempt reduce volume samples dynamic sequence by interpolating volumes between acquired samples. However, these either 2D and/or unable support but periodic variations functional motion In this paper, we present a spatiotemporal volumetric interpolation network (SVIN) designed for 4D images. SVIN introduces dual networks: first is that...

Outbreaks of infectious diseases pose great risks, including hospitalization and death, to public health. Therefore, improving the management outbreaks is important for preventing widespread infection mitigating associated risks. Mobile health technology provides new capabilities that can help better capture, monitor, manage diseases, ability quickly identify potential outbreaks.This study aims develop a disease surveillance (IDS) system comprising mobile app accurate data capturing...

Medical image analysis using supervised deep learning methods remains problematic because of the reliance on large amounts labelled training data. Although medical imaging data repositories continue to expand there has not been a commensurate increase in amount annotated Hence, we propose new unsupervised feature method that learns representations then differentiate dissimilar images an ensemble different convolutional neural networks (CNNs) and K-means clustering. It jointly clustering...