Adaptive clip-limit-based bi-histogram equalization algorithm for infrared image enhancement.

Infrared (IR) images are basically low-contrast in nature; hence, it is essential to enhance the contrast of IR images to facilitate real-life applications. This work proposes a novel adaptive clip-limit-oriented bi-histogram equalization (bi-HE) method for enhancing IR images. HE methods are simple in implementation but often cause over-enhancement due to the presence of long spikes. To reduce long spikes, this work suggests to apply a log-power operation on the histogram, where the log operation reduces the long spikes, and power transformation regains the shape of the histogram. First, a histogram separation point is generated applying the mean of the multi-peaks of the input histogram. After that, an alteration in the input histogram is done using the log-power process. Subsequently, a clipping operation on the altered histogram followed by redistribution of the clipped portion is performed to restrict over-enhancement. Next, the modified histogram is sub-divided using the histogram separation point. Finally, the modified sub-histograms are equalized independently. Simulation results show that the suggested method effectively improves the contrast of IR images. Visual quality evaluations and quantitative assessment demonstrate that the suggested method outperforms the state-of-the-art algorithms.

Detection of malignant melanoma in H&E-stained images using deep learning techniques.

Histopathological images are widely used to diagnose diseases including skin cancer. As digital histopathological images are typically of very large size, in the order of several billion pixels, automated identification of all abnormal cell nuclei and their distribution within multiple tissue sections would assist rapid comprehensive diagnostic assessment. In this paper, we propose a deep learning-based technique to segment the melanoma regions in Hematoxylin and Eosin (H&E) stained histopathological images. In this technique, the nuclei in the image are first segmented using a Convolutional Neural Network (CNN). The segmented nuclei are then used to generate melanoma region masks. Experimental results with a small melanoma dataset show that the proposed method can potentially segment the nuclei with more than 94 % accuracy and segment the melanoma regions with a Dice coefficient of around 85 %. The proposed technique also has a small execution time making it suitable for clinical diagnosis with a fast turnaround time.

An efficient CNN based algorithm for detecting melanoma cancer regions in H&E-stained images.

Histopathological images are widely used to diagnose diseases such as skin cancer. As digital histopathological images are typically of very large size, in the order of several billion pixels, automated identification of abnormal cell nuclei and their distribution within multiple tissue sections would enable rapid comprehensive diagnostic assessment. In this paper, we propose a deep learning-based technique to segment the melanoma regions in Hematoxylin and Eosin-stained histopathological images. In this technique, the nuclei in an image are first segmented using a deep learning neural network. The segmented nuclei are then used to generate the melanoma region masks. Experimental results show that the proposed method can provide nuclei segmentation accuracy of around 90% and the melanoma region segmentation accuracy of around 98%. The proposed technique also has a low computational complexity.

Automated proliferation index calculation for skin melanoma biopsy images using machine learning.

The Proliferation Index (PI) is an important diagnostic, predictive and prognostic parameter used for evaluating different types of cancer. This paper presents an automated technique to measure the PI values for skin melanoma images using machine learning algorithms. The proposed technique first analyzes a Mart-1 stained histology image and generates a region of interest (ROI) mask for the tumor. The ROI mask is then used to locate the tumor regions in the corresponding Ki-67 stained image. The nuclei in the Ki-67 ROI are then segmented and classified using a Convolutional Neural Network (CNN), and the PI value is calculated based on the number of the active and the passive nuclei. Experimental results show that the proposed technique can robustly segment (with 94 % accuracy) and classify the nuclei with a low computational complexity and the calculated PI values have less than 4 % average error.

Automatic Detection of Retinal Lesions for Screening of Diabetic Retinopathy.

OBJECTIVE: Diabetic retinopathy (DR) is characterized by the progressive deterioration of retina with the appearance of different types of lesions that include microaneurysms, hemorrhages, exudates, etc. Detection of these lesions plays a significant role for early diagnosis of DR. METHODS: To this aim, this paper proposes a novel and automated lesion detection scheme, which consists of the four main steps: vessel extraction and optic disc removal, preprocessing, candidate lesion detection, and postprocessing. The optic disc and the blood vessels are suppressed first to facilitate further processing. Curvelet-based edge enhancement is done to separate out the dark lesions from the poorly illuminated retinal background, while the contrast between the bright lesions and the background is enhanced through an optimally designed wideband bandpass filter. The mutual information of the maximum matched filter response and the maximum Laplacian of Gaussian response are then jointly maximized. Differential evolution algorithm is used to determine the optimal values for the parameters of the fuzzy functions that determine the thresholds of segmenting the candidate regions. Morphology-based postprocessing is finally applied to exclude the falsely detected candidate pixels. RESULTS AND CONCLUSIONS: Extensive simulations on different publicly available databases highlight an improved performance over the existing methods with an average accuracy of and robustness in detecting the various types of DR lesions irrespective of their intrinsic properties.

Pre-cancer risk assessment in habitual smokers from DIC images of oral exfoliative cells using active contour and SVM analysis.

Habitual smokers are known to be at higher risk for developing oral cancer, which is increasing at an alarming rate globally. Conventionally, oral cancer is associated with high mortality rates, although recent reports show the improved survival outcomes by early diagnosis of disease. An effective prediction system which will enable to identify the probability of cancer development amongst the habitual smokers, is thus expected to benefit sizable number of populations. Present work describes a non-invasive, integrated method for early detection of cellular abnormalities based on analysis of different cyto-morphological features of exfoliative oral epithelial cells. Differential interference contrast (DIC) microscopy provides a potential optical tool as this mode provides a pseudo three dimensional (3-D) image with detailed morphological and textural features obtained from noninvasive, label free epithelial cells. For segmentation of DIC images, gradient vector flow snake model active contour process has been adopted. To evaluate cellular abnormalities amongst habitual smokers, the selected morphological and textural features of epithelial cells are compared with the non-smoker (-ve control group) group and clinically diagnosed pre-cancer patients (+ve control group) using support vector machine (SVM) classifier. Accuracy of the developed SVM based classification has been found to be 86% with 80% sensitivity and 89% specificity in classifying the features from the volunteers having smoking habit.

Retinal blood vessel extraction using tunable bandpass filter and fuzzy conditional entropy.

BACKGROUND AND OBJECTIVES: Extraction of blood vessels on retinal images plays a significant role for screening of different opthalmologic diseases. However, accurate extraction of the entire and individual type of vessel silhouette from the noisy images with poorly illuminated background is a complicated task. To this aim, an integrated system design platform is suggested in this work for vessel extraction using a sequential bandpass filter followed by fuzzy conditional entropy maximization on matched filter response. METHODS: At first noise is eliminated from the image under consideration through curvelet based denoising. To include the fine details and the relatively less thick vessel structures, the image is passed through a bank of sequential bandpass filter structure optimized for contrast enhancement. Fuzzy conditional entropy on matched filter response is then maximized to find the set of multiple optimal thresholds to extract the different types of vessel silhouettes from the background. Differential Evolution algorithm is used to determine the optimal gain in bandpass filter and the combination of the fuzzy parameters. Using the multiple thresholds, retinal image is classified as the thick, the medium and the thin vessels including neovascularization. RESULTS: Performance evaluated on different publicly available retinal image databases shows that the proposed method is very efficient in identifying the diverse types of vessels. Proposed method is also efficient in extracting the abnormal and the thin blood vessels in pathological retinal images. The average values of true positive rate, false positive rate and accuracy offered by the method is 76.32%, 1.99% and 96.28%, respectively for the DRIVE database and 72.82%, 2.6% and 96.16%, respectively for the STARE database. Simulation results demonstrate that the proposed method outperforms the existing methods in detecting the various types of vessels and the neovascularization structures. CONCLUSIONS: The combination of curvelet transform and tunable bandpass filter is found to be very much effective in edge enhancement whereas fuzzy conditional entropy efficiently distinguishes vessels of different widths.

Blood vessel extraction and optic disc removal using curvelet transform and kernel fuzzy c-means.

This paper proposes an automatic blood vessel extraction method on retinal images using matched filtering in an integrated system design platform that involves curvelet transform and kernel based fuzzy c-means. Since curvelet transform represents the lines, the edges and the curvatures very well and in compact form (by less number of coefficients) compared to other multi-resolution techniques, this paper uses curvelet transform for enhancement of the retinal vasculature. Matched filtering is then used to intensify the blood vessels' response which is further employed by kernel based fuzzy c-means algorithm that extracts the vessel silhouette from the background through non-linear mapping. For pathological images, in addition to matched filtering, Laplacian of Gaussian filter is also employed to distinguish the step and the ramp like signal from that of vessel structure. To test the efficacy of the proposed method, the algorithm has also been applied to images in presence of additive white Gaussian noise where the curvelet transform has been used for image denoising. Performance is evaluated on publicly available DRIVE, STARE and DIARETDB1 databases and is compared with the large number of existing blood vessel extraction methodologies. Simulation results demonstrate that the proposed method is very much efficient in detecting the long and the thick as well as the short and the thin vessels with an average accuracy of 96.16% for the DRIVE and 97.35% for the STARE database wherein the existing methods fail to extract the tiny and the thin vessels.