Search

Gastrointestinal (GI) diseases are prevalent medical conditions that require accurate and timely diagnosis for effective treatment. To address this, we developed the Multi-Fusion Convolutional Neural Network (MF-CNN), a deep learning framework that strategically integrates and adapts elements from six deep learning models, enhancing feature extraction and classification of GI diseases from endoscopic images. The MF-CNN architecture leverages truncated and partially frozen layers from existing models, augmented with novel components such as Auxiliary Fusing Layers (AuxFL), Fusion Residual Block (FuRB), and Alpha Dropouts (αDO) to improve precision and robustness. This design facilitates the precise identification of conditions such as ulcerative colitis, polyps, esophagitis, and healthy colons. Our methodology involved preprocessing endoscopic images sourced from open databases, including KVASIR and ETIS-Larib Polyp DB, using adaptive histogram equalization (AHE) to enhance their quality. The MF-CNN framework supports detailed feature mapping for improved interpretability of the model's internal workings. An ablation study was conducted to validate the contribution of each component, demonstrating that the integration of AuxFL, αDO, and FuRB played a crucial part in reducing overfitting and efficiency saturation and enhancing overall model performance. The MF-CNN demonstrated outstanding performance in terms of efficacy, achieving an accuracy rate of 99.25%. It also excelled in other key performance metrics with a precision of 99.27%, a recall of 99.25%, and an F1-score of 99.25%. These metrics confirmed the model's proficiency in accurate classification and its capability to minimize false positives and negatives across all tested GI disease categories. Furthermore, the AUC values were exceptional, averaging 1.00 for both test and validation sets, indicating perfect discriminative ability. The findings of the P-R curve analysis and confusion matrix further confirmed the robust classification performance of the MF-CNN. This research introduces a technique for medical imaging that can potentially transform diagnostics in gastrointestinal healthcare facilities worldwide. [ABSTRACT FROM AUTHOR]

A brain tumor develops as a result of uncontrolled and rapid cell proliferation. If not treated in its early stages, it might result in death. Despite several significant efforts and positive outcomes, accurate segmentation and classification remain challenging jobs. The variations in tumor size, shape, and location provide a substantial challenge for brain tumor diagnosis. Therefore, identifying brain tumors manually is challenging, time-consuming, and prone to mistakes. Consequently, there is now a need for high-accuracy automated computer-assisted diagnostics. This paper proposes a novel brain tumor detection method based on a machine learning classifier. Initially, the brain tumor images are collected from the “Magnetic Resonance Imaging (MRI)” database. In the preprocessing stage, anisotropic filtering and “Adaptive Histogram Equalization (AHE)” are performed to remove the noise and enhance the image contrast respectively. Then the images are segmented using “Enhanced Fruitfly Optimization-based Otsu segmentation (EFO-OTSU)”. The feature extraction is done using “Principal Component Analysis (PCA)” and “Discrete Wavelet Transform (DWT)”. We propose Boosted “Multi-Gradient Support Vector Machine (BMG-SVM)”to use the retrieved characteristics to divide the picture into the tumor and non-tumor sections. Further to enhance the classification performance, we employ the “Black Monkey Optimization (BMO)” algorithm. A few currently used approaches are contrasted with the simulation results of the suggested technique. The final findings show that the suggested technique outperforms the other methods in terms of effectiveness.

In modern culture, the rise in consumption of alcohol has caused many issues and its potentially adverse effects on human health are a well-known reality. The image of Iris helps to diagnose the alcohol drinkers efficiently. The prevailing methodology can take up plenty of time to execute the process, and may also produce less accuracy. This paper has proposed an effective method for the classification of alcohol drinkers as well as the identification of iris damaged levels utilizing MDLNN. Initially, the Log Gabor (LG), HOG feature, GLAC feature, LGXP, alongside canny edge detection (CED), features are extracted as of the alcohol drinker’s image. Next, the extracted features are selected utilizing BFO. Next, the image is classified as a more, medium, or less drunk with the help of MDLNN. Later on, the AHE is used to enhance the contrast of the drunken person’s iris image. Then, the image’s foreground is enhanced by utilizing HGBFDWT. Thereafter, segmentation is performed based upon the mask value utilizing OTMO. Finally, find what proportion (percentage) the iris is damaged based upon the Euclidean distance betwixt the original iris image of the drunk person and the segmented damaged level of the iris image. In an experimental assessment, the proposed work attains better accuracy than the prevailing methodologies.

In modern culture, the rise in consumption of alcohol has caused many issues and its potentially adverse effects on human health are a well-known reality. The image of Iris helps to diagnose the alcohol drinkers efficiently. The prevailing methodology can take up plenty of time to execute the process, and may also produce less accuracy. This paper has proposed an effective method for the classification of alcohol drinkers as well as the identification of iris damaged levels utilizing MDLNN. Initially, the Log Gabor (LG), HOG feature, GLAC feature, LGXP, alongside canny edge detection (CED), features are extracted as of the alcohol drinker's image. Next, the extracted features are selected utilizing BFO. Next, the image is classified as a more, medium, or less drunk with the help of MDLNN. Later on, the AHE is used to enhance the contrast of the drunken person's iris image. Then, the image's foreground is enhanced by utilizing HGBFDWT. Thereafter, segmentation is performed based upon the mask value utilizing OTMO. Finally, find what proportion (percentage) the iris is damaged based upon the Euclidean distance betwixt the original iris image of the drunk person and the segmented damaged level of the iris image. In an experimental assessment, the proposed work attains better accuracy than the prevailing methodologies. [ABSTRACT FROM AUTHOR]

The present work aimed to evaluate the reproducibility of radiomics features derived from manual delineation and semiautomatic segmentation after enhancement using the Contrast Limited Adaptive Histogram Equalization (CLAHE) and Adaptive Histogram Equalization (AHE) techniques on a benign tumor of two-dimensional (2D) mammography images. Thirty mammogram images with known benign tumors were obtained from The Cancer Imaging Archive (TCIA) datasets and were randomly selected as subjects. The samples were enhanced for semiautomatic segmentation sets using the Active Contour Model in MATLAB 2019a before analysis by two independent observers. Meanwhile, the images without any enhancement were segmented manually. The samples were divided into three categories: (1) CLAHE images, (2) AHE images, and (3) manual segmented images. Radiomics features were extracted using algorithms provided by MATLAB 2019a software and were assessed with a reliable intra-class correlation coefficient (ICC) score. Radiomics features for the CLAHE group (ICC = 0.890 ± 0.554, p 0.05). Features in all three categories were more robust for the CLAHE compared to the AHE and manual groups. This study shows the existence in variation for the radiomics features extracted from tumor region that are segmented using various image enhancement techniques. Semiautomatic segmentation with image enhancement using CLAHE algorithm gave the best result and was a better alternative than manual delineation as the first two techniques yielded reproducible descriptors. This method should be applicable for predicting outcomes in patient with breast cancer.

The research work proposes a novel triangular fuzzy membership (TFM) function-based contrast limited adaptive histogram equalization (CLAHE) for biometric image enhancement. Biometric images have wide applications in the areas of verification and authentication systems. For accurate identification and verification, pre-processing of captured biometric images becomes essential. When the region of interest is smaller than the original image, a variation of histogram equalization called adaptive histogram equalization (AHE) is used. AHE enhances contrast of images by considering local regions. Along with local contrast, noise in those regions also get amplified by using AHE. This amplification of noise can be resolved by applying a contrast limited AHE (CLAHE) which limits the contrast in the enhanced local regions by clipping the histogram at a pre-fixed limit. CLAHE yields good results by limiting the contrast and enhancing local regions, but it is image invariant since it uses pre-determined clip limit for limiting contrast. The proposed research work TFM-CLAHE puts forward the idea of image variant, automatic clip value determinant algorithm for enhancement. The algorithm employs triangular fuzzy membership function to determine clip-limit and limits contrast by clipping the histogram at the computed clip-level. TFM function computes the clipping parameter by considering intensities of pixels. The computed fuzzy clip-limit overrides the pre-defined limit. Consequently, the clipping parameter varies according to the image under consideration and yields better enhancement results. The proposed work is experimented on multimodal biometric images acquired from Chinese Academy of Science, Institute of Automation Iris, Face and Fingerprint databases. TFM-CLAHE computes appropriate clipping limit for each of these heterogenous images. The results of the proposed work are evaluated on the grounds of images' average information content, mean square error, peak signal noise ratio, natural image quality evaluator, no-reference free energy based robust metric, blind image quality measure of enhanced images and no reference quality metric for contrast distortion. The results show good enhancement and these are compared with existing conventional image enhancement techniques. [ABSTRACT FROM AUTHOR]