Papers by Carlos M. Travieso
Melanoma diagnosis depends on the experience of doctors. Symmetry is one of the most important fa... more Melanoma diagnosis depends on the experience of doctors. Symmetry is one of the most important factors to measure, since asymmetry shows an uncontrolled growth of cells, leading to melanoma cancer. A system for melanoma detection in diagnosing melanocytic diseases with high sensitivity is proposed here. Two different sets of features are extracted based on the importance of the ABCD rule and symmetry evaluation to develop a new architecture. Support Vector Machines are used to classify the extracted sets by using both an alternative labeling method and a structure divided into two different classifiers which prioritize sensitivity. Although feature extraction is based on former works, the novelty lies in the importance given to symmetry and the proposed architecture, which combines two different feature sets to obtain a high sensitivity, prioritizing the medical aspect of diagnosis. In particular, a database provided by Hospital Universitario de Gran Canaria Doctor Negrín was tested, obtaining a sensitivity of 100% and a specificity of 66.66% using a leave-one-out validation method. These results show that 66.66% of biopsies would be avoided if this system is applied to lesions which are difficult to classify by doctors.
IntechOpen eBooks, Jun 19, 2024
Supervised learning requires labeled data to train models and then make predictions from new inpu... more Supervised learning requires labeled data to train models and then make predictions from new input data. Deep Learning (DL) methods require immense amounts of training data and processing power to provide reasonable results. In computer vision applications, and more specifically in despeckling SAR (Synthetic Aperture Radar) images, due to the speckle content, there is no ground truth available. To test the performances of despeckling filters, the common approach is tocorrupt synthetic images with a suitable speckle model and then, after filtering, well-known metrics are obtained. Then, filters are tested on actual SAR data, and specific metrics for SAR are evaluated. However, even the most elaborated speckle models are far from accounting for the complex mechanisms related to SAR images. In this paper, a methodology to design a realistic dataset to overcome these limitations is proposed. Actual SAR images of the same scene but acquired with the same sensor on different dates are dow...
Apples are one of the most common fruit on the planet. It is rich in iron, fiber, antioxidants an... more Apples are one of the most common fruit on the planet. It is rich in iron, fiber, antioxidants and other nutritive quality; which are incredibly important for human body and brain. The quality of an apple gets affected once they are chopped. This paper presents a non-destructive image processing based algorithm that identifies the presence of enzymatic browning in chopped apples for the determination of its nutrients loss. The proposed imperative assemblage of this image processing algorithm makes it flexible, automatic and non-destructive. The quantification of enzymatic browning in chopped apples has been obtained with high precision using this proposed imaging based method. The machine learning based on strategic selection of discriminatory statistical features of chopped apples extracted in wavelet domain makes it a novel approach. 85% of accuracy has been achieved by using machine learning based Support Vector Machine (SVM) classifier.
The present work presents a biometric identifier system using the combination of two different fe... more The present work presents a biometric identifier system using the combination of two different features: hands shape (finger lengths and width) and hand-written signature contour. Signature database contains 300 different signers with 24 signatures and the hand database has 144 owners with 10 images. The study covers three different classifiers: Hidden Markov Models (HMM), Support Vector Machines (SVM) and a combination of both using the Fisher Kernel. Systems are evaluated separately and in conjunction, giving in each case 100% of identification success rate for the combined classifier. The combination of features gives better results when reducing the training set than the independent systems.
Biomedical Signal Processing and Control, Mar 1, 2016
The Optic disc (OD) nerve head region form basis for study and analysis of various eye pathologie... more The Optic disc (OD) nerve head region form basis for study and analysis of various eye pathologies. The shape, contour and size of OD is vital in classification and grading of retinal diseases like glaucoma. There is a need to develop fast and efficient algorithms for large scale retinal disease screening. With this in mind, this paper present a novel framework for fast and fully automatic detection of OD and its accurate segmentation in digital fundus images. The methodology involves optic disc centre localization followed by removal of vascular structure by accurate inpainting of blood vessels in the optic disc region. An adaptive threshold based Region Growing technique is then employed for reliable segmentation of fundus images. The proposed technique achieved significant results when tested on standard test databases like MESSIDOR and DRIVE with average overlapping ratio of 89% and 87% respectively. Validation experiments were done on a labeled dataset containing healthy and pathological images obtained from a local eye hospital achieving an appreciable 91% average OD segmentation accuracy
International Journal for Population Data Science, Aug 30, 2018
Take-down policy If you believe that this document breaches copyright please contact us providing... more Take-down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
This paper proposes a method of zero-watermarking scheme to generate a unique digital identificat... more This paper proposes a method of zero-watermarking scheme to generate a unique digital identification tag having patient's identity without tampering the details of the medical information present in the image. Using the features of the original medical image, the digital identification tag is encrypted and a unique master key is created which can be decrypted by using the original medical image at the receiver's end. This is performed by using strategic decomposition of digital wavelet transform along with singular value decomposition to obtain a unique identity vector which is used in a function along with the image's watermark to create the master share for that image. Images of the fundus comprise of very fine features like macula, optic disks and blood vessels. Traditional watermarking techniques might tamper with the crucial information stored in the medical image. Even the smallest of changes can result in noise and distortion of the fine features of the medical image and can affect the final decision which would be made by the doctor. The method proposed in this paper does not tamper with any details of the image in order of watermarking the digital identification tags onto the image.
The quality of the fish is mainly altered by different cooling methods, exporting, handling etc. ... more The quality of the fish is mainly altered by different cooling methods, exporting, handling etc. In this paper a nondestructive framework is proposed for identification of fish freshness using image processing techniques. In this paper skin tissue is selected as focal tissue for basic analysis and identification of freshness of fish from fish images. Statistical features are extracted in the HSV color space which gives degradation pattern for fish freshness which is used to design the framework for identification of fish freshness. The experiment result indicates monotonic degradation pattern. Experiments were carried on fish images and results are encouraging. The maximum classification accuracy of contributed methodology is 96.66%.
Neural Computing and Applications, Oct 19, 2019
Cardiovascular diseases are one of the most fatal diseases across the globe. Clinically, conventi... more Cardiovascular diseases are one of the most fatal diseases across the globe. Clinically, conventional stethoscope is used to check the medical condition of a human heart. Only a trained medical professional can understand and interpret the heart auscultations clinically. This paper presents a machine learning-based automatic classification system based on heart sounds to diagnose cardiac disorders. The proposed framework involves strategic processing and framing of heart sound to extract discriminatory features for machine learning. The most prominent features are selected and used to train a supervised classifier for automatic detection of cardiac diseases. The biological abnormalities disturbing the physical functioning of the heart cause variations in the auscultations, which is strategically used in terms of some discriminatory features for machine learning-based automatic classification. The proposed method achieved 97.78% accuracy with the equal error rate of 2.22% for abnormal and normal heart sound classification. The experimental results exhibit that the performance of the proposed method in proper diagnosis of the cardiac diseases is high in terms of accuracy and has low error rate which makes the proposed algorithm suitable for real-time applications.
JMIR mental health, Jun 22, 2018
Tecnología en Marcha, Nov 16, 2022
Many countries are struggling for COVID-19 screening resources which arises the need for automati... more Many countries are struggling for COVID-19 screening resources which arises the need for automatic and low-cost diagnosis systems which can help to diagnose and a large number of tests can be conducted rapidly. Instead of relying on one single method, artificial intelligence and multiple sensors based approaches can be used to decide the prediction of the health condition of the patient. Temperature, oxygen saturation level, chest X-ray and cough sound can be analyzed for the rapid screening. The multi-sensor approach is more reliable and a person can be analyzed in multiple feature dimensions. Deep learning models can be trained with multiple chest x-ray images belonging to different categories to different health conditions i.e. healthy, COVID-19 positive, pneumonia, tuberculosis, etc. The deep learning model will extract the features from the input images and based on that test images will be classified into different categories. Similarly, cough sound and short talk can be trained on a convolutional neural network and after proper training, input voice samples can be differentiated into different categories. Artificial based approaches can help to develop a system to work efficiently at a low cost. Palabras clave Red neuronal convolucional; detección COVID-19; aprendizaje profundo; sensor múltiple. Resumen Muchos países están luchando por los recursos de detección de COVID-19, lo que plantea la necesidad de sistemas de diagnóstico automáticos y de bajo costo que puedan ayudar a diagnosticar y que se pueda realizar una gran cantidad de pruebas rápidamente. En lugar de depender de un solo método, se pueden utilizar la inteligencia artificial y enfoques basados en múltiples sensores para decidir la predicción del estado de salud del paciente. La temperatura, el nivel de saturación de oxígeno, la radiografía de tórax y el sonido de la tos se pueden analizar para la detección rápida. El enfoque de múltiples sensores es más confiable y una persona puede ser analizada en múltiples dimensiones de características. Los modelos de aprendizaje profundo se pueden entrenar con múltiples imágenes de rayos X de tórax que pertenecen a diferentes categorías para diferentes condiciones de salud, es decir, saludable, COVID-19 positivo, neumonía, tuberculosis, etc. El modelo de aprendizaje profundo extraerá las características de las imágenes de entrada y en base a eso, las imágenes de prueba se clasificarán en diferentes categorías. De manera similar, el sonido de la tos y la conversación corta se pueden entrenar en una red neuronal convolucional y, después de un entrenamiento adecuado, las muestras de voz de entrada se pueden diferenciar en diferentes categorías. Los enfoques basados en materiales artificiales pueden ayudar a desarrollar un sistema que funcione de manera eficiente a bajo costo.
2017 4th IEEE Uttar Pradesh Section International Conference on Electrical, Computer and Electronics (UPCON), Oct 1, 2017
Objectives: White matter (WM) impairments involving both motor and extra-motor areas have been we... more Objectives: White matter (WM) impairments involving both motor and extra-motor areas have been well-documented in amyotrophic lateral sclerosis (ALS). This study tested the potential of diffusion measurements in WM for identifying ALS based on support vector machine (SVM). Methods: Voxel-wise fractional anisotropy (FA) values of diffusion tensor images (DTI) were extracted from 22 ALS patients and 26 healthy controls and served as discrimination features. The revised ALS Functional Rating Scale (ALSFRS-R) was employed to assess ALS severity. Feature ranking and selection were based on Fisher scores. A linear kernel SVM algorithm was applied to build the classification model, from which the classification performance was evaluated. To promote classifier generalization ability, a leave-one-out cross-validation (LOOCV) method was adopted. Results: By using the 2,400∼3,400 ranked features as optimal features, the highest classification accuracy of 83.33% (sensitivity = 77.27% and specificity = 88.46%, P = 0.0001) was achieved, with an area under receiver operating characteristic curve of 0.862. The predicted function value was positively correlated with patient ALSFRS-R scores (r = 0.493, P = 0.020). In the optimized SVM model, FA values from several regions mostly contributed to classification, primarily involving the corticospinal tract pathway, postcentral gyrus, and frontal and parietal areas. Conclusions: Our results suggest the feasibility of ALS diagnosis based on SVM analysis and diffusion measurements of WM. Additional investigations using a larger cohort is recommended in order to validate the results of this study.
Computing, Jun 28, 2018
AbstractThere are different reasons like pests, weeds, and diseases which are responsible for the... more AbstractThere are different reasons like pests, weeds, and diseases which are responsible for the loss of crop production. Identification and detection of different plant diseases is a difficult task in a large crop field and it also requires an expert manpower. In this paper, the proposed method uses adaptive intensity based thresholding for automatic segmentation of powdery mildew disease which makes this method invariant to image quality and noise. After the segmentation of powdery mildew disease from leaf images, the affected area is quantified which makes this method efficient for grading the level of disease infection. The proposed method is tested on the comprehensive dataset of leaf images of cherry crops, which achieved good accuracy of 99%. The experimental results indicate that proposed method for segmentation of powdery mildew disease affected area from leaf image of cherry crops is convincing and computationally cheap.
Exudates are one of the abnormalities present in the eye which can lead to vision loss. Many fund... more Exudates are one of the abnormalities present in the eye which can lead to vision loss. Many fundus images consist of artifacts which occur during acquisition and hamper the accuracy of exudates detection. There is still a need to
This work shows an approach for the writer identification using off-line handwritten writing. The... more This work shows an approach for the writer identification using off-line handwritten writing. The contour of handwritten words is calculated. That feature is transformed by a Hidden Markov Model, in order to get a hyperdimensionality of new features. Later, a support vector machine classifier is applied to analyze the grade of discrimination. The approach reaches an accuracy between 98,60% and 100%, according to the type of word. This proposal presents a robust and novel system for the based-writing identification in comparison versus the state-of-the-art; and shows the study of its efficiency according to different off-line handwritten words.
This work presents a biometric approach for spider identification based on transform domain and S... more This work presents a biometric approach for spider identification based on transform domain and Support Vector Machines as classifier. The dataset is composed by 185 images of spider web. The goal of this work is to use the structure of spider web for identifying the kind of spider. The experiments were done using two different of segmentation blocks and the analysis of the whole and center of the spider web. The best accuracy is reached after to run the different combinations.
With the widespread of Monkeypox and increase in the weekly reported number of cases, it is obser... more With the widespread of Monkeypox and increase in the weekly reported number of cases, it is observed that this outbreak continues to put the human beings in risk. The early detection and reporting of this disease will help monitoring and controlling the spread of it and hence, supporting international coordination for the same. For this purpose, the aim of this paper is to classify three diseases viz. Monkeypox, Chikenpox and Measles based on provided image dataset using trained standalone DL models (InceptionV3, EfficientNet, VGG16) and Squeeze and Excitation Network (SENet) Attention model. The first step to implement this approach is to search, collect and aggregate (if require) verified existing dataset(s). To the best of our knowledge, this is the first paper which has proposed the use of SENet based attention models in the classification task of Monkeypox and also targets to aggregate two different datasets from distinct sources in order to improve the performance parameters. ...
Uploads
Papers by Carlos M. Travieso