Publications by authors named "Hernan Dario Vargas Cardona"
Article Synopsis
- * The CITOBOT-COL project focuses on creating CITOBOT, a portable cervical cancer screening device, through a human-centered design approach and the use of AI, with four design iterations leading to CITOBOT v4.
- * The device's prototypes were tested by experts and validated through focus groups, emphasizing high-quality image capture, AI algorithms for image classification, and risk assessment to enhance cervical cancer screening in resource-limited communities.
Article Synopsis
- - The integration of artificial intelligence (AI) in cancer research is growing, particularly in analyzing diagnostic images for cervical cancer, with a review of literature conducted under systematic guidelines.
- - A comprehensive search identified 32 studies from 2009 to 2022, highlighting various AI algorithms like support vector machines, deep learning methods, and their performance in diagnosing cervical cancer from sources such as digital colposcopy and cervicography.
- - Results indicated that deep learning techniques, especially convolutional neural networks, achieved over 97% accuracy in diagnosis, suggesting promising advancements, yet emphasizing the need for more research to confirm these outcomes.
Global cervical cancer incidence and mortality have remained a major public health problem. Depending on the quality and coverage of preventive programs, and the capacity of health care systems, different screening tests are used, with the Pap smear being the most widely implemented. Several difficulties have been reported in accessing timely detection, causing late cervical cancer diagnosis.
View Article and Find Full Text PDFMulti-output Gaussian processes for enhancing resolution of diffusion tensor fields.
Annu Int Conf IEEE Eng Med Biol Soc
August 2016
Article Synopsis
- - Second order diffusion tensor (DT) fields are crucial in clinical fields like brain fiber mapping and diagnosing neuro-degenerative diseases, but low spatial resolution in MRI leads to challenges in accurately capturing tissue structures.
- - The study introduces a novel feature-based interpolation method using multi-output Gaussian processes (MOGP) to enhance the spatial resolution of DT fields by treating eigenvalues and Euler angles of diffusion tensors as separate but connected outputs.
- - This MOGP method outperforms existing techniques for DT interpolation in terms of accuracy and effectively maintains important characteristics of diffusion tensors, showing performance comparable to advanced methods like Generalized Wishart processes.
Sparse representation of MER signals for localizing the Subthalamic Nucleus in Parkinson's disease surgery.
Annu Int Conf IEEE Eng Med Biol Soc
October 2016
Article Synopsis
- * The study demonstrates that analyzing microelectrode recording (MER) signals with sparse representation techniques improves the accuracy of STN identification.
- * Three methods—Method of Frames (MOF), Best Orthogonal Basis (BOB), and Basis Pursuit (BP)—show superior performance compared to traditional signal processing methods, achieving over 98% classification accuracy in tests.
Article Synopsis
- This paper discusses a method for automatically identifying biomedical signals, specifically Microelectrode Recordings (MER) and Electrocardiography (ECG) signals, using unsupervised learning techniques.
- The approach combines classic and Bayesian estimation theories, employing Gaussian mixture models with two estimation methods: the Expectation-Maximization (EM) algorithm and variational inference.
- The results demonstrate an accuracy of over 85% for MER and 90% for ECG in classifying these signals, outperforming traditional classifiers like Naive Bayes and K-nearest neighbor.
Multi-patient learning increases accuracy for Subthalamic Nucleus identification in deep brain stimulation.
Annu Int Conf IEEE Eng Med Biol Soc
August 2013
Article Synopsis
- * Recent efforts are focusing on developing automatic systems that use supervised learning to accurately localize brain regions across different patients.
- * This study demonstrates that utilizing multi-task learning to share information between patients enhances accuracy in targeting the Subthalamic Nucleus, outperforming traditional methods in real datasets.
NEUROZONE: on-line recognition of brain structures in stereotactic surgery--application to Parkinson's disease.
Annu Int Conf IEEE Eng Med Biol Soc
August 2013
Article Synopsis
- The paper introduces NEUROZONE, a software system designed to assist in the precise positioning of microelectrodes during Deep Brain Stimulation surgeries by analyzing microelectrode recordings (MER) for real-time brain structure recognition.
- NEUROZONE includes features for offline database processing and classifier training, enhancing the automatic identification of brain target areas while aiding medical specialists in reducing potential side effects from misidentification.
- The software has been successfully tested at the Institute for Epilepsy and Parkinson of the Eje Cafetero in Colombia, achieving over 85% accuracy in identifying the Subthalamic Nucleus using a naive Bayes classifier.