Examinando por Autor "Bardhi, Ornela"

Mostrando 1 - 4 de 4
  • Miniatura
    Ítem
    Deep learning models for colorectal polyps
    (MDPI AG, 2021-06-10) Bardhi, Ornela; Sierra-Sosa, Daniel; García-Zapirain, Begoña; Bujanda Fernández de Piérola, Luis
    Colorectal cancer is one of the main causes of cancer incident cases and cancer deaths worldwide. Undetected colon polyps, be them benign or malignant, lead to late diagnosis of colorectal cancer. Computer aided devices have helped to decrease the polyp miss rate. The application of deep learning algorithms and techniques has escalated during this last decade. Many scientific studies are published to detect, localize, and classify colon polyps. We present here a brief review of the latest published studies. We compare the accuracy of these studies with our results obtained from training and testing three independent datasets using a convolutional neural network and autoencoder model. A train, validate and test split was performed for each dataset, 75%, 15%, and 15%, respectively. An accuracy of 0.937 was achieved for CVC-ColonDB, 0.951 for CVC-ClinicDB, and 0.967 for ETIS-LaribPolypDB. Our results suggest slight improvements compared to the algorithms used to date
  • Miniatura
    Ítem
    Factors influencing care pathways for breast and prostate cancer in a hospital setting
    (MDPI, 2021-07-26) Bardhi, Ornela ; García-Zapirain, Begoña ; Nuño Solinís, Roberto
    Breast cancer (BCa) and prostate cancer (PCa) are the most prevalent types of cancers. We aimed to understand and analyze the care pathways for BCa and PCa patients followed at a hospital setting by analyzing their different treatment lines. We evaluated the association between different treatment lines and the lifestyle and demographic characteristics of these patients. Two datasets were created using the electronic health records (EHRs) and information collected through semi-structured one-on-one interviews. Statistical analysis was performed to examine which variable had an impact on the treatment each patient followed. In total, 83 patients participated in the study that ran between January and November 2018 in Beacon Hospital. Results show that chemotherapy cycles indicate if a patient would have other treatments, i.e., patients who have targeted therapy (25/46) have more chemotherapy cycles (95% CI 4.66–9.52, p = 0.012), the same is observed with endocrine therapy (95% CI 4.77–13.59, p = 0.044). Patients who had bisphosphonate (11/46), an indication of bone metastasis, had more chemotherapy cycles (95% CI 5.19–6.60, p = 0.012). PCa patients with tall height (95% CI 176.70–183.85, p = 0.005), heavier (95% CI 85.80–99.57, p < 0.001), and a BMI above 25 (95% CI 1.85–2.62, p = 0.017) had chemotherapy compared to patients who were shorter, lighter and with BMI less than 25. Initial prostate-specific antigen level (PSA level) indicated if a patient would be treated with bisphosphonate or not (95% CI 45.51–96.14, p = 0.002). Lifestyle variables such as diet (95% CI 1.46–1.85, p = 0.016), and exercise (95% CI 1.20–1.96, p = 0.029) indicated that healthier and active BCa patients had undergone surgeries. Our findings show that chemotherapy cycles and lifestyle for BCa, and tallness and weight for PCa may indicate the rest of treatment plan for these patients. Understanding factors that influence care pathways allow a more person-centered care approach and the redesign of care processes.
  • Miniatura
    Ítem
    Machine learning and algorithms applied to ethnographic and biomedical cancer data
    (Universidad de Deusto, 2022-02-18) Bardhi, Ornela; García-Zapirain, Begoña; Facultad de Ingeniería; Programa de Doctorado en Ingeniería para la Sociedad de la Información y Desarrollo Sostenible por la Universidad de Deusto
    Technology has seen an increased presence in the healthcare field for many years now. The last decade especially has seen a boom due to the progress of machine learning techniques and algorithms as well as the digitalization of healthcare records. These records are of different formats, such as text data, images, video, etc. and each requires specific ways to preprocess and analyze it. This thesis tackles important health issues faced in our society through ethnographic and biomedical data analysis using statistical analysis, machine learning and deep learning. The thesis is comprised of three case studies conducted in Ireland, Finland, and Spain, and each follows a different methodology and analysis approach. The first study deals with care pathways, their implementation in the last 20 years around the world, and the Beacon Hospital study. Understanding what factors influence care pathways allow a more person-centered care approach and the redesign of care processes. Four main tasks have been achieved in this study: a literature review of cancer care pathway implementation, an ethnographic study with breast and prostate cancer patients at Beacon Hospital about their perspective on care pathways, creation of two datasets with information coming from electronic health records and one-on-one interviews, and an analysis of the data through statistical analysis to identify the factors influencing care pathways for these two cancer diseases in a hospital setting. The second study is about the use of electronic health records to predict cancer patient survivability employing various machine learning algorithms. A collaboration with a regional hospital in Finland helped to achieve this task. Two steps were taken to predict survivability. The first one was to select the most relevant variables through various feature selection algorithms, and the second one was to perform survival prediction using nine machine learning algorithms. The third and the last study is about colorectal polyps detection using deep learning to prevent colorectal cancer from forming or progressing. The tasks performed to complete it follow a comprehensive review of the published scientific research related to colorectal polyp detection, classification, segmentation, localization, and the implementation of combined convolutional neural networks and autoencoders model to detect colorectal polyps without image preprocessing. All three case studies are accepted for publication in high-impact journals; two are already published online, one is currently in press.
  • Miniatura
    Ítem
    Machine learning techniques applied to electronic healthcare records to predict cancer patient survivability
    (Tech Science Press, 2021-04-13) Bardhi, Ornela; García-Zapirain, Begoña
    Breast cancer (BCa) and prostate cancer (PCa) are the two most common types of cancer. Various factors play a role in these cancers, and discovering the most important ones might help patients live longer, better lives. This study aims to determine the variables that most affect patient survivability, and how the use of different machine learning algorithms can assist in such predictions. The AURIA database was used, which contains electronic healthcare records (EHRs) of 20,006 individual patients diagnosed with either breast or prostate cancer in a particular region in Finland. In total, there were 178 features for BCa and 143 for PCa. Six feature selection algorithms were used to obtain the 21 most important variables for BCa, and 19 for PCa. These features were then used to predict patient survivability by employing nine different machine learning algorithms. Seventy-five percent of the dataset was used to train the models and 25% for testing. Cross-validation was carried out using the Stratified Kfold technique to test the effectiveness of the machine learning models. The support vector machine classifier yielded the best ROC with an area under the curve (AUC) = 0.83, followed by the KNeighbors Classifier with AUC = 0.82 for the BCa dataset. The two algorithms that yielded the best results for PCa are the random forest classifier and KNeighbors Classifier, both with AUC = 0.82. This study shows that not all variables are decisive when predicting breast or prostate cancer patient survivability. By narrowing down the input variables, healthcare professionals were able to focus on the issues that most impact patients, and hence devise better, more individualized care plans.