{"id":26250,"date":"2023-06-01T18:00:43","date_gmt":"2023-06-01T16:00:43","guid":{"rendered":"https:\/\/www.teoresigroup.com\/thesis\/echocardiographic-dataset-creation-and-left-ventricular-hypertrophy-detection-using-a-weakly-supervised-residual-neural-network\/"},"modified":"2023-06-01T18:02:30","modified_gmt":"2023-06-01T16:02:30","slug":"echocardiographic-dataset-creation-and-left-ventricular-hypertrophy-detection-using-a-weakly-supervised-residual-neural-network","status":"publish","type":"thesis","link":"https:\/\/teoresi.sixeleven.it\/it\/thesis\/echocardiographic-dataset-creation-and-left-ventricular-hypertrophy-detection-using-a-weakly-supervised-residual-neural-network\/","title":{"rendered":"Echocardiographic dataset creation and left ventricular hypertrophy detection using a weakly supervised residual neural network"},"content":{"rendered":"\n<div class=\"wp-block-columns align-center row sezione\">\n<div class=\"wp-block-column small-12 medium-10 large-8\">\n<h5 class=\"wp-block-heading has-text-align-center h5\">Abstract<\/h5>\n\n\n\n<p>L\u2019ipertrofia ventricolare sinistra (LVH) \u00e8 un cambiamento strutturale caratterizzato da un incremento nella massa delle pareti ventricolari che pu\u00f2 condurre a sconpenso cardiaco. Lo scopo di questo lavoro \u00e8 usare il deep leaning per individuare automaticamente l\u2019LVH da ecocardiografie.<\/p>\n\n\n\n<p>Abbiamo creato un dataset di circa 10000 immagini e costruito un classificatore a singole immagini basato su ResNet50 per individuare l\u2019LVH. Inoltre, abbiamo applicato l\u2019analisi Grad-CAM per ottenere una validazione visiva del modello.<\/p>\n\n\n\n<p>La rete ha raggiunto un\u2019AUC di 0.99, un\u2019accuracy di 0.94 e un F1-score di 0.94. L\u2019analisi Grad\u2011CAM ha confermato che il modello si concentra su regioni rilevanti per la diagnosi di LVH. In conclusione, la nostra rete ha la capacit\u00e0 di individuare automaticamente l\u2019LVH e anche di localizzare strutture cardiache chiave con solo label a livello di immagine come supervisione.<\/p>\n\n\n\n<h5 class=\"wp-block-heading has-text-align-center\">Obiettivo tesi<\/h5>\n\n\n\n<p>Creare una struttura dati adatta a compiti di machine learning e individuare l\u2019ipertrofia ventricolare sinistra da ecocardiografie usando il deep learning.<\/p>\n\n\n\n<h5 class=\"wp-block-heading has-text-align-center h5\">Metodologia di ricerca<\/h5>\n\n\n\n<p>Abbiamo ottenuto un dataset di circa 10000 immagini che \u00e8 stato diviso in: 80% per i training e validation set usati durante un\u2019addestramento con 3-fold cross validation, e 20% per test set. Abbiamo costruito e addestrato un classificatore basato su ResNet50 usando la libreria Keras di Python. Abbiami usato l\u2019analisi Grad\u2011CAM per ottenere una validazione visiva del modello.<\/p>\n\n\n\n<h5 class=\"wp-block-heading has-text-align-center h5\">Conclusioni<\/h5>\n\n\n\n<p>Il modello ha raggiunto un\u2019accuracy di 0.94, un\u2019AUC di 0.99 e un F1-score di 0.94. L\u2019analisi Grad\u2011CAM ha mostrato che il modello si concentra sulla parete posteriore del ventricolo sinistro che \u00e8 effettivamente una regione rilevante per la diagnosi di LVH.<\/p>\n\n\n\n<h5 class=\"wp-block-heading has-text-align-center h5\">Sviluppi futuri<\/h5>\n\n\n\n<p>Estendere il dataset, introdurre uno step di segmentazione, selezionare solo certi frame dalle ecocardiografie, individuare altre patologie.<\/p>\n<\/div>\n<\/div>\n","protected":false},"featured_media":0,"template":"","university":[177],"thesis_type":[273,275],"keyword":[242,500,502],"class_list":["post-26250","thesis","type-thesis","status-publish","hentry","university-universita-degli-studi-di-napoli-federico-ii-it","thesis_type-artificial-intelligence-it","thesis_type-deep-learning-it","keyword-deep-learning","keyword-echocardiography","keyword-left-ventricular-hypertrophy-2"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v26.8 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Echocardiographic dataset creation and left ventricular hypertrophy detection using a weakly supervised residual neural network - Teoresi Group<\/title>\n<meta name=\"robots\" content=\"noindex, follow\" \/>\n<meta property=\"og:locale\" content=\"it_IT\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Echocardiographic dataset creation and left ventricular hypertrophy detection using a weakly supervised residual neural network - Teoresi Group\" \/>\n<meta property=\"og:description\" content=\"Abstract L\u2019ipertrofia ventricolare sinistra (LVH) \u00e8 un cambiamento strutturale caratterizzato da un incremento nella massa delle pareti ventricolari che pu\u00f2 condurre a sconpenso cardiaco. Lo scopo di questo lavoro \u00e8 usare il deep leaning per individuare automaticamente l\u2019LVH da ecocardiografie. Abbiamo creato un dataset di circa 10000 immagini e costruito un classificatore a singole immagini basato su ResNet50 per individuare...\" \/>\n<meta property=\"og:url\" content=\"https:\/\/teoresi.sixeleven.it\/it\/thesis\/echocardiographic-dataset-creation-and-left-ventricular-hypertrophy-detection-using-a-weakly-supervised-residual-neural-network\/\" \/>\n<meta property=\"og:site_name\" content=\"Teoresi Group\" \/>\n<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/pages\/Gruppo-Teoresi\/118393464917205\" \/>\n<meta property=\"article:modified_time\" content=\"2023-06-01T16:02:30+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/teoresi.sixeleven.it\/wp-content\/uploads\/2021\/02\/og-image.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"1200\" \/>\n\t<meta property=\"og:image:height\" content=\"630\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/teoresi.sixeleven.it\/it\/thesis\/echocardiographic-dataset-creation-and-left-ventricular-hypertrophy-detection-using-a-weakly-supervised-residual-neural-network\/\",\"url\":\"https:\/\/teoresi.sixeleven.it\/it\/thesis\/echocardiographic-dataset-creation-and-left-ventricular-hypertrophy-detection-using-a-weakly-supervised-residual-neural-network\/\",\"name\":\"Echocardiographic dataset creation and left ventricular hypertrophy detection using a weakly supervised residual neural network - Teoresi Group\",\"isPartOf\":{\"@id\":\"https:\/\/teoresi.sixeleven.it\/de\/#website\"},\"datePublished\":\"2023-06-01T16:00:43+00:00\",\"dateModified\":\"2023-06-01T16:02:30+00:00\",\"breadcrumb\":{\"@id\":\"https:\/\/teoresi.sixeleven.it\/it\/thesis\/echocardiographic-dataset-creation-and-left-ventricular-hypertrophy-detection-using-a-weakly-supervised-residual-neural-network\/#breadcrumb\"},\"inLanguage\":\"it-IT\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/teoresi.sixeleven.it\/it\/thesis\/echocardiographic-dataset-creation-and-left-ventricular-hypertrophy-detection-using-a-weakly-supervised-residual-neural-network\/\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/teoresi.sixeleven.it\/it\/thesis\/echocardiographic-dataset-creation-and-left-ventricular-hypertrophy-detection-using-a-weakly-supervised-residual-neural-network\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\",\"item\":\"https:\/\/teoresi.sixeleven.it\/it\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Echocardiographic dataset creation and left ventricular hypertrophy detection using a weakly supervised residual neural network\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/teoresi.sixeleven.it\/de\/#website\",\"url\":\"https:\/\/teoresi.sixeleven.it\/de\/\",\"name\":\"Teoresi Group\",\"description\":\"Engineering for Human life.\",\"publisher\":{\"@id\":\"https:\/\/teoresi.sixeleven.it\/de\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/teoresi.sixeleven.it\/de\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"it-IT\"},{\"@type\":\"Organization\",\"@id\":\"https:\/\/teoresi.sixeleven.it\/de\/#organization\",\"name\":\"Teoresi Group\",\"url\":\"https:\/\/teoresi.sixeleven.it\/de\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"it-IT\",\"@id\":\"https:\/\/teoresi.sixeleven.it\/de\/#\/schema\/logo\/image\/\",\"url\":\"https:\/\/www.teoresigroup.com\/wp-content\/uploads\/2020\/12\/teoresi-logo-color.svg\",\"contentUrl\":\"https:\/\/www.teoresigroup.com\/wp-content\/uploads\/2020\/12\/teoresi-logo-color.svg\",\"width\":1,\"height\":1,\"caption\":\"Teoresi Group\"},\"image\":{\"@id\":\"https:\/\/teoresi.sixeleven.it\/de\/#\/schema\/logo\/image\/\"},\"sameAs\":[\"https:\/\/www.facebook.com\/pages\/Gruppo-Teoresi\/118393464917205\",\"https:\/\/www.instagram.com\/teoresigroup\/\",\"https:\/\/www.linkedin.com\/company\/teoresigroup\/\",\"https:\/\/www.youtube.com\/channel\/UCNdBBSSax2EWPmFk5QzYP8Q\"]}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Echocardiographic dataset creation and left ventricular hypertrophy detection using a weakly supervised residual neural network - Teoresi Group","robots":{"index":"noindex","follow":"follow"},"og_locale":"it_IT","og_type":"article","og_title":"Echocardiographic dataset creation and left ventricular hypertrophy detection using a weakly supervised residual neural network - Teoresi Group","og_description":"Abstract L\u2019ipertrofia ventricolare sinistra (LVH) \u00e8 un cambiamento strutturale caratterizzato da un incremento nella massa delle pareti ventricolari che pu\u00f2 condurre a sconpenso cardiaco. Lo scopo di questo lavoro \u00e8 usare il deep leaning per individuare automaticamente l\u2019LVH da ecocardiografie. Abbiamo creato un dataset di circa 10000 immagini e costruito un classificatore a singole immagini basato su ResNet50 per individuare...","og_url":"https:\/\/teoresi.sixeleven.it\/it\/thesis\/echocardiographic-dataset-creation-and-left-ventricular-hypertrophy-detection-using-a-weakly-supervised-residual-neural-network\/","og_site_name":"Teoresi Group","article_publisher":"https:\/\/www.facebook.com\/pages\/Gruppo-Teoresi\/118393464917205","article_modified_time":"2023-06-01T16:02:30+00:00","og_image":[{"width":1200,"height":630,"url":"https:\/\/teoresi.sixeleven.it\/wp-content\/uploads\/2021\/02\/og-image.jpg","type":"image\/jpeg"}],"twitter_card":"summary_large_image","schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/teoresi.sixeleven.it\/it\/thesis\/echocardiographic-dataset-creation-and-left-ventricular-hypertrophy-detection-using-a-weakly-supervised-residual-neural-network\/","url":"https:\/\/teoresi.sixeleven.it\/it\/thesis\/echocardiographic-dataset-creation-and-left-ventricular-hypertrophy-detection-using-a-weakly-supervised-residual-neural-network\/","name":"Echocardiographic dataset creation and left ventricular hypertrophy detection using a weakly supervised residual neural network - Teoresi Group","isPartOf":{"@id":"https:\/\/teoresi.sixeleven.it\/de\/#website"},"datePublished":"2023-06-01T16:00:43+00:00","dateModified":"2023-06-01T16:02:30+00:00","breadcrumb":{"@id":"https:\/\/teoresi.sixeleven.it\/it\/thesis\/echocardiographic-dataset-creation-and-left-ventricular-hypertrophy-detection-using-a-weakly-supervised-residual-neural-network\/#breadcrumb"},"inLanguage":"it-IT","potentialAction":[{"@type":"ReadAction","target":["https:\/\/teoresi.sixeleven.it\/it\/thesis\/echocardiographic-dataset-creation-and-left-ventricular-hypertrophy-detection-using-a-weakly-supervised-residual-neural-network\/"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/teoresi.sixeleven.it\/it\/thesis\/echocardiographic-dataset-creation-and-left-ventricular-hypertrophy-detection-using-a-weakly-supervised-residual-neural-network\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"","item":"https:\/\/teoresi.sixeleven.it\/it\/"},{"@type":"ListItem","position":2,"name":"Echocardiographic dataset creation and left ventricular hypertrophy detection using a weakly supervised residual neural network"}]},{"@type":"WebSite","@id":"https:\/\/teoresi.sixeleven.it\/de\/#website","url":"https:\/\/teoresi.sixeleven.it\/de\/","name":"Teoresi Group","description":"Engineering for Human life.","publisher":{"@id":"https:\/\/teoresi.sixeleven.it\/de\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/teoresi.sixeleven.it\/de\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"it-IT"},{"@type":"Organization","@id":"https:\/\/teoresi.sixeleven.it\/de\/#organization","name":"Teoresi Group","url":"https:\/\/teoresi.sixeleven.it\/de\/","logo":{"@type":"ImageObject","inLanguage":"it-IT","@id":"https:\/\/teoresi.sixeleven.it\/de\/#\/schema\/logo\/image\/","url":"https:\/\/www.teoresigroup.com\/wp-content\/uploads\/2020\/12\/teoresi-logo-color.svg","contentUrl":"https:\/\/www.teoresigroup.com\/wp-content\/uploads\/2020\/12\/teoresi-logo-color.svg","width":1,"height":1,"caption":"Teoresi Group"},"image":{"@id":"https:\/\/teoresi.sixeleven.it\/de\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/www.facebook.com\/pages\/Gruppo-Teoresi\/118393464917205","https:\/\/www.instagram.com\/teoresigroup\/","https:\/\/www.linkedin.com\/company\/teoresigroup\/","https:\/\/www.youtube.com\/channel\/UCNdBBSSax2EWPmFk5QzYP8Q"]}]}},"acf":[],"_links":{"self":[{"href":"https:\/\/teoresi.sixeleven.it\/it\/wp-json\/wp\/v2\/thesis\/26250","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/teoresi.sixeleven.it\/it\/wp-json\/wp\/v2\/thesis"}],"about":[{"href":"https:\/\/teoresi.sixeleven.it\/it\/wp-json\/wp\/v2\/types\/thesis"}],"version-history":[{"count":2,"href":"https:\/\/teoresi.sixeleven.it\/it\/wp-json\/wp\/v2\/thesis\/26250\/revisions"}],"predecessor-version":[{"id":26253,"href":"https:\/\/teoresi.sixeleven.it\/it\/wp-json\/wp\/v2\/thesis\/26250\/revisions\/26253"}],"wp:attachment":[{"href":"https:\/\/teoresi.sixeleven.it\/it\/wp-json\/wp\/v2\/media?parent=26250"}],"wp:term":[{"taxonomy":"university","embeddable":true,"href":"https:\/\/teoresi.sixeleven.it\/it\/wp-json\/wp\/v2\/university?post=26250"},{"taxonomy":"thesis_type","embeddable":true,"href":"https:\/\/teoresi.sixeleven.it\/it\/wp-json\/wp\/v2\/thesis_type?post=26250"},{"taxonomy":"keyword","embeddable":true,"href":"https:\/\/teoresi.sixeleven.it\/it\/wp-json\/wp\/v2\/keyword?post=26250"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}