{"id":26992,"date":"2026-09-27T20:02:20","date_gmt":"2026-09-27T18:02:20","guid":{"rendered":"https:\/\/www.itspy.cz\/thesis\/analysis-of-medical-data-by-xai-with-a-grad-cam-method\/"},"modified":"2026-10-23T13:44:53","modified_gmt":"2026-10-23T11:44:53","slug":"analysis-of-medical-data-by-xai-with-a-grad-cam-method","status":"publish","type":"thesis","link":"https:\/\/www.itspy.cz\/en\/thesis\/analysis-of-medical-data-by-xai-with-a-grad-cam-method\/","title":{"rendered":"Analysis of Medical Data by XAI with a Grad-CAM method"},"content":{"rendered":"<p>The reliability of modern deep learning models in the<br \/>\nmedical domain is frequently questioned due to their<br \/>\nblack-box nature. Post-hoc explainability techniques from<br \/>\nthe field of explainable artificial intelligence (XAI)<br \/>\noffer a means to improve transparency and assess the<br \/>\nreliability of predictions produced by convolutional neural<br \/>\nnetworks. However, in many medical imaging datasets only<br \/>\nimage-level diagnostic labels are available, and<br \/>\npixel-level annotations of pathological regions are not<br \/>\nprovided, limiting direct quantitative evaluation of<br \/>\nexplanation maps. The research aims to investigate how XAI<br \/>\nmethods, specifically Gradient-weighted Class Activation<br \/>\nMapping (Grad-CAM), can provide reliable explanations for<br \/>\nmedical image classification. For this purpose, MRI images<br \/>\nof brain were used to train a convolutional neural network<br \/>\nto categorize the four stages of dementia in Alzheimer&#8217;s<br \/>\ndisease. To make each prediction transparent, the areas of<br \/>\nthe brain which the trained network used to make the<br \/>\ncategorization on were highlighted using Grad-CAM. The<br \/>\nresulting relevance maps, heatmaps, were evaluated using<br \/>\ntwo approaches: spatial comparison with anatomically<br \/>\ndefined brain regions associated with Alzheimer\u2019s disease<br \/>\nusing atlas overlay, and quantitative faithfulness<br \/>\nassessment using a deletion-based metric, where highly<br \/>\ninfluential regions identified by Grad-CAM were<br \/>\nprogressively removed and the impact on classification<br \/>\nconfidence was measured.<\/p>\n","protected":false},"featured_media":26748,"template":"","meta":{"_acf_changed":false,"_links_to":"","_links_to_target":""},"university":[230,436,222],"thesis-year":[466],"class_list":["post-26992","thesis","type-thesis","status-publish","has-post-thumbnail","hentry","thesis-year-466"],"acf":{"autor":"Alexandra \u010ci\u017em\u00e1rov\u00e1","portret":"","vedouci":"prof. Ing. Elena Zaitseva, PhD."},"_links":{"self":[{"href":"https:\/\/www.itspy.cz\/en\/wp-json\/wp\/v2\/thesis\/26992","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.itspy.cz\/en\/wp-json\/wp\/v2\/thesis"}],"about":[{"href":"https:\/\/www.itspy.cz\/en\/wp-json\/wp\/v2\/types\/thesis"}],"version-history":[{"count":1,"href":"https:\/\/www.itspy.cz\/en\/wp-json\/wp\/v2\/thesis\/26992\/revisions"}],"predecessor-version":[{"id":27047,"href":"https:\/\/www.itspy.cz\/en\/wp-json\/wp\/v2\/thesis\/26992\/revisions\/27047"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.itspy.cz\/en\/wp-json\/wp\/v2\/media\/26748"}],"wp:attachment":[{"href":"https:\/\/www.itspy.cz\/en\/wp-json\/wp\/v2\/media?parent=26992"}],"wp:term":[{"taxonomy":"university","embeddable":true,"href":"https:\/\/www.itspy.cz\/en\/wp-json\/wp\/v2\/university?post=26992"},{"taxonomy":"thesis-year","embeddable":true,"href":"https:\/\/www.itspy.cz\/en\/wp-json\/wp\/v2\/thesis-year?post=26992"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}