{"id":7452,"date":"2026-04-02T11:32:04","date_gmt":"2026-04-02T11:32:04","guid":{"rendered":"https:\/\/www.proefschriftmaken.nl\/portfolio\/ilaria-jansen\/"},"modified":"2026-04-02T11:32:11","modified_gmt":"2026-04-02T11:32:11","slug":"ilaria-jansen","status":"publish","type":"us_portfolio","link":"https:\/\/www.proefschriftmaken.nl\/en\/portfolio\/ilaria-jansen\/","title":{"rendered":"Ilaria Jansen"},"content":{"rendered":"","protected":false},"excerpt":{"rendered":"","protected":false},"author":8,"featured_media":7455,"comment_status":"closed","ping_status":"closed","template":"","meta":{"_acf_changed":false,"footnotes":""},"us_portfolio_category":[45],"class_list":["post-7452","us_portfolio","type-us_portfolio","status-publish","has-post-thumbnail","hentry","us_portfolio_category-new-template"],"acf":{"naam_van_het_proefschift":"Bladder cancer diagnostics","samenvatting":"Er is geen Nederlandse samenvatting beschikbaar. De Engelse samenvatting vind je <a href=\"https:\/\/www.proefschriftmaken.nl\/en\/portfolio\/ilaria-jansen\/\">hier<\/a>.","summary":"Chapter 1 provides a general introduction. First, a detailed description of the epidemiology, recurrence rates, and current risk stratification regarding non-muscle invasive bladder cancer is given. Second, the role of histopathology and digital pathology is explained. Third, a general introduction of artificial intelligence is provided. Finally, the context and rationale for the studies included in this thesis are outlined.\n\nChapter 2 provides an overview of the current status of digital pathology and its future role in clinical practice. First, an introduction is given on the current histopathological workflow, emphasizing the limitations of current practice, in particular the assessment of two-dimensional slides. Second, the possibility of creating three-dimensional tumor reconstructions out of digitized two-dimensional slides is discussed. Ultimately, a future perspective is given; combining three-dimensional tumor reconstructions and computer aided diagnosis to enhance the histological assessment.\n\nChapter 3 assesses the feasibility of creating three-dimensional bladder tumor reconstructions by stacking multiple two-dimensional slides. En-bloc resection specimen of 21 patients were cut and digitized. Per case, 26-30 sections were included. The slides were aligned and the tumor and muscularis propria were manually delineated to create three-dimensional segmentations. The segmentations allowed full three-dimensional visualization and evaluation of the spatial relationship of the bladder tumor and the muscularis propria. Although we show the feasibility of three-dimensional bladder tumor reconstructions, the additional value of these reconstructions still has to be proven.\n\nChapter 4 reports the acquisition of three-dimensional mass spectrometry imaging datasets from fourteen bladder cancer resection specimens. Formalin fixed paraffin embedded tissue samples were used to obtain molecular information on a peptide level by in situ proteolytic digestion on the tissues\u2019 surface. Outlier detection showed that the data was robust and high correlation with histology was seen. Moreover, it was found that on average 33% of the sample has to be measured in order to obtain sufficient coverage of the existing biological variance within a sample. However, the interpretation of signals and the time-consuming technique withhold the implementation of mass spectrometry imaging in the near future.\n\nChapter 5 assesses the association of intravesical tumor location with the one- and five-year recurrence-free survival in patients with non-muscle invasive bladder cancer. Only patients with a primary, solitary, non-muscle invasive tumor are included in the study. Clinical data of 184 patients is retrospectively collected. Tumor areas are defined using the bladder map of the European Association of Urology. Subsequently, the areas are dichotomized into dorsal vs. non-dorsal tumors. The dorsal area is defined as the diamond-shaped area bordered by bladder neck, trigone, posterior wall and orifices. The non-dorsal areas are the lateral walls, dome and anterior wall. Results indicate that patients with a tumor in the non-dorsal area have a worse one- and five-year RSF. However, no association can be assessed with specific tumor location due to the limited number of patients.\n\nChapter 6 describes an automated method to detect and grade urothelial cell carcinoma on histological slides by using deep learning. Histological slides of 232 patients were included and the consensus reading of three specialized pathologists was utilized. Firstly, a segmentation network was trained to segment the urothelium. The segmentations of the first network were used to train a classification network to automatically grade the urothelial cell carcinoma using the World Health Organization 2004 grading. The segmentation network resulted in accurate detection of urothelium. The automated classification correctly graded 76% of the low-grade and 71% of the high-grade according to the consensus. Moderate agreement (\u03ba=0.48 \u00b1 0.14 se) with the consensus reading was seen. These results indicate that deep learning can be used for the fully automated detection and grading of urothelial cell carcinoma. Although the consensus of multiple pathologists is used, a gold standard is difficult to define since histological grading is prone to interobserver variation. A validation study must be done before this technique can be implemented into clinical practice.\n\nChapter 7 describes a deep learning method to predict the one- and five-year recurrence-free survival in patients with non-muscle invasive bladder cancer by combining clinical data with digitized histological slides. Retrospectively collected data of 395 patients and the corresponding histological slides were used. First, as described in Chapter 6 a deep learning network was used to segment the urothelium on the histological slides. Second, a selection network was trained to find the most predictive patches. Ultimately, a classification network was trained to predict the one- and five-year recurrence. The deep learning method showed an area-under-the-curve (AUC) of 0.62 and 0.76 for the one- and five-year recurrence prediction when combining clinical data with histopathology image data. This method shows better performance compared to using histopathology image data or clinical data only. Combining clinical data with histological imaging data has the potential to enhance the risk stratification for patients with NMIBC. A large multicenter study is needed to optimize and validate this proposed approach.\n\nIn conclusion, reconstructing three-dimensional bladder tumors out of whole slide images is feasible and we are able to combine these reconstructions with three-dimensional mass spectrometry imaging data. Moreover, our retrospective analysis shows that tumor located in the dorsal area of the bladder are favorably associated with recurrence-free survival. Additionally, this thesis concludes that deep learning has the potential to improve the risk stratification of patients with non-muscle invasive bladder cancer. Deep learning networks are able to accurately detect and grade urothelial cell carcinoma on histological slides. Moreover, deep learning networks can adequately predict the one- and five-year recurrence-free survival by combining clinical data with whole slide images.","auteur":"Ilaria Jansen","auteur_slug":"ilaria-jansen","publicatiedatum":"4 november 2020","taal":"NL","url_flipbook":"https:\/\/ebook.proefschriftmaken.nl\/ebook\/ilariajansen?iframe=true","url_download_pdf":"","url_epub":"","ordernummer":"FTP-202604021128","isbn":"978-94-6380-974-0","doi_nummer":"","naam_universiteit":"Universiteit van Amsterdam","afbeeldingen":7456,"naam_student:":"","binnenwerk":"","universiteit":"Universiteit van Amsterdam","cover":"","afwerking":"","cover_afwerking":"","design":""},"_links":{"self":[{"href":"https:\/\/www.proefschriftmaken.nl\/en\/wp-json\/wp\/v2\/us_portfolio\/7452","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.proefschriftmaken.nl\/en\/wp-json\/wp\/v2\/us_portfolio"}],"about":[{"href":"https:\/\/www.proefschriftmaken.nl\/en\/wp-json\/wp\/v2\/types\/us_portfolio"}],"author":[{"embeddable":true,"href":"https:\/\/www.proefschriftmaken.nl\/en\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/www.proefschriftmaken.nl\/en\/wp-json\/wp\/v2\/comments?post=7452"}],"version-history":[{"count":1,"href":"https:\/\/www.proefschriftmaken.nl\/en\/wp-json\/wp\/v2\/us_portfolio\/7452\/revisions"}],"predecessor-version":[{"id":7453,"href":"https:\/\/www.proefschriftmaken.nl\/en\/wp-json\/wp\/v2\/us_portfolio\/7452\/revisions\/7453"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.proefschriftmaken.nl\/en\/wp-json\/wp\/v2\/media\/7455"}],"wp:attachment":[{"href":"https:\/\/www.proefschriftmaken.nl\/en\/wp-json\/wp\/v2\/media?parent=7452"}],"wp:term":[{"taxonomy":"us_portfolio_category","embeddable":true,"href":"https:\/\/www.proefschriftmaken.nl\/en\/wp-json\/wp\/v2\/us_portfolio_category?post=7452"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}