The most common type of bladder cancer is urothelial carcinoma, which is among the cancer types with the highest recurrence rate and lifetime treatment cost per patient. Diagnosed patients are stratified into risk groups, mainly based on grade and stage. However, it is well known that correct grading of bladder cancer suffers from intra- and interobserver variability and inconsistent reproducibility between pathologists, potentially leading to under- or overtreatment of the patients. The economic burden, unnecessary patient suffering, and additional load on the health care system illustrate the importance of developing new tools to aid pathologists. We propose a pipeline, called TRI _grade , that will identify diagnostic relevant regions in the whole-slide image (WSI) and collectively predict the grade of the current WSI. The system consists of two main models, trained on weak slide-level grade labels. First, a WSI is segmented into the different tissue classes (urothelium, stroma, muscle, blood, damaged tissue, and background). Next, tiles are extracted from the diagnostic relevant urothelium tissue from three magnification levels (25x, 100x, and 400x) and processed sequentially by a convolutional neural network (CNN) based model. Ten models were trained for the slide-level grading experiment, where the best model achieved an F1-score of 0.90 on a test set consisting of 50 WSIs. The best model was further evaluated on a smaller segmentation test set, consisting of 14 WSIs where low- and high-grade regions were annotated by a pathologist. The TRI _grade pipeline achieved an average F1-score of 0.91 for both the low-grade and high-grade classes.

Background: European treatment guidelines for pTa and pT1 urinary bladder urothelial carcinoma depend highly on stage and WHO-grade. Both the WHO73 and the WHO04 grading systems show some intra- and interobserver variability. The current pilot study investigates which histopathological features are especially sensitive for this undesired lack of reproducibility and the influence on prognostic value. Methods: Thirty-eight cases of primary non-muscle invasive urothelial carcinomas, including thirteen cases with stage progression, were reviewed by three pathologists. Thirteen microscopic features were extracted from pathology textbooks and evaluated separately. Reproducibility was measured using Gwet’s agreement coefficients. Prognostic ability regarding progression was estimated by the area under curve (AUC) of the receiver operating characteristics (ROC) function. Results: The best reproducible features (Gwet’s agreement coefficient above 0.60) were papillary architecture, nuclear polarity, cellular maturation, nuclear enlargement and giant nuclei. Nucleoli was the strongest prognostic feature, and the only feature with an AUC above 0.70 for both grading systems, but reproducibility was not among the strongest. Nuclear polarity also had prognostic value with an AUC of 0.70 and 0.67 for the WHO73 and WHO04, respectively. The other features did not have significant prognostic value. Conclusions: The reproducibility of the histopathological features of the different WHO grading systems varied considerably. Of all the features evaluated, only nuclear polarity was both prognostic and significantly reproducible. Further validation studies are needed on these features to improve grading of urothelial carcinomas.

Correct treatment of urothelial carcinoma patients is dependent on accurate grading and staging of the cancer tumour. This is determined manually by a pathologist by examining the histological whole-slide images (WSI). The large size of these images makes this a time-consuming and challenging task. The WSI contain a variety of tissue types, and a method for defining diagnostic relevant regions would have several advantages for visualization as well as further input to automated diagnosis systems. We propose an automatic multiscale method for classification of tiles from WSI of urothelial carcinoma patients into six classes. Three architectures based on convolutional neural network (CNN) were tested: MONO-CNN (400x), DI-CNN (100x/400x) and TRI-CNN (25x/100x/400x). The preliminary results show that the two multiscale models performed significantly better than the mono-scale model, achieving an F1-score of 0.986, substantiating that utilising multiple scales in the model aids the classification accuracy.

Bladder cancer is a common type of cancer, especially among men in developed countries. Most cancers in the urinary bladder are papillary urothelial carcinomas. They are characterized by a high recurrence frequency (up to 70 %) after local resection. It is crucial for prognosis to discover these recurrent tumours at an early stage, especially before they become muscle-invasive. Reliable prognostic biomarkers for tumour recurrence and stage progression are lacking. This is why patients diagnosed with a non-muscle invasive bladder cancer follow extensive follow-up regimens with possible serious side effects and with high costs for the healthcare systems. WHO grade and tumour stage are two central biomarkers currently having great impact on both treatment decisions and follow-up regimens. However, there are concerns regarding the reproducibility of WHO grading, and stage classification is challenging in small and fragmented tumour material. In Paper I, we examined the reproducibility and the prognostic value of all the individual microscopic features making up the WHO grading system. Among thirteen extracted features there was considerable variation in both reproducibility and prognostic value. The only feature being both reasonably reproducible and statistically significant prognostic was cell polarity. We concluded that further validation studies are needed on these features, and that future grading systems should be based on well-defined features with true prognostic value. With the implementation of immunotherapy, there is increasing interest in tumour immune response and the tumour microenvironment. In a search for better prognostic biomarkers for tumour recurrence and stage progression, in Paper II, we investigated the prognostic value of tumour infiltrating immune cells (CD4, CD8, CD25 and CD138) and previously investigated cell proliferation markers (Ki-67, PPH3 and MAI). Low Ki 67 and tumour multifocality were associated with increased recurrence risk. Recurrence risk was not affected by the composition of immune cells. For stage progression, the only prognostic immune cell marker was CD25. High values for MAI was also strongly associated with stage progression. However, in a multivariate analysis, the most prognostic feature was a combination of MAI and CD25. BCG-instillations in the bladder are indicated in intermediate and high-risk non-muscle invasive bladder cancer patients. This old-fashion immunotherapy has proved to reduce both recurrence- and progression-risk, although it is frequently followed by unpleasant side-effects. As many as 30-50% of high-risk patients receiving BCG instillations, fail by develop high-grade recurrences. They do not only suffer from unnecessary side-effects, but will also have a delay in further treatment. Together with colleagues at three different Dutch hospitals, in Paper III, we looked at the prognostic and predictive value of T1-substaging. A T1-tumour invades the lamina propria, and we wanted to separate those with micro- from those with extensive invasion. We found that BCG-failure was more common among patients with extensive invasion. Furthermore, T1-substaging was associated with both high-grade recurrence-free and progression-free survival. Finally, in Paper IV, we wanted to investigate the prognostic value of two classical immunohistochemical markers, p53 and CK20, and compare them with previously investigated proliferation markers. p53 is a surrogate marker for mutations in the gene TP53, considered to be a main characteristic for muscle-invasive tumours. CK20 is a surrogate marker for luminal tumours in the molecular classification of bladder cancer, and is frequently used to distinguish reactive urothelial changes from urothelial carcinoma in situ. We found both positivity for p53 and CK20 to be significantly associated with stage progression, although not performing better than WHO grade and stage. The proliferation marker MAI, had the highest prognostic value in our study. Any combination of variables did not perform better in a multivariate analysis than MAI alone.