Accurate Diagnosis and Survival Prediction of Bladder Cancer Using Deep Learning on Histological Slides

Abstract

(1) Background: Early diagnosis and treatment are essential to reduce the mortality rate of bladder cancer (BLCA). We aimed to develop deep learning (DL)-based weakly supervised models for the diagnosis of BLCA and prediction of overall survival (OS) in muscle-invasive bladder cancer (MIBC) patients using whole slide digitized histological images (WSIs). (2) Methods: Diagnostic and prognostic models were developed using 926 WSIs of 412 BLCA patients from The Cancer Genome Atlas cohort. We collected 250 WSIs of 150 BLCA patients from the Renmin Hospital of Wuhan University cohort for external validation of the models. Two DL models were developed: a BLCA diagnostic model (named BlcaMIL) and an MIBC prognostic model (named MibcMLP). (3) Results: The BlcaMIL model identified BLCA with accuracy 0.987 in the external validation set, comparable to that of expert uropathologists and outperforming a junior pathologist. The C-index values for the MibcMLP model on the internal and external validation sets were 0.631 and 0.622, respectively. The risk score predicted by MibcMLP was a strong predictor independent of existing clinical or histopathologic indicators, as demonstrated by univariate Cox (HR = 2.390, p < 0.0001) and multivariate Cox (HR = 2.414, p < 0.0001) analyses. The interpretability of DL models can help in the analysis of critical regions associated with tumors to enrich the information obtained from WSIs. Furthermore, the expression of six genes (ANAPC7, MAPKAPK5, COX19, LINC01106, AL161431.1 and MYO16-AS1) was significantly associated with MibcMLP-predicted risk scores, revealing possible potential biological correlations. (4) Conclusions: Our study developed DL models for accurately diagnosing BLCA and predicting OS in MIBC patients, which will help promote the precise pathological diagnosis of BLCA and risk stratification of MIBC to improve clinical treatment decisions.

Recent Advancements in Deep Learning Using Whole Slide Imaging for Cancer Prognosis

“…One promising direction for future research is the exploration of weakly supervised learning models for cancer prognosis. Such models, as proposed by Shao et al [ 80 ], Zheng et al [ 50 ], and Shi et al [ 58 ], could potentially alleviate the reliance on extensive manual annotations, reducing the workload of pathologists and accelerating the development of predictive models. Moreover, the use of advanced techniques like multiresolution deep learning methods and attention mechanisms, as employed by Liu and Kurc [ 38 ] and Jiang et al [ 83 ], should continue to be investigated for more sophisticated analysis and more accurate prediction results.…”

“…The potential of weakly supervised learning models has also been explored, with Shao et al [ 80 ] proposing BDOCOX, a weakly supervised deep ordinal Cox model for survival prediction from WSIs. In similar vein, Zheng et al [ 50 ] developed weakly supervised deep learning models for diagnosing bladder cancer and predicting overall survival rates. Shi et al [ 58 ] proposed a weakly supervised deep learning framework for hepatocellular carcinoma analysis, establishing a ’tumor risk score’ from WSIs that outperformed traditional clinical staging systems in predictive ability.…”

Global research trends of the application of artificial intelligence in bladder cancer since the 21st century: a bibliometric analysis

“…The diagnostic methods for BCa include imaging, cystoscopy, urine tests, etc. and pathological diagnosis is the gold standard ( 3 , 4 ). BCa can be divided into two types in pathology according to the degree of muscle invasion: non-muscle-invasive BCa and muscle-invasive BCa ( 5 ), and prognosis of them is very different.…”

“…BCa can be divided into two types in pathology according to the degree of muscle invasion: non-muscle-invasive BCa and muscle-invasive BCa ( 5 ), and prognosis of them is very different. It requires pathologists’ efforts to identify them, but the manual review of pathologist may sometimes bring mistakes ( 4 ). Imaging tests can help detect BCa early ( 6 ), but they are also influenced by human factors.…”

A Weakly Supervised Deep Learning Model and Human–Machine Fusion for Accurate Grading of Renal Cell Carcinoma from Histopathology Slides

“…Artificial intelligence technology has become one of the most promising fields in computational pathology [ 11 , 12 , 13 , 14 ]. Multiple studies have shown that deep learning algorithms can extract key features from hematoxylin and eosin (H&E)-stained histopathological images, enabling diagnosis and subtyping with comparable or better accuracy than expert pathologists [ 15 , 16 , 17 , 18 , 19 ]. This evidence not only demonstrates the potential of deep learning to improve traditional diagnosis and prediction, but also helps pathologists reduce repetitive and tedious work, freeing up time to handle more complex tasks [ 20 ].…”