Integrating Diagnosis Rules into Deep Neural Networks for Bladder Cancer Staging

Abstract

Bladder cancer is a malignant disease with substantial morbidity and mortality. Bladder cancer staging is crucial to determine the effective treatments of bladder tumors in clinic. As to the superiority of feature learning, Deep Convolutional Neural Networks (DCNN) are widely used to predict the cancer stage based on medical images. However, most existing DCNN-based cancer staging methods are data-driven and neglect the domain knowledge and experiences of clinicians. Besides, the deep neural networks are short of model interpretability and may lead to risky diagnosis. To tackle the problems, we construct the diagnosis rules of bladder cancer staging based on the clinical experiences of tumor penetration into bladder wall. The diagnosis rules are extracted from Magnetic Resonance (MR) images and further integrated into DCNN for joint identification of tumor stage. The experiments validate that the integrated rules improve the model interpretability and guide DCNN to focus on the regions of tumor penetration and thereby produce precise prediction of cancer staging. CCS CONCEPTS • Applied computing → Life and medical sciences; • Computing methodologies → Knowledge representation and reasoning.

PENet: Prior evidence deep neural network for bladder cancer staging

“…and a rule-integrated approach such as RuleNet [25]. We discovered that PENet's performance was superior to each of these other methods.…”

“…Functional characteristics representing percentiles of the cumulative distribution function (CDF), morphological features representing radionics texture features, and morphological features defining tumor shape were retrieved from T2W-MRI, and DW-MRI as input to neural networks for bladder cancer staging [24]. Zhang et al learned infiltration criteria from MR images that are advantageous for tumor staging based on clinical experiences and used the rules into DCNN to increase performance [25]. Using ResNet structure, non-local attention, and image super-resolution processing, [26] developed a model with high performance for CT imaging-based bladder cancer staging.…”

Radiomics-based T-staging of hollow organ cancers

“…Studies of EC ( 29 ), CRC ( 52 ), and BC ( 53 ) have used deep learning-based methods for T-staging. This type studies are relatively lacking, since doctors usually use optical endoscopy to screen polyps in hollow organs.…”

“…These deep learning works use imaging modalities such as PET ( 29 ), CT ( 52 ), and T2-weighted imaging (T2WI) ( 53 ). Deep learning usually requires a large amount of data for training.…”

Deep Neural Networks with Prior Evidence for Bladder Cancer Staging

“…To validate the effectiveness of our method, we compare our method with four state-of-the-art DCNN-based medical image classification methods including ResNet18 [29], DenseNet [30], EvientialNet [28] and another prior-integrated DCNN method RuleNet [18]. Figure 4 and Table II present the comparison of cancer stage classification results produced by different methods.…”