Background: Nonmuscle-invasive bladder cancer is diagnosed, treated, and monitored using cystoscopy. Artificial intelligence (AI) is increasingly used to augment tumor detection, but its performance is hindered by the limited availability of cystoscopic images required to form a large training data set. This study aimed to determine whether stepwise transfer learning with general images followed by gastroscopic images can improve the accuracy of bladder tumor detection on cystoscopic imaging. Materials and Methods: We trained a convolutional neural network with 1.2 million general images, followed by 8728 gastroscopic images. In the final step of the transfer learning process, the model was additionally trained with 2102 cystoscopic images of normal bladder tissue and bladder tumors collected at the University of Tsukuba Hospital. The diagnostic accuracy was evaluated using a receiver operating characteristic curve. The diagnostic performance of the models trained with cystoscopic images with or without stepwise organic transfer learning was compared with that of medical students and urologists with varying levels of experience. Results: The model developed by stepwise organic transfer learning had 95.4% sensitivity and 97.6% specificity. This performance was better than that of the other models and comparable with that of expert urologists. Notably, it showed superior diagnostic accuracy when tumors occupied >10% of the image. Conclusions: Our findings demonstrate the value of stepwise organic transfer learning in applications with limited data sets for training and further confirm the value of AI in medical diagnostics. Here, we applied deep learning to develop a tool to detect bladder tumors with an accuracy comparable with that of a urologist. To address the limitation that few bladder tumor images are available to train the model, we demonstrate that pretraining with general and gastroscopic images yields superior results.

Purpose: Although cell-free DNA (cfDNA) testing is expected to drive cancer precision medicine, little is known about the significance of detecting low-frequency variants in circulating cell-free tumor DNA (ctDNA) in castration-resistant prostate cancer (CRPC). We aimed to identify genomic profile including low-frequency variants in ctDNA from patients with CRPC and investigate the clinical utility of detecting variants with variant allele frequency (VAF) below 1%. Experimental Design: This prospective, multicenter cohort study enrolled patients with CRPC eligible for treatment with abiraterone or enzalutamide. We performed targeted sequencing of pretreatment cfDNA and paired leukocyte DNA with molecular barcodes, and ctDNA variants with a VAF ≥0.1% were detected using an in-house pipeline. We investigated progression-free survival (PFS) and overall survival (OS) after different ctDNA fraction cutoffs were applied. Results: One hundred patients were analyzed (median follow-up 10.7 months). We detected deleterious ATM, BRCA2, and TP53 variants even in samples with ctDNA fraction below 2%. When the ctDNA fraction cutoff value of 0.4% was applied, significant differences in PFS and OS were found between patients with and without defects in ATM or BRCA2 [HR, 2.52; 95% confidence interval (CI), 1.24–5.11; P = 0.0091] and TP53 (HR, 3.74; 95% CI, 1.60–8.71; P = 0.0014). However, these differences were no longer observed when the ctDNA fraction cutoff value of 2% was applied, and approximately 50% of the samples were classified as ctDNA unquantifiable. Conclusions: Detecting low-frequency ctDNA variants with a VAF <1% is important to identify clinically informative genomic alterations in CRPC.

Background: To explore possible solutions to overcome chronic Bacillus Calmette–Guérin (BCG) shortage affecting seriously the management of non-muscle invasive bladder cancer (NMIBC) in Europe and throughout the world, we investigated whether non-maintenance eight-dose induction BCG (iBCG) was comparable to six-dose iBCG plus maintenance BCG (mBCG). Methods: This observational study evaluated 2669 patients with high- or highest-risk NMIBC who treated with iBCG with or without mBCG during 2000–2019. The patients were classified into five groups according to treatment pattern: 874 (33%) received non-maintenance six-dose iBCG (Group A), 405 (15%) received six-dose iBCG plus mBCG (Group B), 1189 (44%) received non-maintenance seven−/eight-dose iBCG (Group C), 60 (2.2%) received seven−/eight-dose iBCG plus mBCG, and 141 (5.3%) received only ≤5-dose iBCG. Recurrence-free survival (RFS), progression-free survival, and cancer-specific survival were estimated and compared using Kaplan–Meier analysis and the log-rank test, respectively. Propensity score-based one-to-one matching was performed using a multivariable logistic regression model based on covariates to obtain balanced groups. To eliminate possible immortal bias, 6-, 12-, 18-, and 24-month conditional landmark analyses of RFS were performed. Results: RFS comparison confirmed that mBCG yielded significant benefit following six-dose iBCG (Group B) in recurrence risk reduction compared to iBCG alone (groups A and C) before (P < 0.001 and P = 0.0016, respectively) and after propensity score matching (P = 0.001 and P = 0.0074, respectively). Propensity score-matched sequential landmark analyses revealed no significant differences between groups B and C at 12, 18, and 24 months, whereas landmark analyses at 6 and 12 months showed a benefit of mBCG following six-dose iBCG compared to non-maintenance six-dose iBCG (P = 0.0055 and P = 0.032, respectively). There were no significant differences in the risks of progression and cancer-specific death in all comparisons of the matched cohorts. Conclusions: Although non-maintenance eight-dose iBCG was inferior to six-dose iBCG plus mBCG, the former might be an alternative remedy in the BCG shortage era. To overcome this challenge, further investigation is warranted to confirm the real clinical value of non-maintenance eight-dose iBCG.

Cold-inducible RNA-binding protein (CIRP) and RNA-binding motif protein 3 (RBM3) are two evolutionarily conserved RNA-binding proteins that are structurally related to hnRNPs and upregulated in response to moderately low temperatures in mammalian cells. Although contributions of splicing efficiency, the gene promoters activated upon mild hypothermia and the transcription factor Sp1 to induction of CIRP have been reported, precise mechanisms by which hypothermia and other stresses induce the expression of mammalian cold-inducible proteins (CIPs) are poorly understood. By screening the serine/arginine-rich splicing factors (SRSFs), we report that the transcript and protein levels of SRSF5 were increased in mammalian cells cultured at 32 °C. Expression of SRSF5 as well as CIRP and RBM3 were also induced by DNA damage, hypoxia, cycloheximide and hypotonicity. Immunohistochemical studies demonstrated that SRSF5 was constitutively expressed in male germ cells and the level was decreased in human testicular germ cell tumors. SRSF5 facilitated production of p19 H-RAS, and increased sensitivity to doxorubicin in human U-2 OS cells. Induction of CIPs was dependent on transient receptor potential vanilloid 4 (TRPV4) channel protein, but seemed independent of its ion channel activity. These findings indicate a previously unappreciated role for the TRP protein in linking environmental stress to splicing.

We analyzed the efficacy and toxicity of cabazitaxel (CBZ) at high and low initial doses in Japanese patients with docetaxel-resistant castration-resistant prostate cancer (CRPC). We retrospectively evaluated 118 patients who received CBZ for docetaxel-resistant CRPC in 10 university hospitals in Japan between 2014 and 2016. The rate of decrease of prostate-specific antigen (PSA), adverse events, progression-free survival (PFS), and overall survival (OS) were compared between patients receiving initially high (≥22.5 mg/m2, n = 36) and low (≤20 mg/m2, n = 80) CBZ doses. Factors associated with survival and grade 4 neutropenia were evaluated. PSA values decreased by > 50% in 22 patients (19%), with a higher frequency in the high-dose group than in the low-dose group (29 and 14%, P = 0.073). The median PFS time for the all-patient, high- and low-dose groups was 2.8 months (95% confidence interval [CI] 1.9–4.4), 2.1 months (1.2–5.5), and 3.0 months (2.0–4.4), respectively (P = 0.904). The median OS times were 16.3 months (95% CI 9.7–30.9), 30.9 months (11.8–47.4), and 10.2 months (8.6–20), respectively (P = 0.020). In multivariate analyses, PFS was significantly associated with existing bone metastasis at diagnosis (P = 0.005) and OS with PSA > 100 ng/ml (P = 0.007), hemoglobin < 12 g/dl (P = 0.030), and low initial CBZ dose (P = 0.030). Grade 4 neutropenia occurred in 53 patients (45%) and was associated with a low CBZ dose (hazard ratio 0.21, 95% CI 0.08–0.59, P = 0.002). CBZ at a higher initial dose may have similar response rate and response duration, but longer survival duration after treatment with higher toxicity than a lower initial dose for docetaxel-resistant CRPC in Japanese patients.