Background Clinically significant prostate cancer (PCa) diagnosis at MRI requires accurate and efficient radiologic interpretation. Although artificial intelligence may assist in this task, lack of transparency has limited clinical translation. Purpose To develop an explainable artificial intelligence (XAI) model for clinically significant PCa diagnosis at biparametric MRI using Prostate Imaging Reporting and Data System (PI-RADS) features for classification justification. Materials and Methods This retrospective study included consecutive patients with histopathologic analysis-proven prostatic lesions who underwent biparametric MRI and biopsy between January 2012 and December 2017. After image annotation by two radiologists, a deep learning model was trained to detect the index lesion; classify PCa, clinically significant PCa (Gleason score ≥ 7), and benign lesions (eg, prostatitis); and justify classifications using PI-RADS features. Lesion- and patient-based performance were assessed using fivefold cross validation and areas under the receiver operating characteristic curve. Clinical feasibility was tested in a multireader study and by using the external PROSTATEx data set. Statistical evaluation of the multireader study included Mann-Whitney U and exact Fisher-Yates test. Results Overall, 1224 men (median age, 67 years; IQR, 62-73 years) had 3260 prostatic lesions (372 lesions with Gleason score of 6; 743 lesions with Gleason score of ≥ 7; 2145 benign lesions). XAI reliably detected clinically significant PCa in internal (area under the receiver operating characteristic curve, 0.89) and external test sets (area under the receiver operating characteristic curve, 0.87) with a sensitivity of 93% (95% CI: 87, 98) and an average of one false-positive finding per patient. Accuracy of the visual and textual explanations of XAI classifications was 80% (1080 of 1352), confirmed by experts. XAI-assisted readings improved the confidence (4.1 vs 3.4 on a five-point Likert scale; P = .007) of nonexperts in assessing PI-RADS 3 lesions, reducing reading time by 58 seconds (P = .009). Conclusion The explainable AI model reliably detected and classified clinically significant prostate cancer and improved the confidence and reading time of nonexperts while providing visual and textual explanations using well-established imaging features. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Chapiro in this issue.

Objective: To compare the efficacy of multiparametric magnetic resonance imaging (mpMRI)‐directed and micro‐ultrasonography (micro‐US)‐directed biopsy for detecting clinically significant (Grade Group >1) prostate cancer (csPCa). Materials and Methods: A total of 203 patients were prospectively enrolled at three institutions across Germany and Austria in the period from January 2019 to December 2019. During each biopsy, the urologist was blinded to the mpMRI report until after the micro‐US targets had been assessed. After unblinding, targets were then sampled using software‐assisted fusion, followed by systematic samples. The primary outcome measure was non‐inferiority of micro‐US to detect csPCa, with a detection ratio of at least 80% that of mpMRI. Results: A total of 79 csPCa cases were detected overall (39%). Micro‐US‐targeted biopsy detected 58/79 cases (73%), while mpMRI‐targeted biopsy detected 60/79 (76%) and non‐targeted (completion sampling) samples detected 45/79 cases (57%). mpMRI‐targeted samples alone detected 7/79 (9%) csPCa cases which were missed by micro‐US‐targeted and non‐targeted samples. Three of these seven were anterior lesions with 2/7 in the transition zone. Micro‐US‐targeted samples alone detected 5/79 (6%) and completion sampling alone detected 4/79 cases (5%). Micro‐US was non‐inferior to mpMRI and detected 97% of the csPCa cases detected by mpMRI‐targeted biopsy (95% CI 80–116%; P  = 0.023). Conclusions: This is the first multicentre prospective study comparing micro‐US‐targeted biopsy with mpMRI‐targeted biopsy. The study provides further evidence that micro‐US can reliably detect cancer lesions and suggests that micro‐US biopsy might be as effective as mpMRI for detection of csPCA. This result has significant implications for increasing accessibility, reducing costs and expediting diagnosis.

Objective: To compare outcomes of patients with lymph node ( LN )‐positive urothelial carcinoma of the bladder ( UCB ) treated with or without cisplatin‐based combined adjuvant chemotherapy ( AC ) after radical cystectomy ( RC ). Patients and Methods: We retrospectively analysed 1523 patients with LN ‐positive UCB , who underwent RC with bilateral pelvic LN dissection. All patients had no evidence of disease after RC . AC was administered within 3 months. Competing‐risks models were applied to compare UCB ‐related mortality. Results: Of the 1523 patients, 874 (57.4%) received AC . The cumulative 1‐, 2‐ and 5‐year UCB ‐related mortality rates for all patients were 16%, 36% and 56%, respectively. Administration of AC was associated with an 18% relative reduction in the risk of UCB ‐related death (subhazard ratio 0.82, P = 0.005). The absolute reduction in mortality was 3.5% at 5 years. The positive effect of AC was detectable in patients aged ≤70 years, in women, in pT3 –4 disease, and in those with a higher LN density and lymphovascular invasion. This study is limited by its retrospective and non‐randomised design, selection bias, the absence of central pathological review and lack in standardisation of LN dissection and cisplatin‐based protocols. Conclusion: AC seems to reduce UCB ‐related mortality in patients with LN ‐positive UCB after RC . Younger patients, women and those with high‐risk features such as pT3 –4 disease, a higher LN density and lymphovascular invasion appear to benefit most. Appropriately powered prospective randomised trials are necessary to confirm these findings.

Background: Bacillus Calmette-Guerin (BCG) is the standard of care for adjuvant intravesical instillation therapy for intermediate- and high-risk non-muscle invasive bladder cancer (NMIBC) after complete transurethral resection. Increasing evidence suggests that there are marked differences in outcomes according to BCG substrains. BCG-Moreau was recently introduced to the European market to cover the issue of BCG shortage, but there are little data regarding the oncologic efficacy. Methods: We retrospectively analyzed 295 consecutive patients, who received adjuvant intravesical instillation therapy with BCG-Moreau for intermediate- and high-risk NMIBC between October 2007 and April 2013 at a single institution. The end points of this study were time to first recurrence and progression to muscle-invasive disease. Results: Median age was 66 years (interquartile range 59-74, mean 65.9 years). According to the EAU risk group, 76 patients presented with intermediate-risk and 219 patients with high-risk NMIBC. The 5-year recurrence-free survival and progression-free survival rate was 64.8% (95% CI 52.8-74.4) and 81.4% (95% CI 65.2-90.2), respectively. Conclusions: BCG-Moreau is an effective substrain for adjuvant instillation therapies of NMIBC, and outcomes appear to be comparable to series using other substrains. During worldwide shortage of BCG-TICE, Connaught and RIVM, BCG-Moreau may serve as an equally effective alternative.

Background/Objectives: Artificial intelligence (AI), particularly natural language processing (NLP) models such as ChatGPT, presents novel opportunities for patient education and informed consent. This study evaluated ChatGPT’s use as a support tool for informed consent before penile prosthesis implantation (PPI) in patients with erectile dysfunction (ED) following radical prostatectomy. Methods: ChatGPT-4 answered 20 frequently asked questions across four categories: ED and treatment, PPI surgery, complications, and postoperative care. Three senior urologists independently rated information quality using the DISCERN instrument on a Likert scale ranging from 1 (poor quality) to 5 (good quality). Readability was assessed using the Flesch Reading Ease (FRE) and Flesch–Kincaid Grade Level (FKGL) formulas, and inter-rater reliability was measured using intraclass correlation coefficients. Results: The inter-rater reliability coefficient was 0.76 (95% CI 0.71–0.80). Mean DISCERN scores indicated moderate quality: 2.79 ± 0.92 for ED and treatment, 2.57 ± 0.98 for surgery, 2.65 ± 0.86 for complications, and 2.74 ± 0.90 for postoperative care. High scores (>4) were achieved for clarity and relevance, while complex issues, such as risks and alternative treatments, scored the lowest (<2). The FRE scores ranged from 9.8 to 28.39, and FKGL scores ranged from 14.04 to 17.41, indicating complex readability suitable for college-level comprehension. Conclusions: ChatGPT currently provides variable and often inadequate quality information without sufficient comprehensibility for informed patient decisions, indicating the need for further improvements in quality and readability.

The aim of this paper was to compare the perioperative and postoperative results of photoselective vaporization of the prostate with the GreenLight-XPS 180 Watt System (PVP) and transurethral resection of the prostate (TURP). This retrospective study included 140 men who underwent PVP and 114 men who underwent TURP for symptomatic benign prostate enlargement (BPE) between June 2010 and February 2015. The primary outcome measures were the patient reported outcome, operative results, International Prostate Symptom Score-Quality of Life (IPSS-QoL), complication rates, catheterization time, and length of hospital stay. The median follow-up times were 27 months (range 14–44) for the PVP group and 36 months (range 25–47) for the TURP group. The patient characteristics were well balanced in both groups with a median age of 71 years (PVP group) vs. 70 years (TURP group) and a comparable prostate volume (median 50 mL in the PVP group vs. 45 mL in the TURP group). The IPSS-QoL was significantly higher in the PVP group than in the TURP group (median 22 + 4; range 16–27 + 3−5 vs. median 19 + 3; range 15−23 + 3−4; p = 0.02). Men undergoing PVP were more likely to be on anticoagulants (PVP group n = 23; 16% vs. TURP group n = 2; 2%, p < 0.001). The median operation time (OT; min) for both procedures was comparable with 68 min (PVP group; range 53–91) vs. 67 min (TURP group; range 46–85). The rate of severe intraoperative bleeding was significantly lower in the PVP group than in the TURP group (n = 7; 5% vs. n = 16; 14%; p = 0.01). The postoperative catheterization time and length of hospital stay was significantly lower in the PVP group (median 1–2 days; range 1–4) vs. the TURP group (median 2–4 days; range 2–5; both p < 0.001). Complication rates (Clavien-Dindo classification ≥III) based on the follow-up data showed no statistically significant difference between the PVP group and the TURP group (n = 6; 4% vs. n = 6; 5%; p = 0.28). The IPSS on follow-up showed an equivalent reduction in symptoms for both treatment modalities (IPSS-QoL of 5 + 1; range 2–11 + 0−2 for both). There were no differences concerning urge (PVP group n = 3; 2% vs. TURP group n = 3; 3%; p = 0.90) and men were similarly satisfied with the postoperative outcome (PVP group 92% vs. TURP group 87%; p = 0.43). The PVP group was associated with a shorter hospitalization time and showed a reduced risk of bleeding, despite patients remaining on anticoagulants, without increasing the overall operative time. There was no difference in the patient reported outcome for both procedures.