Evaluation of Artificial Intelligence-Based Gleason Grading Algorithms "in the Wild".

The biopsy Gleason score is an important prognostic marker for prostate cancer patients. It is, however, subject to substantial variability among pathologists. Artificial intelligence (AI)-based algorithms employing deep learning have shown their ability to match pathologists' performance in assigning Gleason scores, with the potential to enhance pathologists' grading accuracy. The performance of Gleason AI algorithms in research is mostly reported on common benchmark data sets or within public challenges. In contrast, many commercial algorithms are evaluated in clinical studies, for which data are not publicly released. As commercial AI vendors typically do not publish performance on public benchmarks, comparison between research and commercial AI is difficult. The aims of this study are to evaluate and compare the performance of top-ranked public and commercial algorithms using real-world data. We curated a diverse data set of whole-slide prostate biopsy images through crowdsourcing containing images with a range of Gleason scores and from diverse sources. Predictions were obtained from 5 top-ranked public algorithms from the Prostate cANcer graDe Assessment (PANDA) challenge and 2 commercial Gleason grading algorithms. Additionally, 10 pathologists (A.C., C.R., J.v.I., K.R.M.L., P.R., P.G.S., R.G., S.F.K.J., T.v.d.K., X.F.) evaluated the data set in a reader study. Overall, the pairwise quadratic weighted kappa among pathologists ranged from 0.777 to 0.916. Both public and commercial algorithms showed high agreement with pathologists, with quadratic kappa ranging from 0.617 to 0.900. Commercial algorithms performed on par or outperformed top public algorithms.

A Prostate Cancer "Nimbosus": Genomic Instability and SChLAP1 Dysregulation Underpin Aggression of Intraductal and Cribriform Subpathologies.

BACKGROUND: Intraductal carcinoma (IDC) and cribriform architecture (CA) represent unfavorable subpathologies in localized prostate cancer. We recently showed that IDC shares a clonal ancestry with the adjacent glandular adenocarcinoma. OBJECTIVE: We investigated for the co-occurrence of "aggression" factors, genomic instability and hypoxia, and performed gene expression profiling of these tumors. DESIGN, SETTING, AND PARTICIPANTS: A total of 1325 men were treated for localized prostate cancer from four academic institutions (University Health Network, CHU de Québec-Université Laval, Memorial Sloan Kettering Cancer Center [MSKCC], and Erasmus Medical Center). Pathological specimens were centrally reviewed. Gene copy number and expression, and intraprostatic oxygenation were assessed. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: IDC/CA was separately assessed for biochemical relapse risk in the Canadian and MSKCC cohorts. Both cohorts were pooled for analyses on metastasis. RESULTS AND LIMITATION: Presence of IDC/CA independently predicted for increased risks of biochemical relapse (HRCanadian 2.17, p<0.001; HRMSKCC 2.32, p=0.0035) and metastasis (HRpooled 3.31, p<0.001). IDC/CA+ cancers were associated with an increased percentage of genome alteration (PGA [median] 7.2 vs 3.0, p<0.001), and hypoxia (64.0% vs 45.5%, p=0.17). Combinatorial genomic-pathological indices offered the strongest discrimination for metastasis (C-index 0.805 [clinical+IDC/CA+PGA] vs 0.786 [clinical+IDC/CA] vs 0.761 [clinical]). Profiling of mRNA abundance revealed that long noncoding RNA, SChLAP1, was the only gene expressed at >3-fold higher (p<0.0001) in IDC/CA+ than in IDC/CA- tumors, independently corroborated by increased SChLAP1 RNA in situ hybridization signal. Optimal treatment intensification for IDC/CA+ prostate cancer requires prospective testing. CONCLUSIONS: The poor outcome associated with IDC and CA subpathologies is associated with a constellation of genomic instability, SChLAP1 expression, and hypoxia. We posit a novel concept in IDC/CA+ prostate cancer, "nimbosus" (gathering of stormy clouds, Latin), which manifests as increased metastatic capacity and lethality. PATIENT SUMMARY: A constellation of unfavorable molecular characteristics co-occur with intraductal and cribriform subpathologies in prostate cancer. Modern imaging for surveillance and treatment intensification trials should be considered in this adverse subgroup.

Exploring prostate cancer genome reveals simultaneous losses of PTEN, FAS and PAPSS2 in patients with PSA recurrence after radical prostatectomy.

The multifocal nature of prostate cancer (PCa) creates a challenge to patients' outcome prediction and their clinical management. An approach that scrutinizes every cancer focus is needed in order to generate a comprehensive evaluation of the disease, and by correlating to patients' clinico-pathological information, specific prognostic biomarker can be identified. Our study utilized the Affymetrix SNP 6.0 Genome-wide assay to investigate forty-three fresh frozen PCa tissue foci from twenty-three patients. With a long clinical follow-up period that ranged from 2.0-9.7 (mean 5.4) years, copy number variation (CNV) data was evaluated for association with patients' PSA status during follow-up. From our results, the loss of unique genes on 10q23.31 and 10q23.2-10q23.31 were identified to be significantly associated to PSA recurrence (p < 0.05). The implication of PTEN and FAS loss (10q23.31) support previous reports due to their critical roles in prostate carcinogenesis. Furthermore, we hypothesize that the PAPSS2 gene (10q23.2-10q23.31) may be functionally relevant in post-operative PSA recurrence because of its reported role in androgen biosynthesis. It is suggestive that the loss of the susceptible region on chromosome 10q, which implicates PTEN, FAS and PAPSS2 may serve as genetic predictors of PSA recurrence after radical prostatectomy.

Genome-wide investigation of multifocal and unifocal prostate cancer-are they genetically different?

Prostate cancer is widely observed to be biologically heterogeneous. Its heterogeneity is manifested histologically as multifocal prostate cancer, which is observed more frequently than unifocal prostate cancer. The clinical and prognostic significance of either focal cancer type is not fully established. To investigate prostate cancer heterogeneity, the genetic profiles of multifocal and unifocal prostate cancers were compared. Here, we report observations deduced from tumor-tumor comparison of copy number alteration data of both focal categories. Forty-one fresh frozen prostate cancer foci from 14 multifocal prostate cancers and eight unifocal prostate cancers were subjected to copy number variation analysis with the Affymetrix SNP 6.0 microarray tool. With the investigated cases, tumors obtained from a single prostate exhibited different genetic profiles of variable degrees. Further comparison identified no distinct genetic pattern or signatures specific to multifocal or unifocal prostate cancer. Our findings suggest that samples obtained from multiple sites of a single unifocal prostate cancer show as much genetic heterogeneity and variability as separate tumors obtained from a single multifocal prostate cancer.

Cysteine-rich secretory protein 3 and ß-microseminoprotein on prostate cancer needle biopsies do not have predictive value for subsequent prostatectomy outcome.

OBJECTIVES: • To investigate whether cysteine-rich secretory protein 3 (CRISP-3) and/or ß-microseminoprotein (ß-MSP) expression in diagnostic prostate needle biopsies have predictive value for prostate cancer (PC) on radical prostatecomy (RP). • To evaluate their potential clinical implementation in a preoperative setting. PATIENTS AND METHODS: • In total, 174 participants from the European Randomized Study of Screening for Prostate Cancer, Rotterdam section, treated by RP for PC were included in the present study. • CRISP-3 and ß-MSP immunohistochemistry was performed on corresponding diagnostic needle biopsies. • Outcome was correlated with clinicopathological parameters (prostate-specific-antigen, PSA; number of positive biopsies; Gleason score, GS; pT-stage; surgical margins at RP) and significant PC at RP (pT3/4, or GS > 6, or tumour volume ≥ 0.5 mL) in the total cohort (n= 174) and in a subgroup with low-risk features at biopsy (PSA ≤ 10 ng/ml, cT ≤ 2, PSA density <0.20 ng/mL/g, GS < 7 and ≤ 2 positive biopsy cores; n= 87). RESULTS: • ß-MSP and CRISP-3 expression in PC tissue was heterogeneous, with variable staining intensities occurring in the same tissue specimen. • High expression of ß-MSP significantly correlated with GS < 7 at RP; it was not a predictor for significant PC at RP neither in the total group (n= 174; odds ratio, OR, 0.319; 95% confidence interval, CI, 0.060-1.695; P= 0.180), nor in the low-risk group (n= 87; OR, 0.227; 95% CI, 0.040-1.274; P= 0.092). • CRISP-3 expression was not related to clinicopathological parameters, and did not predict significant PC at RP in the total group (n= 174; OR, 1.056; 95% CI, 0.438-2.545; P= 0.904) or the low-risk group (n= 87; OR, 1.856; 95% CI, 0.626-5.506; P= 0.265). CONCLUSIONS: • High ß-MSP expression correlated with low GS in subsequent RP specimens, supporting the view that ß-MSP exerts a tumour-suppressive effect. • No significant prognostic value of ß-MSP or CRISP-3 in prostate needle biopsies for significant PC at RP was found. • ß-MSP or CRISP-3 do not have additional value in the therapeutic stratification of patients with PC.