Your manuscript "Artificial Intelligence as a Second Reader in Prostate Core Needle Biopsies: The Effect on Diagnostic Agreement and Gleason Score" has been reviewed by the Archives of Pathology & Laboratory Medicine. It has been determined, however, that a major revision is necessary before this manuscript can be accepted for publication. The reviewer comments are shown below. Please review these comments and make the requested corrections. Use the link below to submit the revised version. In addition to the reviewer comments below, please do not include any author information within the revised text file. Prior to peer review, the author initials (F.A.) were removed from the text in 2 places and the Acknowledgement was removed. This was done to better ensure a blinded peer review. Please add the full sentences (with the initials) to the cover letter so that information can be added back to the text later. Do not embed the figures and tables within the main text. And the full text should be double-spaced. The title and running title entered in the submission site differ from the title and running title included on the title page. Please add a Context section to the Abstract. This Original Article submission requires a structured abstract of no more than 250 words using the following sections: Context, Objective, Design, Results, and Conclusions. The Objective section should begin as "Objective: To..." Also, the "Conclusion" heading on page 13 should be plural. Please expand the in-text callouts for the figures so all images are included (eg, cite "Figure 2, A and B" instead of "Figure 2"). The P values must be edited per American Medical Association style. Specifically, make sure to express as a capital P in italics and delete all the zeros to the left of the decimal. Exact P values must be stated unless P<.001 or P>.99 in which case one of these values should be stated. P values equal to or greater than .01 should be expressed to 2 significant figures. The figure legends should be listed together at the end of the text, before the reference list. The legends should include the specific magnifications and stains used. Any acronym used in a legend or on a figure should be defined within the legend. Any acronym used in a table should be defined below that table, even if spelled out in the text or a previous table. Each figure within a composite should be labeled, cited in the text, and described in the legend. Figure 2B consists of 2 images and should be labeled as B and C. We require much higher quality figures. Your figures should be submitted in either a TIFF, EPS, Word, or PowerPoint format (depending on the type of figure) instead of JPEG. Attached is a Word document with additional information on the submission of high-resolution figures. Remove the trademark symbol from Figure 1 and use a period instead of comma within numbers for decimals (eg, 3.3% instead of 3,3%). If Figure 1 was originally created in Word or PPT, we suggest you submit in that native format. Line art (such as figures 1, 6A, 6B, and 7) should be submitted at a resolution of 1200 dpi and sized at a 1- or 2-column width (3.37 inches - 7 inches wide). Photomicrographs should be submitted with a resolution of at least 300 dpi and sized at 1-column width for a single image or a 2-column width for a composite. Also, please create composites for figures 2, 3, and 6. Figures 2A-C, the figures could be stacked A over B over C in a 1-column width. A composite for 3A and B should have the images either stacked at a 1-column width or placed next to each other at a 2-column width (A on the left, B on the right). For figures 6A and B, the composite should probably have the 2 figures next to each other at a 2-column width. For all composites, place labels in the lower left corner directly on the images (do not place boxes behind labels). Remove the titles (eg, "Changes in Pathologist ...") across the top of Figures 6A and 6B; this information should be incorporated into the legends included in the main text. Add labels A, B, and C to the lower left corner of each of the 3 graphs in Figure 7. All authors must submit a completed Authorship Form; a blank form is attached for your convenience. These completed forms can be faxed to us at 847-832-8400, scanned and emailed, or sent as a picture taken with a smart phone; please include your manuscript number on all the forms. Please revise the manuscript using the link below: https://archivesofpathology.msubmit.net/cgi-bin/main.plex?el=A5PQ6Izk1A7Vul5I7A9ftd0ii2s8IrXUwkywQmnyeZAQZ When submitting a revised manuscript, you must submit a Revision Worksheet that clearly lists all of the changes you made to your manuscript. The revision worksheet is available at the submission home page under the Required Manuscript Forms link. Please note that you have 3 months in which to submit your revised manuscript. If you wish to request a revision extension, you must contact the editorial office. Make certain to include your manuscript number in your request e-mail. The reviewers are given access to a version of your text that contains line numbers. Since most reviewers use these numbers in their comments, I have attached this version for your convenience. Also attached is a Word document with information to assist authors with the revision of their articles and the loading of their revised files into the submission site. If you have any questions, please contact the Editorial Office. Sincerely, Alain C. Borczuk, MD Editor in Chief Archives of Pathology & Laboratory Medicine ------------------------------------------------------------------------------ Reviewer #1 (Comments for the Author): The authors evaluated the effect of PaigeProstate AI on diagnostic agreement and Gleason Grading in prostate needle biopsy. They found that AI could improve diagnostic agreement and Gleason grading consistency among different pathologists. The manuscript is well written, and the findings are interesting. However, there are some minor issues that remain to be addressed. 1. The pathological features of prostate biopsy specimens need to be provided, such as Gleason Score, tumor volume, PNI, unusual pattern, IDC, AIP, and treatment. 2. What are the backgrounds of 4 pathologists? How many years in practice? How many prostate biopsies do they diagnose per year? 3. Are there any unusual patterns in prostatic adenocarcinoma, such atrophic, hyperplastic, mucinous, and foamy glands? Can the AI recognize these unusual patterns? 4. How does the AI perform in areas of IDC, high-grade PIN, and small cell carcinoma? 5. Can the PI recognize perineural invasion and extraprostatic extension on prostate biopsy? 6. Can the AI recognize treated prostate cancer? 7. If the AI can achieve a negative predictive value of 100%, is it still necessary for pathologists to review the negative biopsy slides? Reviewer #2 (Comments for the Author): The authors report a retrospective study analyzing the utility of artificial intelligence (AI) as a decision support tool for improving diagnostic agreement and Gleason score on needle biopsies of the prostate gland. The cohort included 836 H&E stained digitized images from 60 consecutive cases. The study was performed in 2 parts. In the first part the AI results were compared to the original pathology report which included immunohistochemical findings. In the second part, 4 pathologists independently evaluated the cases without AI assistance. Following a wash out period the pathologists re-evaluated the cases with AI support. Based on the findings, the authors conclude that AI is a useful decision support tool, improving diagnostic agreement and consistency in Gleason grading. This study confirms findings provided in multiple prior publications. While it does not provide any novel information, at the very least it can serve as a validation set for the authors to implement AI in the routine evaluation of needle biopsies of the prostate in their institutions. Grade groups are an important component of in risk group classification in localized prostate cancer for both the American Urological Association (AUA) and the European Urological Association (EUA). It is one of the factors used for risk grouping into low risk, intermediate risk-favorable, intermediate risk-unfavorable and high risk categories. It would be useful to know how AI influenced grade grouping but also risk grouping, since the latter will determine therapy. This information with strengthen the manuscript significantly, although I understand the investigators will need to gather additional data, specifically, pre-biopsy serum PSA and clinical stage. The authors report a consecutive series of 60 cases. This approach does not guarantee sufficient variability in Gleason grades across the entire spectrum. For example, does correlation with Gleason grade and grade groups differ with the size (length) of the focus of tumor? Is the coefficient of correlation lower in cases where the secondary Gleason pattern is minimal (5%-10%)? Another issue not addressed in the is study in the use on immunohistochemistry in establishing a diagnosis, particularly in phase 2 of the project. How often would the reviewing pathologist be influenced to perform immunohistochemistry based on the AI findings? How often would the not perform immunohistochemistry based on the AI findings?