Date/Time
Date(s) - Wed 10 December
12:00 - 13:00
Location
Seminar Room, Malaghan Institute, VUW
Malaghan Institute seminar
Artificial intelligence in prostate pathology
Artificial intelligence (AI) is an emerging tool in many fields of diagnostic pathology including the assessment of prostate biopsies. Several commercially available AI systems designed specifically for prostate biopsy reporting have been launched. They have the potential to assist in both diagnosis and grading of prostate cancer and also in the assessment of other prognostic factors. A possible benefit of the use of AI is a greater degree of standardization of e.g. grading. AI may also help pathologists by avoiding the accidental underdiagnosis of small lesions. However, it is important that pathologists are aware of the limitations and risks of these tools. AI for clinical use needs to be trained on large, high-quality datasets that have been assessed by experienced pathologists. The quality of AI-generated reports will not be better than that of the training datasets. A problem in prostate pathology is the plethora of benign mimics of prostate cancer and unusual morphological variants of cancer. It is difficult to find case collections that are large enough for the training of the most unusual morphologies. Another problem that has been underestimated in the development of AI for histopathology is the interlaboratory variability of cutting and staining of sections and the variation in data output from different scanners used to produce digital slides. A commercial system developed in a specific setting may not be valid in another hospital. There is a risk that clinicians will be impressed by the precision in AI-based reports of e.g. tumor extent and grade percentages, which may exceed the actual precision of the AI systems and also the clinical relevance of the generated information. We also need to be aware of the risk that users will rely too much on the output of a technical system and lose their confidence as diagnostic pathologists. It is ultimately the responsibility of the reporting pathologist to accept or reject the diagnosis proposed by AI.
Bio: Prof Lars Egevad
• Consultant at Dept of Pathology at the Karolinska University Hospital, Stockholm since 1998.
• Scientist/group head IARC/WHO, Lyon 2006-2008
• Professor of Pathology, the Karolinska Institute, Stockholm since 2010
• Secretary of the International Society of Urological Pathology (ISUP) 2005-2013, President 2015-2017
• Co-author of the 3rd, 4th and 5th editions of the WHO Classification of Tumours of the Urinary System and Male Genital Organs
• Member of the organizing committee of 6 ISUP consensus meetings on urological cancer 2005-2015 and 3 WHO consensus meetings 1999-2010
• Invited presentations, lectures and platform presentations in 47 countries
• Organizer and/or lecturer of more than 50 post-graduate courses in prostate pathology in Sweden, Japan, United Kingdom, Singapore, Germany and USA
• Peer-reviewed publications: 432 (Scopus 2025). Number of citations 39747, H index 92 (Google scholar)
• Awarded The Koss medal 2008 and the Grawitz medal 2022 (ISUP)
• Honorary Fellow of the New Zealand Society of Pathology (HonFNZSP) 2016
• Honorary Fellow of the Royal College of Australasian Pathologists (HonFRCPA) 2022
Microsoft Teams
Join the meeting now
Meeting ID: 428 790 497 847 18
Passcode: XZ9YT7DU
Note that if you are coming from outside the Malaghan Institute, please allow time to sign-in at reception for their start at 12pm.