ARTIFICIAL INTELLIGENCE
The highest cost of delaying AI adoption in pathology
Large initial set up costs have resulted in slow adoption of full digital pathology, and as a consequence the integration of artificial intelligence is also lagging behind other fields of medicine. Here Anna Correas, David Tellez and Diana Rosentul consider the potential costs of this slower adoption of new technology.
While deep learning technologies undeniably present remarkable potential for improving diagnostic accuracy and optimising laboratory processes, the integration of artificial intelligence (AI) within clinical pathology lags significantly behind other medical fields like radiology, which benefited from early adoption of digital imaging. Factors such as the
substantial upfront expenses associated with digital pathology systems and AI integration, coupled with insufficient reimbursement structures, remain significant barriers.1,2
and ethical dimensions, and outlines practical strategies to help bridge the gap between innovation and implementation.
The challenge of inter-observer variability
One of the most pressing issues in pathology is inter-observer variability, the discrepancies in diagnostic interpretation among pathologists examining the same histopathological sample. Studies have documented substantial variations in grading various cancers3-7
assessments in many areas, including liver pathology,8 pathology.10
cervical cytology,9 But what is the cost
of delaying the benefits of these tools? This article explores the hidden costs of this hesitancy, going beyond financial considerations to encompass clinical
and other and renal
The occurrence of these
discrepancies can be as high as 50% in some breast evaluations6
colorectal cancer assessments.7
and 53% in early The latter
study linked such variability to potential alternative treatment strategies in up to 30% of the cases.7
Several factors contribute to this
variability, including: n Experience and training: differences in pathologists’ expertise and training levels can lead to varied interpretations.
n Microtomy and staining quality: subtle variations in tissue preparation can affect visual cues.
n Subjective interpretation of diagnostic criteria: despite established guidelines, their application can be inconsistent due to subjective factors during feature assessment.
n Workload and fatigue: high case volumes can impact attention and consistency.
Figure 1 illustrates this variability in breast cancer grading, showing significant differences even among pathologists within the same centre.4
These
In digital pathology, AI tools can be trained to recognise specific morphological features, providing pathologists with valuable support in analysing digital histopathology samples.
WWW.PATHOLOGYINPRACTICE.COM OCTOBER 2025
discrepancies can result in over- or under- treatment of patients, directly impacting
39
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56
