A scientific publication has recently appeared on an early Health Technology Assessment (HTA) of an AI algorithm for the detection of tumour budding in colorectal cancer. As part of this study, a two-round Delphi survey was conducted amongst an international panel van gastro-intestinale pathologists. The panel assessed a series of conditions that the algorithm must meet in order to enable its use in day-to-day clinical practice.
Of the 29 propositions put forward, consensus was reached on 21. These related to technical, organisational, ethical and legal aspects of implementation. No consensus was reached on the remaining eight statements. For example, there was no agreement on the optimal point at which the AI algorithm should be deployed within the diagnostic process: before, during or after the pathologist’s assessment. This outcome suggests that it is not only the timing of the algorithm’s execution that is important, but also the point at which the information generated by the algorithm is utilised.
There was also no consensus reached on the possibility of completely replacing the pathologist with the algorithm. However, the responses clearly indicated a rejection of this application. This finding is in line with the current view on the use of artificial intelligence within pathology and healthcare, whereby AI is primarily seen as a tool to support experts and not as a replacement for professional expertise.
The results emphasise that the successful implementation of AI in clinical practice requires attention to a range of technical, organisational, ethical and legal preconditions. Furthermore, the findings show that, in various areas, there is not yet full