DIGITAL PATHOLOGY
Fig 3. AI image analysis with a macrodissection algorithm highlights tumour content in an easy-to-read heatmap. A pathologist has drawn the yellow annotation to highlight which region should be dissected.
scoring systems like those used for clinical reporting of IHC DAB-stained patient samples. With high case volumes and important treatment decisions at stake, an overhaul of the methods utilised to evaluate IHC expression is in order. Automated, AI-powered IHC evaluation algorithms enable pathologists to quantify IHC expression quickly and reliably in a standardised way.12,13
For instance, AI can scan and
evaluate the expression of breast or lung cancer biomarkers in an accurate and reproducible manner (Fig 2). These algorithms can provide both visual and quantitative feedback about expression levels, can be run before the pathologist receives the case, and results can easily be incorporated into the final synoptic report. This both improves the process of IHC evaluation and enables pathologists to focus on more challenging cases that require human expertise, such as borderline expression cases.
Macrodissection workflows Genetic analysis of cancer is becoming increasingly commonplace and molecular testing is likely to become even more utilised as we deepen our understanding of cancer biology and
continue to develop drugs targeted at specific genetic aberrations.14
However,
the tissue fixation process can have a degradative effect on the genetic material within a sample and formalin-fixed, paraffin-embedded tissues often contain an admixture of both neoplastic and non-neoplastic tissues. Samples sent for genetic testing require further dissection to remove the non-neoplastic tissue from downstream analysis and concentrate tumour genetic content. To facilitate this, manual macrodissection workflows were created and are currently being used in many anatomic and molecular pathology laboratories. Manual macrodissection workflows involve pathologists hand-annotating tumour content on slides using a pen and microscope and estimating the percentage of tumour cells within the region marked for molecular analysis. The annotated slides are then used by laboratory technicians to guide the dissection of unstained serial sections by eye using a razor blade. This process is highly manual, requires a great deal of technical skill, is prone to error, and is not auditable.
Multiple steps in the process can be improved through the implementation
of digital pathology and automation through AI. AI algorithms can quantify and highlight tumour content present on digitised WSIs, providing a useful guide for pathologists to use as they annotate tumour to be dissected (Fig 3). These annotations can then be exported to a robotic slide macrodissector, which can receive digitised annotations and use them to guide robotic macrodissection, with accuracy of up to 100 µm in some cases. In head-to-head comparisons, automated slide macrodissection has been shown to yield just as much genetic content as manual methods, with improved precision and enhanced sample quality.15
Furthermore, the
manual estimation of tumour content has been shown to be highly subjective and suffers from intra- and inter-reporter variability.16,17
This can result in samples
being sent for molecular testing which do not contain the required amount of concentrated tumour cells for a successful test, thus wasting funds on a failed expensive molecular test and wasting precious tissue. On the other hand, patient samples that contain a high level of concentrated tumour cells within the H&E-stained tissue section can be sent directly for molecular testing, forgoing the requirement for pathologists to annotate the region for dissection.
The automated AI algorithms that quantify tumour content within WSIs allow a fully quantifiable and standardised method to report the percentage of tumour cells within a patient sample. This efficiency gain can be further augmented by automatically sending slides, based on automated tumour content analysis, to specific worklists in a laboratory’s image management system, thus bypassing the requirement for pathologist analysis. For example, if a sample is below the threshold of tumour concentration for a specific molecular test it can be automatically marked as inviable, or if
WWW.PATHOLOGYINPRACTICE.COM FEBRUARY 2025
47
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 |
Page 57 |
Page 58 |
Page 59 |
Page 60
