Reliability of quantitative imagine analysis in Personalised Immunotherapy

Hansol Lee¹, Tuba Nur Gide^2,3 and James S. Wilmott^2,3
1. The School of Medicine, The University of Notre Dame Australia, Sydney, NSW, Australia
2. Melanoma Institute Australia, North Sydney, NSW, Australia
3. Central Clinical School, The University of Sydney, Sydney, NSW, Australia

Introduction:

Melanoma Institute Australia (MIA) created a Personalised Immunotherapy Platform, currently in Phase II clinical trials, aiming to best guide advanced melanoma treatment based on clinical and pathological factors. However, immunohistochemistry (IHC) pathologist scoring of tumour markers is a major rate limiting step. An algorithm-based quantitative analysis of IHC tumour slides would significantly fast-track pathological assessment. This study investigated whether quantitative image analysis can reliably replace pathologist scoring for the Personalised Immunotherapy Platform.

Methods:

IHC was performed on excised melanoma tumour samples for drug target markers LAG-3 (n= 44), PD-L1 (n = 42) and TIGIT (n = 44) and pathologist scoring was obtained for each marker. The HALO image analysis platform was used to develop a cell segmentation algorithm and threshold for drug target positivity; analysis reports were generated for each IHC image.

Results:

Compared to pathologist scoring, quantitative pathology reported higher median positivity for LAG-3 (1.5% [95% CI 0.6-2.9] vs 1% [95% CI 0.1-1.0]) and lower median positivity for PD-L1 (1% [95% CI 0.24-2.0] vs 5% [95% CI 0.1-10.0]) and TIGIT (9.6% [95% CI 5.9-13] vs 30% [95% CI 20-30]). Using the Wilcoxon test, the median of differences between quantitative analysis and pathologist scoring was lowest for LAG-3 (0.21%, p = 0.04), followed by PDL-1 (-3.18%, p = 0.0003) and TIGIT (-16.52%, p < 0.0001). There was strong correlation between pathologist and HALO scoring for LAG-3 (rs = 0.59, p < 0.0001), PDL1 (rs = 0.80, p < 0.0001) and TIGIT (rs = 0.61, p < 0.0001).

Conclusion:

Quantitative analysis scoring has a promising potential to replace pathologist scoring in the Personalised Immunotherapy Platform, enabling patients to receive their report quicker to best guide further treatment. However, further optimisations are required to reduce the scoring discrepancies.