Open-source digital image analysis of whole-slide ...

Providence St. Joseph HealthProvidence St. Joseph Health Digital CommonsSociety for Immunotherapy of Cancer 2018 AnnualMeeting Posters Earle A. Chiles Research Institute Collection

11-2018

Open-source digital image analysis of whole-slidemultiplex immunohistochemistryNikhil LonbergEarle A Chiles Research Institute, Robert W Franz Cancer Center, Providence Portland Medical Center, Portland, OR, USA

Carmen Ballesteros-MerinoEarle A Chiles Research Institute, Robert W Franz Cancer Center, Providence Portland Medical Center, Portland, OR, USA,[email protected]

Shawn JensenEarle A Chiles Research Institute, Robert W Franz Cancer Center, Providence Portland Medical Center, Portland, OR, USA,[email protected]

Bernard A FoxEarle A Chiles Research Institute, Robert W Franz Cancer Center, Providence Portland Medical Center, Portland, OR, USA,[email protected]

Follow this and additional works at: https://digitalcommons.psjhealth.org/sitc2018

Part of the Oncology Commons

This Book is brought to you for free and open access by the Earle A. Chiles Research Institute Collection at Providence St. Joseph Health DigitalCommons. It has been accepted for inclusion in Society for Immunotherapy of Cancer 2018 Annual Meeting Posters by an authorized administrator ofProvidence St. Joseph Health Digital Commons. For more information, please contact [email protected].

Recommended CitationLonberg, Nikhil; Ballesteros-Merino, Carmen; Jensen, Shawn; and Fox, Bernard A, "Open-source digital image analysis of whole-slidemultiplex immunohistochemistry" (2018). Society for Immunotherapy of Cancer 2018 Annual Meeting Posters. 14.https://digitalcommons.psjhealth.org/sitc2018/14

Open-source digital image analysis of whole-slide multiplex immunohistochemistry Nikhil Lonberg1,*, Carmen Ballesteros-Merino1, Shawn M Jensen1, Bernard A. Fox1.

1Earle A. Chiles Research Institute, Providence Cancer Center, Portland, Oregon, USA.

ABSTRACT Successful digital image analysis (DIA) of cancer tissue is accurate and reproducible. These points of emphasis have brought procedures like the tissue microarray (TMA) and hotspot regions of interest (ROI) under scrutiny. The nature in which a pathologist selects TMAs and ROIs is conducive to bias. Whole Slide Imaging (WSI) offers a solution in its unbiased region selection and consideration of a larger tissue sample. However, options for software that can handle such large throughput are scarce. Additionally, while multiplex immunohistochemistry (mIHC) is becoming popular (Feng et al.), documentation of its digital analysis tools remains minimal. The combination of these procedures potentiates a deeper understanding of the tumor microenvironment. This study presents the whole-slide mIHC analysis capabilities of QuPath, an open-source application developed at Queen's University Belfast . Methods. A multiplex fluorescent stain panel was performed on patient samples. The slides were imaged and cells were detected and segmented in QuPath. QuPath parallelizes its workload to manage whole-slide throughput efficiently. Custom scripts were written that exhibit machine-learning and thresholding techniques to aggregate cell phenotype totals. Additionally, cell detection numbers were generated for specific ROIs and compared to a proprietary DIA software. All scripts and protocols in this study are made public for replication and improvement by the community. Results. QuPath's automated cell segmentation and classification were demonstrated as a proof-of-concept for whole-slide multiplex immunohistochemistry analysis. Across an entire slide, cells positive for multiple markers were effectively segmented and properly phenotyped. Conclusion. Open-source applications have become a driving force for innovation and collaboration in the field of digital image analysis. In litigating the strengths and weaknesses of QuPath for whole-slide mIHC analysis we aim to advance the field's knowledge of available software tools and bring attention to necessary points of growth in this rapidly changing industry.

REFERENCES 1. Bankhead P, Loughrey MB, Fernandez JA, Dombrowski Y, McArt DG,

Dunne PD, McQuaid S, Gray RT, Murray LJ, Coleman HG, James JA, Salto-Tellez M, Hamilton PW. Qupath: Open source software for digital pathology image analysis. Sci Rep. 2017;7:1-7.

2. Blom S, Paavolainen L, Bychkov D, Turkki R, Maki-Teeri P, Hemmes A, Valimaki K, Lundin J, Kallioniemi O, Pellinen T. Systems pathology by multiplexed immunohistochemistry and whole-slide digital image analysis. Sci Rep. 2017;7:1-13.

3. Feng Z, Jensen SM, Messenheimer DJ, Farhad M, Neuberger M, Bifulco CB, Fox BA. Multispectral Imaging of T and B Cells in Murine Spleen and Tumor. J Immunol. 2016;196:3943-3950.

CONCLUSIONS

Preliminary results using QuPath open-source software to analyze whole-slide multiplex IHC suggests that it holds promise as an approach to assess complex fluorescent images.

*COMMENT

This study was undertaken as an undergraduate student summer project and may be a component of a senior thesis project

Hurdles for Technology & Opportunity Multiplex IHC technology provides unprecedented opportunities to investigate tissues. However, there are weaknesses in the way cells are enumerated. In an attempt to standardize and better understand best practices for evaluating T cell populations and expression of activation markers, a T cell Activation Marker Proficiency Panel (TAMPP) has been initiated. Populations of human T cells, extensively characterized by FACS, have been prepared into cell blocks. Slides will be sent out containing sections of the characterized T cells for participating labs to perform multiplex IHC, analyze and return results. Results will be summarized, site ID blinded and data shared, along with FACS phenotypes, with the participating labs. This academic-based collaboration is open to participation by all groups (academic & commercial) using any multiplex technology with results prepared for publication. For more info email [email protected]

0 5 0 0 1 0 0 0 1 5 0 0 2 0 0 0

0

5 0 0

1 0 0 0

1 5 0 0

2 0 0 0

P D L 1 +

Q u P a t h

In

Fo

rm

R2

= 0 . 9 0

0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0

0

1 0 0

2 0 0

3 0 0

4 0 0

5 0 0

F o x P 3

Q u P a t h

In

Fo

rm

R2

= 0 . 8 9

0 5 0 0 1 0 0 0 1 5 0 0

0

5 0 0

1 0 0 0

1 5 0 0

C D 8 + P D L 1 -

Q u P a t h

In

Fo

rm

R2

= 0 . 0 0 4

0 5 0 0 1 0 0 0 1 5 0 0 2 0 0 0

0

5 0 0

1 0 0 0

1 5 0 0

2 0 0 0

C D 8 + P D L 1 +

Q u P a t h

In

Fo

rm

R2

= 0 . 9 9

0 1 0 0 0 2 0 0 0 3 0 0 0 4 0 0 0

0

1 0 0 0

2 0 0 0

3 0 0 0

4 0 0 0

O T H E R

Q u P a t h

In

Fo

rm

R2

= 0 . 7 6

0 2 0 0 0 4 0 0 0 6 0 0 0

0

2 0 0 0

4 0 0 0

6 0 0 0

C e l l S e g m e n t a t i o n

Q u P a t h

In

Fo

rm

R2

= 0 . 8 8

0 1 0 2 0 3 0 4 0

0

1 0

2 0

3 0

4 0

F o x P 3

Q u P a t h

In

Fo

rm

R2

= 0 . 0 3 3

0 1 0 0 0 2 0 0 0 3 0 0 0

0

1 0 0 0

2 0 0 0

3 0 0 0

P D L 1 +

Q u P a t h

In

Fo

rm

R2

= 0 . 4 5

0 2 0 4 0 6 0 8 0 1 0 0

0

2 0

4 0

6 0

8 0

1 0 0

C D 8 + P D L 1 -

Q u P a t h

In

Fo

rm

R2

= 0 . 8 5

0 1 0 0 2 0 0 3 0 0 4 0 0

0

1 0 0

2 0 0

3 0 0

4 0 0

C D 8 + P D L 1 -

Q u P a t h

In

Fo

rm

R2

= 0 . 2 6

0 1 0 0 0 2 0 0 0 3 0 0 0 4 0 0 0

0

1 0 0 0

2 0 0 0

3 0 0 0

4 0 0 0

O T H E R

Q u P a t h

In

Fo

rm

R2

= 0 . 8 0

0 1 0 0 0 2 0 0 0 3 0 0 0 4 0 0 0

0

1 0 0 0

2 0 0 0

3 0 0 0

4 0 0 0

C e l l S e g m e n t a t i o n

Q u P a t h

In

Fo

rm

R2

= 0 . 5 8

STROMA TUMOR

Correlation of cell numbers for each phenotype in 4 regions of interest using InForm (y-axis) versus QuPath (x-axis) . Tissue segmented (Stroma – left column, Tumor- right column) with each corresponding software.

DAPI FoxP3 CD8 Cytokeratin PDL1

Supported by: The Harder Family, Robert W Franz, Elsie Franz-Finley, Lynn and Jack Loacker, Wes and Nancy Lematta, Chiles Foundation, Murdock Trust and Providence Medical Foundation

Whole slide mIHC images of NSCLC tissue (left) and Breast Cancer Tissue (right) are shown above. Analysis algorithm was developed for each tissue and used to segment and classify the cells in each tissue using QuPath software.

DAPI FoxP3 CD8 Cytokeratin PDL1

On the left a HNSCC mIHC slide is shown, the region of interest to be analyzed is shown in yellow. In the middle column is a zoomed in image showing mIHC image (top), tissue segmented (red-tumor, green-stroma; middle), and cell classification (bottom). The right column shows InForm analysis of the same region of interest: mIHC image (top) and cell classification (bottom). At the bottom: Analysis algorithm was developed in Head and Neck tissue slide, and it was also applied in Non-Small Cell Lung Cancer (NSCLC) tissue and Breast Carcinoma.

QuPath InForm

Tiss

ue

segm

enta

tio

n

Cel

l cla

ssif

icat

ion

Cel

l cla

ssif

icat

ion

NSCLC Tissue Breast Cancer Tissue

mIH

C

mIH

C