ECP 2017 - Abstract Book

SOX 9 expression in normal endometrium and endometrial

carcinoma.

Method:

Immunostain of SOX 9 was performed in 32 cases of endometrial

carcinoma and 30 cases of benign endometrium. Stains of Sox 9 were scored

as weak (no staining 0, weak staining 1+) and strong (Stains 2+ and 3+)

nuclear staining and results were interpreted by two pathologists.

Results:

The expression of Sox 9 was only identified in nuclei of endometrial

glandular epithelium and not in endometrial stroma. Over expression of Sox 9

is identified in majority of endometrial carcinoma 66 % (21/32) while the

staining pattern of Sox 9 is weak or none in majority of normal endometrium.

Only 19 % (6/31) of normal endometrium has strong but focal nuclear stain-

ing (

= 0.0003).

Conclusion:

This study shows significant overexpression of SOX 9 protein

in nuclei of endometrial carcinoma compared to normal endometrium. This

overexpression may play a role in the carcinogenesis of endometrial carcino-

ma.

OFP-04-012

Differentiating primary pulmonary squamous cell carcinoma from

squamous cell carcinoma of the cervix metastatic to the lung:

Histological and immunohistochemistry study

D. Montezuma

, R. Vieira, A. L. Cunha, R. Henrique, C. Bartosch

IPO-Porto, Pathology, Portugal

Objective:

Distinction between primary and metastatic pulmonary squamous

cell carcinoma(SCC) is difficult. We aimed to define histological and immu-

nohistochemical features helpful for the differential diagnosis of pulmonary

SCC metastases originating from the cervix.

Method:

Retrospective review of primary pulmonary SCC(

= 22) and lung

metastases from cervix SCC(

= 21) diagnosed at IPO-Porto(1995

–

2017).

Clinicopathological data was retrieved and histological features were re-

evaluated. Immunohistochemistry study with ER, PR, TTF-1, p63, and

CK7 was performed. Statistical analysis was done to compare groups.

Results:

Compared to patients with primary pulmonary SCC, those with

cervix SCC lung metastases were younger(mean age: 54 years vs. 69 years)

and more frequently presented with multinodular disease(31% vs. 14 %).

Median time between primary cervical SCC and lung metastasis was

30 months. Histologically, cervix SCC lung metastases predominantly fea-

tured large regular nests with moderate cytological atypia, while primary lung

SCC often presented small irregular nests with severe atypia. Keratinization

was more frequent in cervix SCC metastases(60% vs. 32 %).

Immunostaining showed diffuse p63 and absence of TTF-1, ER and PR

expression in all cases. CK7 was more often positive(76% vs. 64 %), with

stronger and more diffuse staining in metastatic SCC(median Allred-

score:7vs.4).

Conclusion:

An integrated approach, comprising clinical, histological and

immunohistochemical features, is essential for the differential diagnosis be-

tween primary and metastatic SCC originating from the cervix.

OFP-05-001

Template based synoptic reporting improves oncological pathology

reports regarding data content and clarity of data layout

K. Aumann

, P. Bronsert, J. Asberger, G. Gitsch, B. Passlick, U.

Wetterauer, D. Hauschke, G. Kayser, M. Werner

Surgical Pathology, Dept. of Pathology, Freiburg, Germany

Objective:

Traditionally, pathology reports have been textual with a high

degree of variability. In part, they miss some of the information needed for

e.g. therapy decision. Here, we describe a TNM-adapted toolset including a

PIS-integrated structured template that contributes to improving pathology

reports of prostate, lung, and breast resection specimens.

Method:

Pathology reports of oncological prostate (

= 1049), lung

(

= 878) and breast (

= 4181) resection specimens were classified into

descriptive reports (DR), structured reports (SR), and template based

synoptic reports (TBSR). The report types were compared regarding the

content of organ specific essential data, summarized in an essential data

score (EDS), and regarding the time a gynecologist needed to detect all

essential data within a subset of breast specimens reports.

Results:

All 11 ED of prostatectomy specimens were included in 2.7 % of

DR, 43.5 % of SR and 97.2 % of TBSR with a statistically highly signif-

icant difference (

< 0.0001). A high-score EDS of 10 was measured in

2.6 % of DR of lung resection specimens, 16.4 % of SR, and 88.4 % of

TBSR (

< 0.0001). Regarding reports of breast resection specimens a

full-score EDS of 9 was seen in 4 % of DRs, in 21.4 % of SRs, and in

72.3 % of TBSRs (

< 0.0001). Mean data detection time decreased sig-

nificantly from 26 to 20 and 14 s in DRs, SRs, and TBSRs, respectively.

Conclusion:

Structuring improve the quality of pathology reports

reflected by an increased content of essential data and a high clarity of

data layout resulting in a rapid detection of essential data by clinicians.

OFP-05-002

Automatic cell detection in the bone marrow by a convolutional neu-

ral network-based approach

C.-A. Weis

, A. Marx, F. Zoellner

University Medical Centre, Institute of Pathology, Mannheim, Germany

Objective:

Automatic detection of cells is an important step for the anal-

ysis of tissue composition and spatial arrangement. Partially overlapping

cell nuclei (either two are erroneously detected as one or one bizarre

configured is detected as two) are one major challenge, which we intend

to address by a convolutional-neural network based cell segmentation

approach.

Method:

The approach is implemented in MATLAB with the free-

available convolutional neuronal network toolbox MatConvNet

(www.vlfeat.org/matconvnet/)

Results:

The detection process can be divided into several steps: i) By

applying a sliding window the input image is decomposed into many small

tiles. ii) These tiles are analyzed by a convolutional network (pre-trained to

classify the tile content) leading to a tile-wise probability value for contain-

ing one single cell. iii) To take different cell-sizes into account, the steps i

and ii are repeated with sliding windows of different sizes. iv) To avoid

double-detection and to optimize the bounding box a combination of non-

maxima-suppression and a second convolutional network (pre-trained to

classify the deviation of the box size and position) is applied.

Conclusion:

The results will be compared to manual segmentation and its

robustness against image variations (e.g. illumination changes) will be the

objective of ongoing investigations.

OFP-05-003

Feasibility of real-time digital pathology by the Panoptiq

™

imaging

system compared with conventional light microscopy in diagnosing

cervicovaginal cytology cases

R. Groen

, A. Yoshikawa, K. Abe, H.-S. Yoon, Y. Akazawa, T. Nitanda,

J. Fukuoka

Nagasaki University Hospital, Dept. of Pathology, Japan

Objective:

Digital pathology has been increasingly gaining the attention

of pathologists worldwide. However, the application of digital cytology is

relatively unexplored. The Panoptiq

™

imaging system enables the oper-

ator to combine low-power panoramic digital images with high-power Z-

stacks of regions of interest with a significantly smaller image size than

that obtained by whole slide scanning. This study aimed to evaluate the

Monday, 4 September 2017, 17:15

–

19:15, G106-107

OFP-05 IT in Pathology

Virchows Arch

(

2017

)

471

(

Suppl 1

–

S352

S14