Table 1 Summary of studies about AI-based systems used for Ki-67 assessment in PanNENs.

From: Ki-67 assessment of pancreatic neuroendocrine neoplasms: Systematic review and meta-analysis of manual vs. digital pathology scoring

AUTHOR, YEAR

COUNTRY

N° OF CASES

GENDER

MATERIAL

GRADING

MANUAL COUNT

DIA

Bagci, 2012

USA-Japan

21

N/A

SRS

WD

EE, ECM, CC/PI

N/S

Remes, 2012

Finland

31a

N/A

N/S

WD

ECM of at least 2000 cells (hotspots)

Publicly available ImmunoRatio software, capturing five different image fields (minimum of 400 tumor cells per picture, altogether 2000 cells)

Fung, 2012

USA

16b

N/A

CB

WD

N/S

Automated Cellular Imaging System III (ACIS, Dako, Carpinteria, CA, USA) at 20x objective in 3 tumor “hotspots”

Goodell, 2012

USA

45

22 M, 22FΩ

SRS

WD

ECM

VIAS (Ventana): count in 1 hotspot; count in 10 consecutive random fields

Tang, 2012

USA

12c

N/A

N/S

WD

1. ECM of >2000 cells

2. EE

Aperio immunohistochemistry nuclear quantitative image analysis (QIA) algorithm analyzing representative images scanned at 20x magnification

Cimic, 2014

USA

28

10 M, 18 F

SRS

WD

EE

Free software available online (Immunoratio.com)

van Velthuysen, 2014

The Netherlands

6d

N/A

N/S

N/A

EE at x20

ImageJ freeware at different magnifications (20x and 40x).

Reid, 2015

USA, Turkey, Japan, Korea

68

33 M, 35 F

N/S

26 G1, 39 G2, 3 G3

1. EE at intermediate power (x10 objective)

2. ECM on the x20 objective

3. CC/PI

4. Careful, extensive, and exhaustive analysis by an expert.

Automated cellular image cytometer (ACISs III, Dako) scanned the entire slide at x4 and 3 hotspots were selected

Kroneman, 2015

USA

97

51 M, 46 F

N/S

N/A

1. EE

2. ECM of at least 500 tumor cells

Automated Cellular Imaging System (ACIS) (Dako) to select 8 to 10 hotspots within the hottest staining region(s) of the tumor present on the slide

Mejias, 2015

USA

21

N/A

N/S

7 G1, 14 G2

N/S

Ventana Image-VIAS

Neely, 2016

USA

24

N/A

CB

N/A

CC/PI, selection of 3 hotspots

Calculation of PI on 3 hotspots with a DIA software algorithm

Burdette, 2016

USA

57

N/A

N/S

WD

CC/PI, selection of 6 hotspots

Whole slide scanning with Aperio ImageScope, manual revision and selection of 6 hotspots, Aperio immunohistochemistry nuclear quantitative analysis algorithm

Jin, 2016

USA

58

33 M, 25 F

CB and SRS

31 G1, 23 G2, 4 G3

CC/PI of at least 500 tumor cells. For cases where TTCN was less than 500 on the entire slide, all tumor cells were counted.

Publicly available ImmunoRatio software. Basic mode was used for analysis

Conemans, 2017

The Netherlands

69

N/A

SRS

57 G1, 11 G2, 1 G3

ECM 2000 cells (hotspot)

Digital quantification of Ki67 LI (PACS, Sectra AB, Linköping, Sweden) on manually selected hotspots

Niazi, 2018

USA

33

N/A

Biopsy

WD

N/S

Deep learning method to automatically differentiate between NET and non-tumor regions based on images of Ki67 stained biopsies

Dere, 2019

Turkey

8e

N/S

N/S

N/A

ECM of 500 to 2000 tumor cells

Software designed by Technology Faculty of the institution

Sajjan, 2019

USA

50f

N/S

N/S

N/A

N/S

Ki67-stained whole slide images were captured and the tumor area with the greatest mitotic activity was manually identified. The Ki67-positive cells were counted in 0.5 mm2 using Ventana Virtuoso software

Owens, 2020

UK

42

N/A

N/S

G1 and G2, NOS

CC/PI, 1 hotspot

Open-source image analysis program QuPath version 0.1.34 analyzing the same hotspot regions used for the manual Ki67 assessments. Each hotspot was classified into tumor and stromal compartments using a detection classifier based on training regions

Saadeh, 2020

Jordan

3g

N/S

N/S

WD

CC/PI of at least 1000 tumor cells

ImageJ

Satturwar, 2020

USA

39h

N/S

CB

N/A

1. EE

2. CC/PI of up to 3 hotspot at ×20 magnification

Aperio immunohistochemistry color convo-luted, nuclearV9 quantitative image analysis algorithm (Leica Biosystems)

Lea, 2021

Norway

21i

N/S

SRS and biopsy

N/A

ECM of 500 to 2000 tumor cells

Visiopharm image analysis software (Hoersholm, Denmark) measured Ki67and PHH3 on IHC slides

including 500 to 2000 tumor cells

Boukhar, 2021

USA

3j

1 M, 2 F

N/S

2 G2, 1 G3

CC/PI of hotspot images

Two DIA platforms: QuantCenter and HALO

TOTAL

13/22 USA, 6/22 Europe, 3/22 Asia and mixed

752

50.5% M, 49.5% F

12 N/S; 4 SRS, 3 CB, 1 biopsy, 2 other

55.3% G1, 40.6% G2, 4.1% G3, NOS

–

–

  1. Abbreviations: AI Artificial intelligence; MC Manual count; DIA Digital image analysis; PanNENs Pancreatic neuroendocrine neoplasms; CB Cell blocks; SRS Surgical resection specimens; NET Neuroendocrine tumor; N/A Not available; EE Eyeball estimation; ECM Eye-counting with microscope; CC/PI Camera captured/printed image; N/S Not specified; WD Well-differentiated; M Male; F Female; PI Proliferation index; PHH3 Phosphohistone H3; IHC Immunohistochemistry, NOS Not otherwise specified.
  2. Notes: aThis study investigated a total of 51 cases, 31 with pancreatic origin and 20 with ileal origin; bThis study investigated a total of 22 cases, 16 with pancreatic origin (including 3 liver metastases) and 6 with gastro-intestinal origin (including 4 liver metastases); cThis study investigated a total of 27 cases, 12 with pancreatic origin, 12 originated from small bowel and 3 with rectal origin; dThis study investigated a total of 73 cases, 2 with gastric origin, 18 originated from small bowel, 8 with colonic origin, 18 with pulmonary origin, and 6 with pancreatic origin and 21 liver metastases; eThis study investigated a total of 50 cases, 26 with gastric origin, 10 with appendiceal origin, 3 with colorectal origin, 3 with ileal origin and 8 with pancreatic origin; fThis study investigated a total of 134 cases, 6 with gastric origin, 64 originated from small bowel, 6 originated from large bowl, 7 with appendiceal origin, 31 with mesenterial origin and 50 with pancreatic origin; gThis study investigated a total of 20 cases, 3 with pancreatic origin, 2 with gastric origin, 2 with duodenal and ampullary origin, 7 with jejunal and ileal origin, 2 with appendiceal origin and 2 with colonic origin; hThis study investigated 50 cases, 39 with pancreatic origin and 11 liver metastases; iThis study investigated a total of 159 cases, 2 with esophageal origin, 9 with gastric origin, 54 originated from small bowel, 1 originated form Meckel’s diverticulum, 31 with appendiceal origin, 21 with pancreatic origin, 15 with colonic origin, 14 with rectal origin and 7 liver metastases and metastases with unknown primary tumor; jThis study investigated a total of 25 cases, 3 with pancreatic origin, 5 with ileal origin, 5 with duodenal origin, 2 with gastric origin, 3 nodal metastases, 1 ileal metastasis, 5 liver metastases and 1 diaphragmatic metastasis; Ωthis study reported data on a total of 45 cases but the total number of patients was 44: there were 22 females (one had two tumors, for a total of 23 tumors) and 22 males.
Back to article page