Table 4 The characteristic of datasets from various hospitals, Male: M; Female: F

From: A deep learning based smartphone application for early detection of nasopharyngeal carcinoma using endoscopic images

Characteristics

SZH Dataset

FSH Dataset

LZH Dataset

No. of collected images

119,620

66,716

15,133

No. of patients

10,934

7568

1891

Race of the patients

Asian

Asian

Asian

No. of qualified images

23,788

14,285

1267

No. of patients with definitive diagnosis

7948

4743

1267

Age range/Mean

4–96/37

2–95/35

3–97/46

Sex: n (%)

M: 4019(50.57%)

F: 3929(49.43%)

M: 2638(55.63%)

F: 2105(44.37%)

M: 591(46.65%)

F: 676(53.35%)

Level of hospital

Tertiary

General Hospitals

Tertiary

General Hospitals

Secondary

General Hospitals

Location of hospital

Shenzhen,

Guangdong, China

Foshan,

Guangdong, China

Leizhou,

Guangdong, China

The operating systems and endoscopes for obtaining images

System: Olympus CV-170 Digital Endo-Vision

Endoscope: Matrix E2, XION GmbH

System: STORZ Xenon Nova 300

Endoscope: STORZ Image 1 HUB H3-Z

System: SD-HD668P (Shenda endoscope)

Endoscope: Shenda, J0200G

Image resolution(pixel)

~850 × 850

~720 × 570

~480 × 540