Table 5 Comparison on (a) hardware and (b) computing efficiency.

From: Artificial intelligence-assisted fast screening cervical high grade squamous intraepithelial lesion and squamous cell carcinoma diagnosis and treatment planning

Method

CPU

RAM (GB)

GPU

(a)

Proposed method

Intel Xeon Gold 6134 CPU @ 3.20GHz 16

128

4 GeForce GTX 1080Ti

U-net19

Intel Xeon CPU E5-2650 v2 @ 2.60GHz 16

32

1 GeForce GTX 1080Ti

SegNet20

Intel Xeon CPU E5-2650 v2 @ 2.60GHz 16

32

1 GeForce GTX 1080Ti

AraĂşj et al.17

Intel Xeon E5-2643 @ 3.40 GHz 6

251

4 GeForce GTX Titan-X

Method

Inference time in seconds (per WSIa)

Inference pixels (per second)

 

(b)

Proposed method

210

21,604,662

 

U-net19

4370

1,038,196

 

SegNet20

4024

1,127,500

 

AraĂşj et al.17

219\(^{l}\)

20,681,142\(^{l}\)

 
  1. aThe size of the WSI in this evaluation is 4,536,979,200 pixels (99,600\(\times \)45,552 pixels). bAraúj et al.’s method17 takes 0.07 s for processing a patch with 1392 \(\times \) 1040 pixels; b219 s = \(\left\lfloor \frac{99,600 \times 45,552}{1,392 \times 1040} \times 0.07 \right\rfloor \); 20,681,142 pixels = \(\left\lfloor 1392 \times 1040 \div 0.07 \right\rfloor \).
Back to article page