A Review of Machine-Vision-Based Analysis of Wireless Capsule Endoscopy Video

Table 2

Summary of abnormality types, features and claimed experimental results.

Paper	Pathology				Feature			Experiment results
	Bleeding	Polyp	Tumor	Ulcer	Color	Texture	Shape	Image count	Sensitivity	Specificity
Chen et al. [3]	√				√			1,857,657	71.54%	N/A
Lau and Correia [6]	√				√			1,705	88.30%	N/A
Liu and Yuan [7]	√				√			800	99.73%	98.89%
Giritharan et al. [8]	√				√			275,000	83.1%	93.6%
Penna et al. [9]	√				√			1,111	92%	88%
Karargyris and Bourbakis [10]	√						√	N/A	N/A	N/A
Li and Meng [11, 12]	√			√		√		200	92.6%	91%
Li and Meng [13, 14]				√		√		100	93.28%	91.46%
Karargyris and Bourbakis [15]				√			√	50	100%	67.5%
Iakovidis et al. [16]		√				√		4,000	87.5%	N/A
Karargyris and Bourbakis [17]		√					√	50	75%	73.3%
Li et al. [18]		√					√	300	89.8%	82.5%
Barbosa et al. [19]			√			√		192	98.7%	96.6%
Barbosa et al. [20]			√			√		600	97.4%	97.5%
Karkanis et al. [21]			√			√		N/A	N/A	N/A
Li and Meng [22]			√			√		300	97.33%	96%

Sensitivity = Number of True Postives/(Number of True Postives + Number of False Negatives). Sensitivity of 100% means no positives are incorrectly marked as negative. In other words, the test recognizes all positives.
Specificity = Number of True Negatives/(Number of True Negatives + Number of False Postives). Specificity of 100% means no negatives are incorrectly marked as positive. In other words, the test recognizes all negatives.
N/A: data is not available.