Rapid Communication
Copyright ©2008 The WJG Press and Baishideng.
World J Gastroenterol. Jan 28, 2008; 14(4): 563-568
Published online Jan 28, 2008. doi: 10.3748/wjg.14.563
Table 1 Dataset of the 253 patients with atrophic body gastritis: original 29 input variables used by the standard artificial neural network (ANN) and input variables selected as the most informative by the T&T and IS System and the TWIST Protocol during the input data optimisation procedures
Input variablesStandard ANNT&T and IS systemTWIST protocol
Positive clinical history for peptic ulcerXXX
Active H pylori infectionXXX
Age at diagnosis of ABG1XX
ABG without associated neoplastic lesionsX
ABG with associated neoplastic gastric lesionsXXX
Referred for anemiaXX
Referred for long-standing dispepsiaX
Referred for dermatological disordersXXX
Referred for endocrinological disordersXX
Referred for neurological disordersXX
Presence of metaplastic atrophyXXX
Presence of multifocal atrophic gastritisXX
Actual smokerXX
Previous smokerXX
Never smokedXX
Positive familiy history for gastric cancerX
Positive family history for peptic ulcerXX
Presence of extrathyroidal autoimmune diseasesXXX
Presence of extragastric neoplasmsXXX
Presence of pernicious anemiaXXX
Presence of iron deficiency anemiaXXX
Absence of anemiaXXX
Mean corpuscular volume1XXX
Fasting gastrin1X
Pepsinogen I1XXX
Presence of parietal cell antibodiesXXX
Total number of input variables used292416
Table 2 Predictive performance of linear discriminant analyses (LDA) and artificial neural networks (ANN) in recognizing atrophic body gastritis patients with and without thyroid disease (%)
SensitivitySpecificityMean accuracy
Standard ANN (29 variables)69.059.864.4
T&T and IS (24 variables)78.8b70.574.7
TWIST protocol (16 variables)81.8d69.975.8