TY - JOUR
TI - Machine learning for gastric cancer detection a logistic regression approach
AU - Pouliakis, A.
AU - Foukas, P.
AU - Triantafyllou, K.
AU - Margari, N.
AU - Karakitsou, E.
AU - Damaskou, V.
AU - Koufopoulos, N.
AU - Zoi, T.
AU - Nifora, M.
AU - Gouloumi, A.-R.
AU - Panayiotides, I.G.
AU - Tzivras, M.
JO - International Journal of Reliable and Quality E-Healthcare (IJRQEH)
PY - 2020
VL - 9
TODO - 2
SP - 48-58
PB - IGI Global
SN - 2160-9551, 2160-956X
TODO - 10.4018/IJRQEH.2020040104
TODO - null
TODO - The objective of this study is the investigation of the potential value of a logistic regression model for the classification of gastric cytological data. The model was based on the morphological features of cell nuclei. The aim was the discrimination of benign from malignant nuclei and subsequently patients. Cytological images of gastric smears were analyzed by an image analysis system capable to extract cell nuclear features. Measurements from 50% of the patients were selected as a training set for model creation, while the measurements from the remaining patients were used as test set to validate the results. Furthermore, a model for the classification of individual patients, based on the classification of their cell nuclei has been developed. This approach set gave a correct classification at the level of 90% on the training and test sets on the nuclear level. Concluding the application of morphometric feature selection in combination with logistic regression may offer useful and complementary information about the potential of malignancy of gastric nuclei and patient cases. © 2020 International Journal of Abdominal Wall and Hernia Surgery. All rights reserved.
ER -