Premium
Diagnostic imaging of focal calvarial bone lesions. Comparison of statistical and neural network models [in Spanish]
Author(s) -
Arana Estanislao
Publication year - 1998
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.598206
Subject(s) - receiver operating characteristic , cranial vault , medicine , radiology , logistic regression , computed tomography , medical imaging , artificial neural network , statistical analysis , nuclear medicine , artificial intelligence , skull , computer science , mathematics , statistics , surgery
This thesis assesses the accuracy of logistic regression (LR) and artificial neural networks (ANN) in the diagnosis of calvarial lesions using computed tomography (CT). The importance of the different features needed for the diagnosis in both models is also analyzed. The models were developed from CT images of 167 patients (aged 0.5–84 years) with calvarial lesions as their only known disease. All patients were studied with plain films and CT. Other imaging techniques were used when available. The clinical and CT data were used for developing LR and ANN models. Both models were tested with the jacknife (leave‐out‐one) method. The best ANNs were obtained varying iterations and hidden neurons by selecting the one with higher area under the receiver operating characteristic curve (ROC). The final results of each model were compared by means of area under ROC curves. The lesions were 73.1% benign and 26.9% malignant, although there was no statistically significant difference between LR and ANN in differentiating benign and malignant lesions. In characterizing every histologic diagnosis, ANN was statistically superior to LR ( p <0.001). ANNs were demonstrated to be adequate in diagnosing the most common lesions in the cranial vault and in helping the radiologist with misleading and infrequent appearances. ANN discovered hidden interactions among variables that were missed in the statistical analysis.