System validation

The assisted diagnosis process using automated systems is imperfect. The result obtained from a classification system represents a probability rather than a correct answer with irrefutable certainty. Different diagnostic measures are thus employed to verify and assure that the results are repeatable and to validate the ability of a system to identify the presence or absence of disease.

In particular, random cross-validation (tenfold) was used in these experiments. The available data were used for data training (70%), and the remaining data (30%) to test the proposed model³³. It is important to note that the folds were randomly assembled using a shuffle-split methodology in its stratified version to guarantee a proportional distribution in each set³⁴. Each classification approach was evaluated using logistic regression, support vector machine, decision tree, random forest, multilayer perceptron, and TabNet. For assessing the performance of each model, diagnostic measures such as sensitivity, specificity, accuracy, and precision are used. Additionally, the area under the curve (AUC) of the receiver operating characteristics (ROC) was determined for each model^48,49.

TP = true positive

TN = true negative

FP = false positive

FN = false negative