Classification of lymphomas images with polynomial strategy: An application with Ridge regularization

Histological image analysis through systems to aid diagnosis plays an important role in medicine with supplementary reading for the specialist’s diagnosis. This work proposes a method based on the association of extracted features by fractal techniques, regularization and polynomial classifier. The feature vectors were classified by applying the cross-validation technique with 10 folds. The evaluation of the results occurred through metrics such as accuracy (ACC) and imbalance accuracy metric (IAM). The proposed approach achieved significant results for all metrics with non-Hodgkin lymphoma lesion sets. The proposed approach provided values around 0.97 of IAM and 99% of ACC for investigated groups. These results are considered relevant to studies in the literature and the association of Hermite polynomial and regularization can contribute to the detection of the lesions by supporting specialists in clinical practices.