A Method For Multiclass Lymphoma Classification Based on Morphological and Non-Morphological Descriptors

Abstract

Lymphoma is one of the most common types of cancer and its treatment can be more effective if the disease variant is correctly diagnosed. Many works have been done using computer vision and machine learning to classify the images. This work presents lymphoma based on histological a method using simple descriptors and a decision tree-based ensemble classifier, aiming to maintaing the interpretability of the data and understand what information in most important to the classification task. We use morphological and non morphological descriptors extracted from the cells nuclei, a feature selection method based on principal component analysis (PCA), and a gradient boosting decision tree (GBDT) method for multiclass classification. Our approach achieves an average accuracy of 0.932. this result is close to those obtained in the state of the art, while it uses simpler descriptors and better interpretable classification models.