基于相对变换的骨肉瘤分类算法

The Classification of Osteosarcoma Based on Relative Transformation

摘要: 作为一种常见的骨科疾病，骨肉瘤属于恶性程度甚高、预后极差且转移较快的骨原发性恶性肿瘤。由于该病多发于青少年且危害很大，因此，早期发现、早期诊断和早期治疗便成为治疗骨肉瘤的关键。将机器学习中的基于近邻的局部分类器引入到骨肉瘤的数据分类中来，极大的提高了分类的自动性以及效果。然而由于骨肉瘤数据可能存在稀疏、噪声和非平衡等问题，如此算法的效果往往不佳。本文根据认知的相对性规律提出了基于相对变换的局部均值分类算法，通过相对变换将数据的原始空间变换到相对空间，在相对的空间中度量数据的相似性更符合人们的直觉，从而提高了数据之间的可区分性，同时在一定条件下相对变换还能抑制噪声的影响。实验结果表明，相对局部均值算法具有非常好的分类效果，可以有效地辅助临床医生。

Abstract: As a common disease in department of orthopedics, osteosarcoma is a malignant tumor with high malignancy and poor prognosis. Because the disease often occurs in young people and is very harmful, therefore, early detection, early diagnosis and early treatment are key to the treatment of osteosarcoma. In this paper, local classifier based the nearest neighbor is introduced into the classification of osteosarcoma data, which greatly improves the classification of the automatic and effect. However when dealing with the sparse, noisy and imbalance data, it cannot guarantee to obtain good performance. Based on the relative cognitive law, this paper proposes a feasible strategy called relative local mean center classifier by using the relative transformation to local mean center classifier. The relative space is constructed which may be more line with people’s in-tuition. It should be indicated that relative transformation can improve the distinguishing ability among data points and diminish the impact of noise on classification. The experimental result shows that relative local mean center classifier has a very good classification effect, and can effec-tively assist clinicians.