Noise Reduction with Wavelet Transform for Clustering Time Series of Tumor Volumes

Personalizing the drug administration during chemotherapy is a promising direction in healthcare. However, due to the lack of an extensive dataset for optimizing treatment, mathematical models are essential to establish a relationship between the response of the tumor to the drug. The parameters of the mathematical model can be considered as the unique characteristics of the patients, thus after identification the therapy can be individualized. Nonetheless, determining these parameters poses a challenge due to the varied and intricate patterns of tumor growth. In a previous study, a clustering approach was developed to group the different time series of tumor volumes comparing their dynamic by the authors. The approach was tested via in silico dataset. In this work, we tested the method on multiple time series of tumor volumes of real experimental mice, not only virtual ones. The results were evaluated by comparing the median of the parameters and the time series for each cluster to the original testing data. The evaluation revealed that the median discrepancy was a mere 43 mm3. In future applications, this methodology may be employed to categorize patients into clusters based on tumor dynamics, enabling the development of therapeutic strategies tailored not only to individual patients but also to groups sharing similar characteristics.