Cancer Genetics and Epigenetics 2024, Vol.12, No.3, 144-156 http://medscipublisher.com/index.php/cge 149 monitoring and updating. As new data become available and treatment protocols evolve, models need to be retrained and validated to maintain their accuracy and relevance. This ongoing process ensures that predictive models remain effective tools for improving patient outcomes in the dynamic landscape of cancer treatment. Figure 1 Deep learning architectures (Adopted from Xu et al., 2019) Image caption: The neural architecture includes ResNet CNNs merged with an RNN, and was trained on baseline and follow-up scans. The input axial slices of 50x50 mm2 centered on the selected seed point were used as inputs to the model. They were spaced 5 mm apart; the center slice is on the same axial slice as the seed point. Deep learning networks are trained on natural RGB images and thus need three image slices for input. The outputs of each CNN model are input into the RNN, with a GRU for time-varying inputs. Masking was performed on certain inputs of the CNN so that the recurrent network takes missed scans into account. The final softmax layer provides the prediction (Adopted from Xu et al., 2019) 5 Case Studies and Applications 5.1 Breast cancer In the treatment of breast cancer, circulating tumor DNA (ctDNA) as a non-invasive biomarker to predict the response of neoadjuvant therapy (NST) is receiving more and more attention. The study by Zhou et al. (2022) showed that the detection and persistence of ctDNA can significantly predict the response effect of NST. In a study, they analyzed 93 genes of 193 patients with early breast cancer, designed a patient specific ctDNA tracking detection program, and found that the presence of ctDNA during the mid-term treatment (MT) was significantly related to the higher residual cancer burden (RCB) (Zhou et al., 2022). The study found that among 145 patients with baseline (BL) plasma samples, 63 (43.4%) detected ctDNA at BL. Among these patients, 25 (39.7%) still detected ctDNA at MT, and 15 (23.8%) still had ctDNA at the end of treatment (EOT). Further analysis showed that out of 31 patients who detected ctDNA during MT, 30 (96.8%) were treatment unresponsive (RCB II and III), and only 1 patient achieved RCB I. In addition, patients without RCB 0 detected ctDNA at MT, only 6.7% of RCB I patients detected ctDNA at MT, while 30.6% and 29.6% of RCB II and III patients detected ctDNA at MT, respectively. From these data, it can be seen that ctDNA detection in the mid-term of NST can serve as a negative marker for predicting tumor response. This means that the persistent presence of ctDNA may indicate ineffective treatment and require adjustment of treatment strategies (Figure 2). The results of this study underscore the potential of liquid biopsy in the treatment of breast cancer. Compared with traditional imaging methods, liquid biopsy has the advantages of non-invasive and repeatable, especially suitable for frequent monitoring of early breast cancer patients. In terms of clinical application, this study supports the effectiveness of ctDNA as an early response marker, which can help identify patients who may not require breast surgery after NST, thereby reducing unnecessary invasive treatment. 5.2 Lung cancer In the treatment of non-small cell lung cancer (NSCLC), the integration method combines multiple data types, including imaging data, molecular markers, and clinical data, to improve the prediction accuracy of treatment
RkJQdWJsaXNoZXIy MjQ4ODYzNQ==