Over 50% of patients diagnosed with breast cancer require chemotherapy as part of their treatment. The chemotherapy is increasingly administered before (neoadjuvant chemotherapy = NAC) rather than after breast surgery. Some benefits of NAC are the possibility of breast-conserving surgery due to the downsizing of the tumor load and reducing the risk of recurrence of the tumor.
Depending on the subtype of breast cancer, in about 8 to 39% of patients, no residual tumor (pathologic complete response = pCR) is found after NAC. Currently, these patients however still receive surgery since the selection of patients with pCR can only be done reliably by histopathological examination after excision of the (former) tumor location. Since patients can suffer from pain and/or deformation of the breast after breast surgery, it would greatly improve the quality of life if surgery could be omitted in these patients.
Recently, deep learning approaches have emerged as a promising tool for predicting the tumor response to NAC, but with limitations like the need for manual tumor segmentation and the usage of 2D MR images. The manual segmentation is burdensome and takes much time since the radiologist needs to do the segmentation slice by slice. Also, after NAC, tumor boundaries are unclear and complicated to determine. This can give errors in the segmentation, which can influence the performance of the prediction. To improve on the assessment of the prediction of breast tumor response to NAC, the goal of this graduation project is to train a convolutional neural network that can predict the response of the tumor to NAC by analyzing 3D pre-NAC and post-NAC MR images without the need for tumor segmentation.
To join the presentation via Microsoft Teams click here