Neoadjuvant therapy is the standard diagnosis and treatment strategy for locally advanced rectal cancer defined by MRI in order to achieve tumor regression, thus affecting the selection of surgical strategy and circumferential margin, improving the safety of operation and the prognosis of patients. This study focused on the related clinical factors such as tumor regression before and after neoadjuvant therapy, combined with preoperative high-dimensional features such as radiomics, to predict the related factors of tumor regression of locally advanced rectal cancer, and validate it with multicenter. In order to develop an accurate model that can be applied to the real world and stratify the risk of locally advanced rectal cancer patients before treatment.

The patients with locally advanced rectal cancer were collected retrospectively, and the relevant information such as clinical baseline characteristics, imaging data and preoperative/postoperative pathological data were collected and integrated, applying the method of deep learning to construct the model, in order to predict and evaluate the risk factors (invasion of mesorectal fascia, status of cancer nodule, long-term prognosis, tumor recurrence, etc.) which are important in clinical diagnosis and treatment. After the model was established, prospective studies were carried out to validate the model, continue training and enrich the effectiveness of the model.

Inclusion Criteria:

1. Primary rectal adenocarcinoma was diagnosed before operation
2. Preoperative MRI staging was diagnosed as stage cT4 rectal cancer
3. Receiving neoadjuvant therapy before operation (including but not limited to chemotherapy, radiotherapy, immunotherapy or targeted therapy, etc.)
4. Middle and low rectal cancer (the distance between the lower margin of the tumor and the anal margin ≤ 12cm measured by MRI)
5. Receive radical resection of rectal cancer

Exclusion Criteria:

1. lack of clinical TNM staging information or other key clinical information about MRI
2. the imaging quality is poor or the MRI image has artifacts

A total of 136 patients (Well T-downstage: 28.68%, 39/136) were included in the training cohort and 50 (Well T-downstage: 30%, 15/50) patients in the test cohort.

Please contact us by Email