Using Artificial Intelligence to Predict Rectal Cancer Outcomes

Taichung Veterans General Hospital

Status: Completed

Conditions: Rectal Cancer Stage III

Treatments: Other: As materials for external validation for the buildup model.; Other: As training material for deep learning model.

Study type: Observational

Funder types: Other

Identifiers

NCT05723965

CE21235B

Details and patient eligibility

About

Investigator retrospective collect cases during 2010-2021 diagnosed as rectal adenocarcinoma with high quality CT images. Local advanced rectal cancer cases were labeled as "disease". Nor were defined " normal".

Using artificial intelligence CNN on jupyter notebook with open phyton code to train and develop models capable to recognizing local advanced rectal cancer. Modify the phyton code for better predict rate and help physician to quickly evaluate disease severity for fresh rectal cancer cases.

Full description

From 2010.10.1~2021.12.31, rectal cancer patients with cT3-4 lesion was included. Collect high quality CT images with DICOM files in tumor segment. cT1-2, low rectal lesions, non-CRC cases were not included. Non-contrast and artificial defect images were also excluded. CT images were labeled as" diseased " when CRM were threatened (<2mm). All images were labeled according to judgment of 2 specialist. The data were separated into 2 parts. One for AI model training and testing, another for external validation. The training testing dataset was achieved by deep learning neural network and evaluating model accuracy performance. Then the model was applied into external validation dataset for real-world testing, evaluating coherent rate between AI and the Dr. decision. Furthermore, to see the cancer survival outcomes according to AI model prediction results.

Enrollment: 720 patients

Sex: All

Ages: 20 to 100 years old

Volunteers: No Healthy Volunteers

Inclusion criteria

• clinical staging T3-4 with high quality CT images.

Exclusion criteria

• 1. not primary malignancy lesion
• 2. not localizing rectum
• 3. T1-2 lesion
• 4. non contrast or poor quality images

Trial design

720 participants in 2 patient groups

rectal cancer lesion images for training

Description:

Rectal cancer lesion images. Images with threatened (\<2mm) circumferential margin of rectal cancer were labeled as "diseased". Otherwise, images were labeled as "normal". Using these materials as training materials for AI deep learning model buildup.

Treatment:

Other: As training material for deep learning model.

rectal cancer lesion images for testing.

Description:

Using the buildup AI deep learning models from training cohort. Evaluating prediction rate of the model and analysis survival outcomes.

Treatment:

Other: As materials for external validation for the buildup model.