AI Models for Predicting TACE Response and Prognosis in Liver Cancer, 2018-2024

160 patients with hepatocellular carcinoma were used to develop and validate AI models predicting treatment response and prognosis after transarterial chemoembolization. The dataset, created by Jiangqin Ma and last updated in April 2026, includes results for clinical, radiomics, and deep learning models with external validation on an independent cohort of 38 patients. Deep learning models achieved an area under the curve of up to 0.96 in the training set.

Use Cases

• Training machine learning models to predict treatment response based on radiomic features extracted from MRI scans.
• Developing deep learning models for prognostic survival analysis in hepatocellular carcinoma patients.
• Comparing the performance of clinical, radiomics, and deep learning models for a medical prediction task.
• Validating AI model performance on an independent external cohort of patients.

Strengths

• Includes results from 160 patients for model training and testing, plus an external validation cohort of 38 patients.
• Model performance is quantitatively reported with area under the curve (AUC) metrics for clinical (0.79), radiomics (0.84), and deep learning (0.96) models in the training set.
• The dataset description provides specific methodological details, including the use of LASSO for feature selection and ResNet50 for the deep learning architecture.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for specific modeling tasks.
• The data is stored in a DOCX file (274.7 KB), which is a tiny dataset and may not contain the raw underlying image or tabular data.

Provenance

Source

figshare

Collection Method

Retrospective study of patients from April 2018 to September 2024.

Time Range

April 2018 to September 2024

Freshness

Last updated 2026-04-28 04:28:55; freshness should be verified.

Geography

null