Predicting Survival after Transarterial Chemoembolization for Hepatocellular Carcinoma Using a Neural Network: A Pilot Study
This repository enables external validation of the artificial neural network published in our article with the same title using the pretrained ANN as described in the manuscript. The parameters included in the model are: Demographics - Gender = 'MALE' or 'FEMALE' - BCLC = Barcelona Clinic Liver Cancer staging system (values: A, B, C, D) - Age = age at second TACE, in years Etiology / concomitant disease - Nicotine abuse, Obesity, Diabetes, Etiology alcohol (in case of liver cirrhosis due to alcohol abuse), Etiology HBV (in case of liver cirrhosis due to chronic hepatitis B infection), Etiology HCV (in case of liver cirrhosis due to chronic hepatitis C infection), Etiology NASH (in case of NASH), Etiology unknown (in case of liver cirrhosis of unknown etiology). All these parameters can take values: 0 = false and 1 = true. Tumor related - Baseline tumor number = number of tumor lesions at baseline - Diffuse tumor = diffuse tumor growth pattern (values: 0 = nodular pattern and 1 = diffuse pattern) - Tumor size 1st, Tumor size 2nd: mRECIST evaluation of cross-sectional imaging prior to first and prior to second TACE - Treatment response: radiological response after first TACE (values: 0 = false and 1 = true) Laboratory / liver function - Sodium 1st, Sodium 2nd, Bilirubin 1st, Bilirubin 2nd, Albumin 1st, Albumin 2nd, AST 1st, AST 2nd, ALT 1st, ALT 2nd, INR 1st, INR 2nd, Thrombocyte count 1st, Thrombocyte count 2nd, AFP 1st, AFP 2nd: laboratory values prior to first and prior to second TACE - Child Pugh score 1st, Child Pugh score 2nd, MELD score 1st, MELD score 2nd: Child Pugh and MELD score prior to first and prior to second TACE Sarcopenia - SMI 1st, SMI 2nd: skeletal muscle index/psoas muscle index measured at the level of the L3 vertebrae prior to first and prior to second TACE Type of TACE: - TACE = 'cTACE' or 'DEB-TACE' Imaging used for response evaluation prior to second TACE: - Imaging = 'CT' or 'MRI'
Steps to reproduce
The python script "ann_evaluate.py" enables external validation of the ANN. The weights of the network are stored in the "results" file. In order to run the script, please download all files. All parameters as described above have to be collected and saved in a comma-separated value file. The file "demo_file.csv" contains a sample of 10 test patients and their respective information. In case of technical problems with the implementation or further questions regarding the network, please feel free to contact the corresponding author.