Combination of radiomic and pathomic features in the prediction of prognoses for tumors

Abstract

The present disclosure, in some embodiments, relates to a method. The method includes using a first machine learning model to generate a first medical prediction associated with a lesion in a medical scan using one or more intra-lesional radiomic features associated with the lesion and the one or more peri-lesional radiomic features associated with a peri-lesional region around the lesion. A second machine learning model is used to generate a second medical prediction associated with the lesion using one or more pathomic features associated with the lesion. A combined medical prediction associated with the lesion is generated using the first medical prediction and the second medical prediction as inputs to a third model.

What is claimed is: 
     
         1 . A method, comprising:
 using a first machine learning model to generate a first medical prediction associated with a lesion in a medical scan using one or more intra-lesional radiomic features associated with the lesion and one or more peri-lesional radiomic features associated with a peri-lesional region around the lesion;   using a second machine learning model to generate a second medical prediction associated with the lesion using one or more pathomic features associated with the lesion; and   generating a combined medical prediction associated with the lesion using the first medical prediction and the second medical prediction as inputs to a third model.   
     
     
         2 . A method, comprising:
 inputting, to a first machine learning model, one or more intra-lesional radiomic features associated with a lesion in a medical scan and one or more peri-lesional radiomic features associated with a peri-lesional region around the lesion, the first machine learning model having been pre-trained to make a medical prediction based on the one or more intra-lesional radiomic features and the one or more peri-lesional radiomic features;   receiving a first medical prediction associated with the lesion from the first machine learning model in response to said inputting;   inputting, to a second machine learning model, one or more pathomic features associated with the lesion, the second machine learning model having been pre-trained to make a medical prediction based on the one or more pathomic features;   receiving a second medical prediction associated with the lesion from the second machine learning model in response to said inputting; and   generating a combined medical prediction associated with the lesion using the first medical prediction and the second medical prediction as inputs to a third model.   
     
     
         3 . The method of  claim 2 , wherein the first machine learning model and the second machine learning model are different from one another. 
     
     
         4 . The method of  claim 3 , wherein the lesion comprises a solid tumor. 
     
     
         5 . The method of  claim 3 , wherein the combined medical prediction comprises a combined prognosis. 
     
     
         6 . The method of  claim 5 , wherein the combined medical prediction is one of disease-free survival (DFS), non-DFS, or a likelihood of DFS. 
     
     
         7 . The method of  claim 3 , wherein the third model comprises a nomogram. 
     
     
         8 . The method of  claim 3 , wherein the third model comprises a machine learning model. 
     
     
         9 . A non-transitory machine readable medium having encoded thereon machine-readable instructions that, when executed, cause a machine to execute the method of  claim 2 .