Methods and apparatus for predicting benefit from immunotherapy using tumoral and peritumoral radiomic features

An apparatus for predicting response to immunotherapy, comprising:

• a processor;
• a memory that stores a set of radiological images of a region of tissue demonstrating cancerous pathology;
• an input/output interface;
• a set of circuits, where the set of circuits includes an image acquisition circuit, a tumoral definition circuit, a peritumoral segmentation circuit, a radiomics circuit, and a classification circuit;
• and
• an interface that connects the processor, the memory, the input/output interface, and the set of circuits:
• where the image acquisition circuit accesses a radiological image from among the set radiological images of a region of tissue demonstrating cancerous pathology, where the radiological image has a plurality of pixels, and where the radiological image includes an annotated region of interest (ROI);
• where the tumoral definition circuit generates a tumoral surface boundary that defines a tumoral volume, based on the ROI;
• where the peritumoral segmentation circuit generates a peritumoral region based on the tumoral surface boundary, and that segments the peritumoral region into a plurality of annular bands;
• where the radiomics circuit extracts a set of discriminative features from the tumoral volume and at least one of the plurality of annular bands based, at least in part, on a discriminability associated with a feature, and a stability associated with the feature;
• and
• where the classification circuit includes a machine learning classifier that classifies the ROI as a responder or a non-responder, based, at least in part, on the set of discriminative features, where the machine learning classifier is a linear discriminant analysis (LDA) classifier, a support vector machine (SVM) classifier, a random forest classifier, or a quadratic discriminant analysis (QDA) classifier.

A computer-readable storage device storing executable instructions that, in response to execution, cause one or more processors of a personalized cancer treatment system to perform operations comprising:

• accessing a set of pre-surgery three-dimensional (3D) computed tomography (CT) images of a region of tissue demonstrating non-small cell lung cancer (NSCLC), where a member of the set of CT images includes a plurality of pixels, and where a member of the set of CT images has a slice thickness;
• annotating a region of interest (ROI) in the image using a 3D slicer approach;
• defining a nodule surface based on the ROI by extracting the nodule surface in three dimensions using an active contour approach, where the nodule surface defines a tumoral volume;
• defining a peritumoral volume using a morphologic dilation of the nodule surface, where the morphologic dilation is based, at least in part, on a number of pixels or a distance, where the number of pixels or distance is determined heuristically;
• segmenting the peritumoral volume into at least three annular bands;
• extracting a set of tumoral features from the tumoral volume, and extracting a set of peritumoral features from at least one of the at least three annular bands, based on a level of discriminability associated with a member of the set of tumoral features or the set of peritumoral features, and a level of stability of a member of the set of tumoral features or the set of peritumoral features, where the level of stability of the member of the set of tumoral features or the level of stability of the member of the set of peritumoral features is based, at least in part, on the slice thickness, and where the set of tumoral features and the set of peritumoral features include an entropy of Gabor feature where S=1, θ=π/8, a Kurtosis of Gabor feature where S=1/2, θ=3π/8, a Variance of Law-Laplacian L5×E5 feature, and a Variance of Gabor feature where S=√2/4 and θ=3π/8;
• providing the set of tumoral features and the set of peritumoral features to a linear discriminant analysis (LDA) machine learning classifier;
• receiving, from the LDA machine learning classifier, a classification of the region of tissue as a responder to checkpoint inhibitor immunotherapy or as a non-responder to checkpoint inhibitor immunotherapy, where the classification is based on the set of tumoral features and the set of peritumoral features;
• and
• generating a personalized checkpoint inhibitor immunotherapy treatment plan based, at least in part, on the classification, where the personalized immunotherapy treatment plan includes an immunotherapy recommendation, an immunotherapy dosage value, or an immunotherapy schedule.