Distinguishing minimally invasive carcinoma and adenocarcinoma in situ from invasive adenocarcinoma with intratumoral and peri-tumoral textural features

1. A non-transitory computer-readable storage device storing computer-executable instructions that when executed control a processor to perform operations, the operations comprising:

accessing a radiological image of a region of lung tissue, where the radiological image includes a ground glass (GGO) nodule;

defining a tumoral region by segmenting the GGO nodule, where defining the tumoral region includes defining a tumoral boundary;

defining a peri-tumoral region based on the tumoral boundary;

extracting a set of radiomic features from the peri-tumoral region and the tumoral region;

providing the set of radiomic features to a machine learning classifier trained to distinguish minimally invasive adenocarcinoma (MIA) and adenocarcinoma in situ (AIS) from invasive adenocarcinoma;

receiving, from the machine learning classifier, a probability that the GGO nodule is invasive adenocarcinoma, where the machine learning classifier computes the probability based on the set of radiomic features;

generating a classification of the GGO nodule as MIA or AIS, or invasive adenocarcinoma, based, at least in part, on the probability; and

displaying the classification.

2. The non-transitory computer-readable storage device of claim 1 , where the radiological image is a non-contrast computed tomography (CT) image having a plurality of slices.

3. The non-transitory computer-readable storage device of claim 1 , where the tumoral region has a diameter less than or equal to two centimeters.

4. The non-transitory computer-readable storage device of claim 1 , where the peri-tumoral region is defined by performing a dilation of the tumoral boundary, where the peri-tumoral region includes a plurality of annular rings.

5. The non-transitory computer-readable storage device of claim 1 , where the set of radiomic features includes at least five radiomic features.

6. The non-transitory computer-readable storage device of claim 5 , where the at least five radiomic features includes four tumoral radiomic features, and one peri-tumoral radiomic feature.

7. The non-transitory computer-readable storage device of claim 6 , where the set of radiomic features includes:

a tumoral co-occurrence of local anisotropy gradients (CoLIAGe) feature;

a first tumoral Haralick feature;

a second, different tumoral Haralick feature;

a tumoral Gabor feature; and

a peri-tumoral CoLIAGe feature.

8. The non-transitory computer-readable storage device of claim 7 , where the peri-tumoral CoLIAGe feature is extracted from a first annular ring extending from the tumoral boundary to 3 mm from the tumoral boundary.

9. The non-transitory computer-readable storage device of claim 1 , where the machine learning classifier is a linear discriminant analysis (LDA) classifier.

10. The non-transitory computer-readable storage device of claim 9 , where the LDA classifier computes the probability with an area under a receiver operating curve (AUC) of at least 0.97, a sensitivity of at least 95.5%, and a specificity of at least 94%.

11. The non-transitory computer-readable storage device of claim 1 , where the machine learning classifier is a diagonal LDA classifier, a quadratic discriminant analysis (QDA) classifier, or a support vector machine (SVM) classifier.

12. The non-transitory computer-readable storage device of claim 1 , the operations further comprising training the machine learning classifier.

13. The non-transitory computer-readable storage device of claim 12 , the operations further comprising testing the machine learning classifier on a multi-institutional testing cohort.

14. The non-transitory computer-readable storage device of claim 1 , the operations further comprising:

generating a personalized treatment plan based, at least in part, on the classification; and

displaying the personalized treatment plan.

15. An apparatus for distinguishing lung nodules, comprising:

a processor;

a memory configured to store a digitized image of a region of tissue that includes a ground glass opacity (GGO) nodule;

an input/output (I/O) interface;

a set of circuits; and

an interface that connects the processor, the memory, the I/O interface, and the set of circuits, the set of circuits comprising:

an image acquisition circuit configured to access a diagnostic image of a region of tissue demonstrating a GGO nodule;

a segmentation circuit configured to:

define a tumoral region by segmenting GGO nodule, where segmenting the GGO nodule includes defining a tumoral boundary; and

define a peri-tumoral region based on a dilation of the tumoral boundary;

a radiomic feature circuit configured to:

extract a set of radiomic features from the peri-tumoral region and the tumoral region;

a nodule classification circuit configured to:

compute a probability that the GGO nodule is invasive adenocarcinoma based on the set of radiomic features; and

generate a classification of the GGO nodule as minimally invasive adenocarcinoma (MIA) or adenocarcinoma in situ (AIS), or invasive adenocarcinoma, based, at least in part, on the probability; and

a display circuit configured to:

display the classification.

16. The apparatus of claim 15 , where the set of radiomic features includes:

a tumoral co-occurrence of local anisotropy gradients (CoLIAGe) feature;

a first tumoral Haralick feature;

a second, different tumoral Haralick feature;

a tumoral Gabor feature; and

a peri-tumoral CoLIAGe feature.

17. The apparatus of claim 15 , where the nodule classification circuit is configured to compute the probability that the GGO nodule is invasive adenocarcinoma using a linear discriminant analysis (LDA) machine learning approach.

18. The apparatus of claim 15 , where the diagnostic image is a non-contrast computed tomography (CT) image having a plurality of slices.

19. The apparatus of claim 15 , the set of circuits further comprising a training and testing circuit configured to train the nodule classification circuit to compute the probability that the GGO nodule is invasive adenocarcinoma using a set of training images, where a member of the set of training images is acquired using different imaging parameters than the diagnostic image.

20. A non-transitory computer-readable storage device storing instructions that when executed by a computer control the computer to perform a method, the method comprising:

accessing a computed tomography (CT) image of a region of lung tissue, where the CT image includes a ground glass (GGO) nodule, the CT image having a plurality of slices, a slice having a thickness;

defining a tumoral region by segmenting the GGO nodule, where defining the tumoral region includes defining a tumoral boundary;

defining a peri-tumoral region based on a morphological dilation of the tumoral boundary, where the peri-tumoral region includes at least one annular ring;

extracting a set of radiomic features from the peri-tumoral region and the tumoral region, where the set of radiomic features includes a tumoral co-occurrence of local anisotropy gradients (CoLIAGe) feature, a first tumoral Haralick feature, a second, different tumoral Haralick feature, a tumoral Gabor feature, and a peri-tumoral CoLIAGe feature;

providing the set of radiomic features to a linear discriminant analysis (LDA) classifier trained to distinguish minimally invasive adenocarcinoma (MIA) and adenocarcinoma in situ (AIS) from invasive adenocarcinoma using a set of training images, where a member of the set of training images is acquired using different imaging parameters than the CT image;

receiving, from the LDA classifier, a probability that the GGO nodule is invasive adenocarcinoma, where the LDA classifier computes the probability based on the set of radiomic features;

generating a classification of the GGO nodule as MIA or AIS, or invasive adenocarcinoma, based, at least in part, on the probability; and

displaying the classification.