Combination of features from biopsies and scans to predict prognosis in SCLC
The present disclosure relates to a non-transitory computer-readable medium storing computer-executable instructions that, when executed, cause a processor to perform operations, including generating an imaging data set having both scan data and digitized biopsy data from a patient with small cell lung cancer (SCLC). Scan derived features are extracted from the scan data and biopsy derived features are extracted from the digitized biopsy data. A radiomic-pathomic risk score (RPRS) is calculated from one or more of the scan derived features and one or more of the biopsy derived features. The RPRS is indicative of a prognosis of the patient.
1. A non-transitory computer-readable medium storing computer-executable instructions that, when executed, cause a processor to perform operations, comprising:
generating an imaging data set comprising both scan data and digitized biopsy data from a patient having small cell lung cancer (SCLC), wherein the scan data includes data from a radiological scan of the patient;
extracting scan derived features from the scan data;
extracting biopsy derived features from the digitized biopsy data; and
calculating a radiomic-pathomic risk score (RPRS) from one or more of the scan derived features and one or more of the biopsy derived features, wherein the RPRS is indicative of a prognosis of the patient.
2. The non-transitory computer-readable medium of claim 1 , further comprising:
utilizing a machine learning classifier to identify predictive scan derived features from the scan derived features and predictive biopsy derived features from the biopsy derived features; and
calculating the RPRS from both the predictive scan derived features and the predictive biopsy derived features.
3. The non-transitory computer-readable medium of claim 1 , further comprising:
identifying a first lesional mask and a first perilesional mask of the scan data; and
extracting the scan derived features from the first lesional mask and the first perilesional mask.
4. The non-transitory computer-readable medium of claim 1 , wherein the digitized biopsy data comprises digitized Hematoxylin and Eosin (H&E) images.
5. The non-transitory computer-readable medium of claim 1 , wherein the radiological scan comprises an x-ray, a computerized tomography (CT) scan, a magnetic resonance imaging (MRI) scan, or a positron emission tomography (PET) scan.
6. The non-transitory computer-readable medium of claim 1 , further comprising:
generating a preparatory imaging data set to comprise preparatory scan data and preparatory digitized biopsy data from a plurality of preparatory patients having SCLC;
extracting training scan derived features from the preparatory scan data;
extracting training biopsy derived features from the preparatory digitized biopsy data; and
calculating a plurality of training radiomic-pathomic risk scores for the plurality of preparatory patients from the training scan derived features and the training biopsy derived features.
7. The non-transitory computer-readable medium of claim 6 , further comprising:
determining a median risk score of the plurality of training radiomic-pathomic risk scores; and
arranging the patient into one of a plurality of different risk groups by comparing the RPRS to the median risk score.
8. The non-transitory computer-readable medium of claim 1 , further comprising:
determining a predicted response to chemotherapy treatment for the patient using the RPRS.
9. A method of predicting a prognosis of small cell lung cancer (SCLC), comprising:
taking a radiological image of a region of interest of a patient having SCLC, the radiologic image comprising scan data;
taking a tissue sample from the region of interest of the patient;
digitizing the tissue sample to form digitized biopsy data;
extracting scan derived features from the scan data;
extracting biopsy derived features from the digitized biopsy data;
identifying predictive scan derived features from the scan derived features and predictive biopsy derived features from the biopsy derived features; and
calculating an RPRS from both the predictive scan derived features and the predictive biopsy derived features, wherein the RPRS is indicative of a prognosis of the patient.
10. The method of claim 9 , wherein the patient has not received chemotherapy for the SCLC prior to taking the radiological image and the tissue sample.
11. The method of claim 9 , further comprising:
identifying a first lesional mask and a first perilesional mask of the scan data; and
extracting the scan derived features from both the first lesional mask and the first perilesional mask.
12. The method of claim 9 , wherein the predictive scan derived features or the predictive biopsy derived features may comprise Haralick features.
13. The method of claim 9 ,
wherein the scan derived features consist of texture features; and
wherein the biopsy derived features consist of shape features.
14. The method of claim 9 ,
wherein the scan derived features comprise texture features and shape features; and
wherein the biopsy derived features comprise texture features and shape features.
15. The method of claim 9 , wherein the radiological image comprises an x-ray, a computerized tomography (CT) scan, a magnetic resonance imaging (MRI) scan, or a positron emission tomography (PET) scan.
16. A prognostic apparatus configured to predict a prognosis for a patient having small cell lung cancer (SCLC), comprising:
a memory configured to store an imaging data set comprising both scan data and digitized biopsy data from a patient having SCLC, wherein the scan data includes data from a radiological scan of the patient;
a feature extraction circuit configured to extract scan derived features from the scan data and biopsy derived features from the digitized biopsy data;
a risk score calculation circuit configured to calculate an RPRS from one or more of the scan derived features and from one or more of the biopsy derived features; and
a classification circuit configured to arrange the patient into one of a plurality of different risk groups based upon the RPRS.
17. The prognostic apparatus of claim 16 , further comprising:
a predictive feature identification circuit configured to identify predictive scan derived features from the scan derived features and predictive biopsy derived features from the biopsy derived features; and
wherein the risk score calculation circuit is configured to calculate the RPRS from both the predictive scan derived features and the predictive biopsy derived features.
18. The prognostic apparatus of claim 16 , further comprising:
a mask identification circuit configured to identify a first lesional mask and a first perilesional mask of the scan data; and
wherein the feature extraction circuit is configured to extract the scan derived features from the first lesional mask and the first perilesional mask.
19. The prognostic apparatus of claim 16 , wherein the digitized biopsy data comprises digitized Hematoxylin and Eosin (H&E) images.
20. The prognostic apparatus of claim 16 ,
wherein the risk score calculation circuit is further configured to determine a median risk score of a plurality of training radiomic-pathomic risk scores determined from imaging data of a plurality of preparatory patients; and
wherein the classification circuit is configured to arrange the patient into one of the plurality of different risk groups by comparing the RPRS to the median risk score.