# US 10,441,225 B2 · Predicting disease recurrence following trimodality therapy in non-small cell lung cancer using computed tomography derived radiomic features and clinico-pathologic features

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 10,441,225 B2
- Document ID: US10441225B2
- Status: Granted
- Kind: B2
- Application number: 16/236,675
- Filed: 2018-12-31
- Published: 2019-08-22
- Granted: 2019-10-15
- Earliest claimed: 2018-02-21
- Assignees: Case Western Reserve University
- Applicants: Case Western Reserve University
- Inventors: Anant Madabhushi; Mohammadhadi Khorrami; Vamsidhar Velcheti
- Web page: https://patentgrep.com/patent/US10441225B2

## CPC

- G06T 7/0014: {using an image reference approach}
- A61B 5/7267: {involving training the classification device}
- G06F 18/2413: based on distances to training or reference patterns
- G06F 18/2433: Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
- G06T 7/0016: {involving temporal comparison}
- G06T 7/11: Region-based segmentation
- G06T 7/155: involving morphological operators
- G06V 10/443: {by matching or filtering}
- G06V 10/764: using classification, e.g. of video objects
- G06T 2207/10081: Computed x-ray tomography [CT]
- G06T 2207/20081: Training; Learning
- G06T 2207/20084: Artificial neural networks [ANN]
- G06T 2207/30061: Lung
- G06T 2207/30064: Lung nodule
- G06T 2207/30096: Tumor; Lesion
- G06V 2201/03: Recognition of patterns in medical or anatomical images
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- A61B: DIAGNOSIS; SURGERY; IDENTIFICATION (usefulness limited to only animals A61D)
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

Embodiments include operations, apparatus, methods and other embodiments that access a baseline CT image of a region of tissue (ROT) demonstrating non-small cell lung cancer (NSCLC), segment a tumoral region represented in the baseline CT image; define a peritumoral region by dilating the tumoral boundary; extract a set of tumoral radiomic features from the tumoral region, a set of peritumoral radiomic features from the peritumoral region, and a set of clinico-pathologic features from the baseline CT image; provide the set of tumoral radiomic features, peritumoral radiomic features, and clinico-pathologic features to a machine learning classifier; receive, from the machine learning classifier, a time-to-recurrence post trimodality therapy (TMT) prediction, based on the set of tumoral radiomic features, peritumoral radiomic features, and clinico-pathologic features; generate a classification of the ROT as an MPR responder or MPR non-responder based, at least in part, on the time-to-recurrence post-TMT prediction; and display the classification.

## Claims

### Claim 1

A non-transitory computer-readable storage device storing instructions that when executed control a processor to perform operations for predicting major pathologic response (MPR) in non-small cell lung cancer (NSCLC), the operations comprising:
accessing a baseline computed tomography (CT) image of a region of tissue (ROT) demonstrating NSCLC, the baseline CT image having a plurality of pixels, a pixel having an intensity;
segmenting a tumoral region represented in the baseline CT image, the segmented tumoral region having a tumoral boundary;
defining a peritumoral region by performing a morphological dilation of the tumoral boundary;
extracting a set of tumoral radiomic features from the tumoral region;
extracting a set of peritumoral radiomic features from the peritumoral region;
extracting a set of clinico-pathologic features from the baseline CT image;
providing the set of tumoral radiomic features, the set of peritumoral radiomic features, and the set of clinico-pathologic features to a machine learning classifier;
receiving, from the machine learning classifier, a time-to-recurrence post-trimodality therapy (TMT) prediction, where the machine learning classifier computes the time-to-recurrence post-TMT prediction based on the set of tumoral radiomic features, the set of peritumoral radiomic features, and the set of clinico-pathologic features;
generating a classification of the ROT as an MPR responder or MPR non-responder by classifying the ROT as an MPR responder or MPR non-responder based, at least in part, on the time-to-recurrence post TMT prediction; and
displaying the classification.

### Claim 2 (depends on 1)

The non-transitory computer-readable storage device of claim 1, where the set of tumoral radiomic features includes a tumoral Laplacian Law feature.

### Claim 3 (depends on 2)

The non-transitory computer-readable storage device of claim 2, where the set of peritumoral radiomic features includes a peritumoral Laplacian Law feature.

### Claim 4 (depends on 3)

The non-transitory computer-readable storage device of claim 3, where the set of clinico-pathologic features includes a lymphovascular invasion (LVI) status feature.

### Claim 5 (depends on 1)

The non-transitory computer-readable storage device of claim 1, where the baseline CT image is a non-contrast, pre-neoadjuvant chemotherapy CT image of a region of tissue demonstrating stage III NSCLC.

### Claim 6 (depends on 1)

The non-transitory computer-readable storage device of claim 1, where segmenting the tumoral region includes automatically segmenting the tumoral region using a region growing technique, thresholding, or a watershed approach.

### Claim 7 (depends on 1)

The non-transitory computer-readable storage device of claim 1, where the peritumoral region is defined using a 15 mm morphological dilation of the tumoral boundary.

### Claim 8 (depends on 1)

The non-transitory computer-readable storage device of claim 1, where the machine learning classifier is a multivariable logistic regression (MLR) classifier trained using three-fold cross-validation on a training set of baseline CT images, where a member of the training set includes at least one baseline CT image of a region of tissue that experienced MPR post TMT, and at least one different baseline CT image of a region of tissue that did not experience MPR post TMT.

### Claim 9 (depends on 1)

The non-transitory computer-readable storage device of claim of claim 1, where MPR is defined as less than or equal to ten percent viable tumor.

### Claim 10 (depends on 1)

The non-transitory computer-readable storage device of claim 1, where classifying the ROT as an MPR responder or MPR non-responder includes classifying the ROT as an MPR responder or MPR non-responder with an area under the curve of at least 0.92.

### Claim 11 (depends on 1)

The non-transitory computer-readable storage device of claim 1, the operations further comprising:
upon determining that the classification is MPR responder:
providing a recommendation of the ROT for TMT; and
displaying the recommendation.

### Claim 12 (depends on 1)

The non-transitory computer-readable storage device of claim 1, where the set of tumoral radiomic features includes an area of lesion feature and a Gabor feature;
where the set of peritumoral radiomic features includes a Laws feature and a Law_Laplacian feature; and
where the set of clinico-pathologic features includes a lymphovascular invasion (LVI) feature and a percent viable tumor feature.

### Claim 13 (depends on 1)

The non-transitory computer-readable storage device of claim 1, where the set of tumoral radiomic features includes a Law_Laplacian W5×E5 feature;
where the set of peritumoral radiomic features includes a Law_Laplacian S5×R5 feature, a Laws W5×S5 feature, and a Law_Laplacian W5×S5 feature;
where the set of clinico-pathologic features includes a percent viable tumor feature; and
where the machine learning classifier generates an overall survival (OS) prediction based, at least in part, on the set of tumoral radiomic features, the set of peritumoral radiomic features, and the set of clinico-pathologic features.

### Claim 14 (depends on 13)

The non-transitory computer-readable storage device of claim 13, the operations further comprising displaying the OS prediction.

### Claim 15 (depends on 1)

The non-transitory computer-readable storage device of claim 1, where the set of tumoral radiomic features includes a Law_Laplacian W5 by E5 feature;
where the set of peritumoral radiomic features includes a Law_Laplacian W5 by S5 feature;
where the set of clinico-pathologic features includes a lymphovascular invasion (LVI) status feature; and
where the machine learning classifier generates a disease free survival (DFS) prediction based, at least in part, on the set of tumoral radiomic features, the set of peritumoral radiomic features, and the set of clinico-pathologic features.

### Claim 16 (depends on 15)

The non-transitory computer-readable storage device of claim 15, the operations further comprising displaying the DFS prediction.

### Claim 17

An apparatus for generating a recommendation that a patient receive trimodality therapy (TMT) for non-small cell lung cancer (NSCLC), the apparatus comprising:
a processor;
a memory configured to store a digitized baseline non-contrast computed tomography (CT) image of a region of tissue (ROT) demonstrating NSCLC;
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 digitized baseline non-contrast CT image of an ROT demonstrating NSCLC, where the baseline CT image includes a tumoral region and a peritumoral region, the baseline CT image having a plurality of pixels, a pixel having an intensity;
a tumoral radiomics circuit configured to extract a set of tumoral radiomic features from the tumoral region, where the set of tumoral radiomic features includes an area of lesion feature and a Gabor feature;
a peritumoral radiomics circuit configured to extract a set of peritumoral radiomic features from the peritumoral region, where the set of peritumoral radiomic features includes a Laws feature and a Law_Laplacian feature;
a clinico-pathologic feature circuit configured to extract a set of clinico-pathologic features from the ROT, where the set of clinico-pathologic features includes a lymphovascular invasion (LVI) feature and a percent viable tumor feature;
a TMT recommendation circuit configured to:
generate a probability that the ROT will experience post-TMT major pathologic response (MPR) based on the set of tumoral features, the set of peritumoral features, and the set of clinico-pathologic features; and
generate a recommendation for TMT based on the probability; and
a display circuit configured to display the recommendation.

### Claim 18 (depends on 17)

The apparatus of claim 17:
where the set of tumoral radiomic features includes a Law_Laplacian W5 by S5 feature;
where the set of peritumoral radiomic features includes a Laws W5 by S5 feature and a Law_Laplacian W5 by S5 feature;
where the set of clinico-pathologic features includes a percent viable tumor feature;
where the TMT recommendation circuit is further configured to:
to generate an overall survival (OS) prediction based on the set of tumoral features, the set of peritumoral features, and the set of clinico-pathologic features; and
generate a recommendation for TMT based on the OS prediction; and
where the display circuit is further configured to display the OS prediction.

### Claim 19 (depends on 17)

The apparatus of claim 17:
where the set of tumoral radiomic features includes a Law_Laplacian W5 by E5 feature;
where the set of peritumoral radiomic features includes a Law_Laplacian W5 by S5 feature;
where the set of clinico-pathologic features includes a lymphovascular invasion (LVI) status feature, a histology feature, and a percentage of viable tumor feature;
where the TMT recommendation circuit is further configured to:
to generate a disease free survival (DFS) prediction based on the set of tumoral features, the set of peritumoral features, and the set of clinico-pathologic features; and
generate a recommendation for TMT based on the DFS prediction; and
where the display circuit is further configured to display the DFS prediction.

### Claim 20

A non-transitory computer-readable storage device storing computer executable instructions that when executed by a computer control the computer to perform a method for predicting outcome of trimodality therapy (TMT), the method comprising:
accessing a pre-neoadjuvant chemotherapy, pre-surgery, baseline computed tomography (CT) image of a region of tissue demonstrating non-small cell lung cancer (NSCLC), the region of tissue including a tumor region and a peritumoral region, the baseline CT image having a plurality of pixels, a pixel having an intensity;
extracting a combined radiomic risk signature from the baseline CT image, where the combined radiomic risk signature includes at least one radiomic feature extracted from the tumoral region, at least one radiomic feature extracted from the peritumoral region, and at least one clinico-pathologic feature;
providing the combined radiomic risk signature to a machine learning classifier, where the machine learning classifier computes a major pathologic response (MPR) prediction based on the combined radiomic risk signature, computing a risk score based on the combined radiomic risk signature, that is predictive of a high risk or low risk of disease free survival (DFS) and generating a DFS prediction based on the risk score, or generating an overall survival (OS) prediction based on the combined radiomic risk signature;
receiving, from the machine learning classifier, the MPR prediction;
generating a TMT outcome prediction based on the MPR prediction, the DFS prediction, or the OS prediction; and
displaying the TMT outcome prediction and at least one of the MPR prediction, the DFS prediction, or the OS prediction.