WO2022118190A1 - Machine learning to predict cancer involvement of lymph nodes - Google Patents

Machine learning to predict cancer involvement of lymph nodes Download PDF

Info

Publication number
WO2022118190A1
WO2022118190A1 PCT/IB2021/061125 IB2021061125W WO2022118190A1 WO 2022118190 A1 WO2022118190 A1 WO 2022118190A1 IB 2021061125 W IB2021061125 W IB 2021061125W WO 2022118190 A1 WO2022118190 A1 WO 2022118190A1
Authority
WO
WIPO (PCT)
Prior art keywords
risk
lymph nodes
level
model
subject
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/IB2021/061125
Other languages
English (en)
French (fr)
Inventor
JR. George WASHKO
Raul San Jose Estepar
Charles Matthew KINSEY
Christopher Scott Stevenson
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Johnson and Johnson Enterprise Innovation Inc
Original Assignee
Johnson and Johnson Enterprise Innovation Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Johnson and Johnson Enterprise Innovation Inc filed Critical Johnson and Johnson Enterprise Innovation Inc
Priority to CN202180092579.7A priority Critical patent/CN116848548A/zh
Priority to AU2021390184A priority patent/AU2021390184A1/en
Priority to JP2023533338A priority patent/JP2023551335A/ja
Priority to MX2023006446A priority patent/MX2023006446A/es
Priority to EP21820328.9A priority patent/EP4256511A1/en
Priority to US18/255,432 priority patent/US12608805B2/en
Priority to CA3203664A priority patent/CA3203664A1/en
Priority to KR1020237022202A priority patent/KR20230117391A/ko
Publication of WO2022118190A1 publication Critical patent/WO2022118190A1/en
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G06T7/0014—Biomedical image inspection using an image reference approach
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/50—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for simulation or modelling of medical disorders
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/09—Supervised learning
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/7715—Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H30/00—ICT specially adapted for the handling or processing of medical images
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H30/00—ICT specially adapted for the handling or processing of medical images
    • G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10068—Endoscopic image
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10081—Computed x-ray tomography [CT]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10088—Magnetic resonance imaging [MRI]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10104—Positron emission tomography [PET]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10116—X-ray image
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10132—Ultrasound image
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20048—Transform domain processing
    • G06T2207/20064—Wavelet transform [DWT]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20076—Probabilistic image processing
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30061—Lung
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30061—Lung
    • G06T2207/30064—Lung nodule
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30096—Tumor; Lesion
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30101—Blood vessel; Artery; Vein; Vascular
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/03—Recognition of patterns in medical or anatomical images

Definitions

  • Computed tomographic (CT) imaging of the chest has become the clinical standard for the assessment of those at risk for lung cancer.
  • CT computed tomographic
  • Detection of such a lung nodule triggers a more extensive clinical evaluation to determine if it is cancerous and if so, has it spread to other structures in the chest or even remote areas of the body. This is done to determine the stage of the lung cancer which in turn dictates the plan for therapeutic intervention.
  • a aspect of lung cancer staging is the determination of its presence or absence in the thoracic lymph nodes (nodal status).
  • the current standard of care involves a staged approach where the nodule is first determined to be benign or malignant and then additional testing such as positron emission tomography (PET) scanning and surgical, endoscopic, or bronchoscopic based sampling of the lymph nodes is completed.
  • PET positron emission tomography
  • Tissue sampling procedures add risk, expense and time to the work up. New, non-invasive methods are needed to mitigate these risks and shorten the time between diagnosis and treatment.
  • machine learning and deep learning-based techniques applied to the initial thoracic CT scan used for lung cancer screening can address these needs.
  • Lung cancer detection, staging and treatment planning is a multistep process that can take weeks or months.
  • the extra time necessitated by this process leads to significant patient anxiety and heightened chances of interval metastases.
  • longer diagnosis to treatment times are associated with a worse overall survival.
  • An exemplary approach to mitigating these concerns is one that allows near simultaneous cancer diagnosis, staging and in the case of bronchoscopic approaches to therapy, treatment.
  • Embodiments of the invention disclosed herein are based upon analysis of thoracic
  • CT scans used to screen for lung cancer in high risk populations identifies patients at high risk for cancer that has metastasized to at least one lymph node (LN) in the chest and therefore require additional testing such as PET scanning and lymph node biopsy. This allows the clinical providers to prioritize testing to those at greatest need. Conversely, patients with suspected lung cancer who are at low risk for having lymph node metastasis could proceed directly to immediate therapy such as surgical resection or combined bronchoscopic diagnosis and treatment. This will eliminate the time between disease diagnosis and therapy.
  • LN lymph node
  • embodiments of the invention disclosed herein enable identification of patients who are high or low risk for having nodal disease thereby allowing clinicians to either prioritize additional testing in high risk individuals or have lung cancer patients with low metastatic risk proceed directly to confirmatory bronchoscopic biopsy immediately followed by bronchoscopically delivered therapy.
  • a method of determining a subject level risk of metastatic cancer of a subject comprising: obtaining one or more images captured from the subject comprising a plurality of lymph nodes of the subject; and predicting the subject level risk of metastatic cancer by applying a risk model to extracted features of the obtained one or more images, the risk model trained to predict lymph node (LN) level risks of metastatic cancer for lymph nodes in images, wherein the risk model is trained using at least labels derived from imputed LN-level risks of metastatic cancer for a set of lymph nodes of a reference individual, the imputed LN-level risks of the set of lymph nodes determined using at least an imputation model that discriminates between cancerous and non-cancerous lymph nodes of training images.
  • LN lymph node
  • predicting the subject level risk of metastatic cancer further comprises: selecting one or more of the plurality of lymph nodes based on their LN-level risk of metastatic cancer predicted by the risk model; and determining the subject level risk of metastatic cancer using the LN-level risk of metastatic cancer predicted for the one or more lymph nodes.
  • selecting one or more of the plurality of lymph nodes comprises identifying the lymph node with the highest probability of LN-level risk, and wherein determining the subject level risk of metastatic cancer comprises assigning the highest probability of LN-level risk as the subject level risk of metastatic cancer.
  • the risk model predicts LN-level risks of metastatic cancer with a greater range of probabilities than the imputed LN-level risks predicted by the imputation model.
  • the set of lymph nodes of a reference individual is selected by: determining a median risk value; and comparing imputed LN-level risks of the lymph nodes of the reference individual to the median risk value.
  • the set of lymph nodes of a reference individual is further selected by: including one or more lymph nodes with LN-level risks greater than the median risk value in the set of lymph nodes. In various embodiments, one or more lymph nodes with LN-level risks less than the median risk value in the set of lymph nodes are excluded from the set of lymph nodes.
  • the median risk value is a median intrasubject LN-level risk of the reference individual.
  • the one or more images comprises a computed tomography (CT) image.
  • CT computed tomography
  • the one or more images comprises are obtained from a thoracic CT scan.
  • methods disclosed herein further comprise determining a stage of the cancer according to the predicted subject level risk of metastatic cancer.
  • methods disclosed herein further comprise selecting a diagnostic or treatment for the subject according to the predicted subject level risk of metastatic cancer.
  • selecting a diagnostic or treatment for the subject comprises selecting either surgical tumor resection or combined bronchoscopic or endoscopic diagnosis and treatment when the predicted subject level risk of metastatic cancer is below a threshold risk value.
  • selecting a diagnostic or treatment for the subject comprises selecting additional diagnostic testing of one or both of performing a PET or PET-CT scan or performing a lymph node biopsy when the predicted subject level risk of metastatic cancer is above a threshold risk value.
  • the imputation model is trained using training images obtained from thoracic CT scans.
  • the training images obtained from thoracic CT scans comprise one or more lymph nodes.
  • the imputation model is trained using training images of the National Lung Screening Trial (NLST).
  • the imputation model is trained using training images of a custom dataset.
  • the imputation model is trained by generating two or more radiomic panels from at least one of the training images.
  • the two or more radiomic panels comprise a Laplacian of Gaussians transformation of the training image and a wavelet transform of the training image.
  • the imputation model is further trained by extracting features from each of the two or more radiomic panels; and training the imputation model using at least the extracted features.
  • the imputation model is further trained by: defining a region of interest (ROI) in the radiomic panels, the defined region of interest comprising a lymph node; extracting features from the ROI in the radiomic panels; and training the imputation model using at least the extracted features.
  • ROI region of interest
  • the extracted features include one or more feature categories of first order features, shape features, gray level co-occurrence matrix 1 (GLCM) features, gray level run length matrix (GLRLM) features, gray level size zone matrix (GLSZM) features, and neighborhood gray tone difference matrix (NGTDM) features.
  • the imputation model is further trained using reference ground truth values indicating a presence or absence of metastasis in training images.
  • the reference ground truth values are derived from Nstage covariable values indicating the stage of the cancer, wherein a cancer stage of “0” or “1” indicate an absence of metastasis and wherein a cancer stage of “2” or “3” indicate a presence of metastasis.
  • the reference ground truth values indicate whether individual lymph nodes are cancerous or non-cancerous. In various embodiments, the reference ground truth values indicating whether individual lymph nodes are cancerous or non-cancerous are back-calculated from the Nstage covariable values based on characteristics of the tumor or lymph nodes.
  • the imputation model is a random forest classifier or a LASSO classifier.
  • the risk model is a random forest classifier or a LASSO classifier.
  • the risk model is trained using supervised learning techniques.
  • the plurality of lymph nodes comprises 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 lymph nodes.
  • the plurality of lymph nodes comprises between 8 to 20 lymph nodes, between 10 to 19 lymph nodes, between 12 to 18 lymph nodes, or between 14 and 16 lymph nodes.
  • the metastatic cancer is metastatic lung cancer.
  • a method of training the risk model of any one of claims 1-35 comprising: applying an imputation model to generate imputed LN- level risks for a plurality of lymph nodes in training images; selecting a set of lymph nodes, wherein the lymph nodes in the set have LN-level risks that are greater than a median risk value; using the LN-level risks of the lymph nodes in the set of lymph nodes as reference ground truths for training the risk model, the risk model able to predict LN-level risks with a greater range of probabilities than the imputed LN-level risks predicted by the imputation model.
  • the risk model is further trained by extracting features from one or more radiomic panels; and training the risk model using at least the extracted features.
  • extracting features from two or more radiomic panels comprises: defining a region of interest (ROI) in the one or more radiomic panels, the defined region of interest comprising a lymph node; extracting features from the ROI in the one or more radiomic panels.
  • the extracted features include one or more feature categories of first order features, shape features, gray level co-occurrence matrix 1 (GLCM) features, gray level run length matrix (GLRLM) features, gray level size zone matrix (GLSZM) features, and neighborhood gray tone difference matrix (NGTDM) features.
  • a system for determining a subject level risk of metastatic cancer of a subject comprises: an imaging device configured to capture one or more images of the subject; and a computing device configured to perform the steps of: obtaining one or more images captured from the subject comprising a plurality of lymph nodes of the subject; and predicting the subject level risk of metastatic cancer by applying a risk model to extracted features of the obtained one or more images, the risk model trained to predict lymph node (LN) level risks of metastatic cancer for lymph nodes in images, wherein the risk model is trained using at least labels derived from imputed LN-level risks of metastatic cancer for a set of lymph nodes of a reference individual, the imputed LN-level risks of the set of lymph nodes determined using at least an imputation model that discriminates between cancerous and non-cancerous lymph nodes of training images.
  • LN lymph node
  • a non-transitory computer readable medium comprises instructions that, when executed by a processor, cause the processor to: obtain one or more images captured from the subject comprising a plurality of lymph nodes of the subject; and predict the subject level risk of metastatic cancer by applying a risk model to extracted features of the obtained one or more images, the risk model trained to predict lymph node (LN) level risks of metastatic cancer for lymph nodes in images, wherein the risk model is trained using at least labels derived from imputed LN-level risks of metastatic cancer for a set of lymph nodes of a reference individual, the imputed LN-level risks of the set of lymph nodes determined using at least an imputation model that discriminates between cancerous and non-cancerous lymph nodes of training images.
  • LN lymph node
  • FIG. 1A depicts a system environment overview for cancer staging by determining a subject level risk of metastatic cancer of a subject, in accordance with an embodiment.
  • FIG. IB depicts a block diagram of the cancer staging system, in accordance with an embodiment.
  • FIG. 2A depicts an example lymph node map identifying lymph node stations, in accordance with an embodiment.
  • FIG. 2B is an example lymph node annotation, in accordance with an embodiment.
  • FIG. 3A depicts an example implementation of the imputation model and the risk model, in accordance with an embodiment.
  • FIG. 3B depicts the implementation of the risk model in further detail, in accordance with an embodiment.
  • FIG. 4 illustrates an example computer for implementing the entities shown in FIG. 1A, IB, 2A, 2B, 3A, and 3B.
  • FIGs. 5A-5R depict example correlation heat maps of different lymph node radiomic features.
  • FIG. 6 depicts lymph node level performance of an imputation model.
  • FIGs. 7A and 7B depict subject level performance using the highest probability score predicted by the LN-level classifier.
  • FIG. 8 depicts lymph node level performance an imputation model in the two step bootstrap approach.
  • FIGs. 9A and 9B depict subject level performance using the bootstrap approach.
  • a letter after a reference numeral indicates that the text refers specifically to the element having that particular reference numeral.
  • a reference numeral in the text without a following letter, such as “individual 315,” refers to any or all of the elements in the figures bearing that reference numeral (e.g. “individual 315” in the text refers to reference numerals “individual 3I5A,” “individual 315B,” “individual 3I5C,” and “individual 315D” in the figures).
  • subject encompasses a cell, tissue, or organism, human or non-human, whether in vivo, ex vivo, or in vitro, male or female.
  • mammal encompasses both humans and non-humans and includes but is not limited to humans, non-human primates, canines, felines, murines, bovines, equines, and porcines.
  • sample of “test sample” can include a single cell or multiple cells or fragments of cells or an aliquot of body fluid, such as a blood sample, taken from a subject, by means including venipuncture, excretion, ejaculation, massage, biopsy, needle aspirate, lavage sample, scraping, surgical incision, or intervention or other means known in the art.
  • Examples of an aliquot of body fluid include amniotic fluid, aqueous humor, bile, lymph, breast milk, interstitial fluid, blood, blood plasma, cerumen (earwax), Cowper’s fluid (pre- ejaculatory fluid), chyle, chyme, female ejaculate, menses, mucus, saliva, urine, vomit, tears, vaginal lubrication, sweat, serum, semen, sebum, pus, pleural fluid, cerebrospinal fluid, synovial fluid, intracellular fluid, and vitreous humour.
  • reference individual refers to an individual with known cancer staging.
  • reference individuals can include healthy individuals (e.g., control individuals).
  • reference individuals can include individuals previously diagnosed with cancer and staged.
  • reference individuals include individuals previously diagnosed with cancer that has metastasized to the lymph nodes, e.g., stage IV cancer.
  • the term “obtaining one or more images” encompasses obtaining one or more images captured from a subject or obtaining one or more images captured from a sample obtained from a subject. Obtaining one or more images can encompass performing steps of capturing the one or more images from the subject or from a sample obtained from the subject. The phrase can also encompass receiving one or more images, e.g., from a third party that has performed the steps of capturing the one or more images from the subject or from a sample obtained from the subject. The one or more images can be obtained by one of skill in the art via a variety of known ways including stored on a storage memory. [0036] The terms “lymph node involvement,” “lymph node metastasis,” or “lymph node cancer” refer to the presence or absence of metastasized cancer in one or more lymph nodes of a subject.
  • lymph node metastatic risk As used interchangeably and generally refer to a risk of metastatic cancer in a specific lymph node.
  • subject level risk of metastatic cancer or “patient level risk of metastatic cancer” are used interchangeably and refer to a likelihood that a cancer in a subject has metastasized.
  • the cancer in the subject is a lung cancer that has metastasized in one or more lymph nodes of the subject.
  • cancer stage or “stage of cancer” refers to a stage of a cancer in a subject.
  • Cancer stage can refer to the TNM staging system which includes stages 0-4. Additionally, cancer stage can refer to nodal cancer stages 0-3. Nodal cancer stage 0 indicates no nodal involvement, nodal cancer stage 1 indicates hilar nodal involvement, and nodal stages 2 or 3 indicate mediastinal involvement.
  • FIG. 1A depicts a system environment overview 100 for cancer staging by determining a subject level risk of metastatic cancer of a subject, in accordance with an embodiment.
  • the system environment 100 provides context in order to introduce a subject 110, an imaging device 120, and a cancer staging system 130 for determining a cancer prediction 140.
  • the cancer prediction 140 determined by the cancer staging system 130 serves as the cancer staging for the subject 110.
  • the cancer prediction 140 determined by the cancer staging system 130 is a predicted level of risk for the subject 110 and serves as a basis for determining a cancer staging for the subject 110.
  • the subject is healthy.
  • the methods for cancer staging described herein can be beneficial for early detection of cancer in the healthy subject.
  • the subject was previously diagnosed with a cancer.
  • the subject is healthy, but is suspected to have a form of cancer (e.g., a subject who has exhibited symptoms associated with a cancer).
  • the type of cancer in the subject is a lung cancer.
  • the methods for cancer staging described herein can be beneficial for identifying and differentiating subjects who are at high risk of metastatic cancer including lymph node involvement and subjects with suspected lung cancer who are at low risk for having lymph node metastasis.
  • Subjects at high risk of metastatic cancer can undergo additional testing such as PET scanning and lymph node biopsy whereas subjects who are at low risk for having lymph node metastasis can undergo immediate therapy such as surgical resection of the lung cancer or combined bronchoscopic diagnosis and treatment.
  • the imaging device 120 captures an image from the subject 110 for further analysis. In various embodiments, the imaging device 120 captures an image from a test sample obtained from the subject 110.
  • the image and/or the sample can be obtained by the individual or by a third party, e.g., a health care provider or a medical professional. Examples of medical professionals include physicians, emergency medical technicians, nurses, first responders, psychologists, phlebotomist, medical physics personnel, nurse practitioners, surgeons, dentists, and any other obvious medical professional as would be known to one skilled in the art.
  • the imaging device 120 captures an image of an anatomical location of the subject 110.
  • Example anatomical locations of a subject can include lungs, thoracic cavity, kidney, liver, pancreas, brain, stomach, intestines, hip, knees, legs, arms, and face.
  • the imaging device 120 captures an image of the thoracic cavity of the subject 110.
  • the imaging device 120 captures an image of one or more lymph nodes of the subject 110.
  • the imaging device 120 captures an image of the thoracic cavity including one or more lymph nodes of the subject 110.
  • the imaging device 120 is one of a computed tomography (CT) scanner, magnetic resonance imaging (MRI) scanner, positron emission tomography (PET) scanner, x-ray scanner, or an ultrasound imaging device.
  • CT computed tomography
  • MRI magnetic resonance imaging
  • PET positron emission tomography
  • the imaging device 120 is a CT scanner that captures one or more images of the subject 110.
  • the imaging device 120 is a CT scanner that captures one or more CT images of the thoracic cavity including one or more lymph nodes of the subject 110.
  • the cancer staging system 130 analyzes one or more images (e.g., images captured by the imaging device 120) and generates the cancer prediction 140.
  • the cancer prediction 140 can be a predicted stage of cancer in the subject 110.
  • the cancer prediction 140 can be a predicted likelihood of cancer lymph node involvement in the subject 110.
  • the cancer prediction 140 can include a recommended clinical approach (e.g., additional testing such as PET scanning and lymph node biopsy) or therapy (e.g., surgical resection and/or drug therapy).
  • the cancer staging system 130 can include one or more computers, embodied as a computer system 400 as discussed below with respect to FIG. 4.
  • the steps described in reference to the cancer staging system 130 are performed in silico.
  • the imaging device 120 and the cancer staging system 130 are employed by different parties. For example, a first party operates the imaging device 120 120 to capture one or more images from the subject 110 and then provides the captured one or more images to a second party which implements the cancer staging system 130 to determine a cancer prediction 140.
  • FIG. IB depicts a block diagram illustrating the computer logic components of the cancer staging system 130, in accordance with an embodiment.
  • the cancer staging system 130 may include a lymph node feature module 145, an imputation module 150, a risk training module 155, a risk deployment module 160, a cancer prediction module 165, and a training data store 170.
  • the lymph node feature module 145 extracts features from images derived from subjects or training subjects.
  • the imputation module 150 trains and deploys an imputation model.
  • the imputation model analyzes extracted features and determines lymph node level risk of cancer for each lymph node in images.
  • the risk training module 155 trains a risk model using training data including at least the lymph node level risk of cancer determined by the imputation model.
  • the risk deployment module 160 implements the risk model to analyze features extracted from images obtained from a subject (e.g., subject 110 in FIG. 1A) to determine a cancer prediction, such as a prediction of subject level metastatic risk or a cancer stage, for the subject 110.
  • the cancer staging system 130 trains and implements at least two models (e.g., an imputation model and a risk model).
  • the cancer staging system 130 generates a cancer prediction, such as subject level metastatic risk or a prediction of a cancer stage, for a subject.
  • the implementation of both the imputation model and the risk model results in an improved cancer prediction.
  • the imputation model learns from the subject-level N-stage of the training data and is applied to features extracted of the lymph nodes to define which lymph nodes are likely to contribute to a positive N-stage.
  • the risk model learns a more refined lymph node metastatic risk probability from the imputed data.
  • the risk model predicts LN-level risks of metastatic cancer with a greater range of probabilities than the imputed LN-level risks predicted by the imputation model. This enables a more accurate determination of subject level risk of metastatic cancer.
  • the components of the cancer staging system 130 are hereafter described in reference to two phases: 1) a training phase and 2) a deployment phase.
  • the training phase refers to the building and training of one or more models based on training data, such as training images captured from reference individuals (e.g., individuals with a known cancer staging). Therefore, the models are trained using the training data such that during the deployment phase, implementation of the models enables the prediction of a cancer staging for a subject (e.g., subject 110 in FIG. 1A).
  • the imputation module 150 can also be implemented to train and deploy a model, hereafter referred to as an imputation model.
  • the imputation module 150 trains the imputation model during the training phase and deploys the imputation model during the deployment phase.
  • the components of the cancer staging system 130 are applied during one of the training phase and the deployment phase.
  • the risk training module 155 and training data store 170 are applied during the training phase to train a risk model.
  • the risk deployment module 160 is applied during the deployment phase.
  • the components of the cancer staging system 130 can be performed by different parties depending on whether the components are applied during the training phase or the deployment phase. In such scenarios, the training and deployment of the prediction model are performed by different parties.
  • the risk training module 155 and training data store 170 applied during the training phase can be employed by a first party (e.g., to train a risk model) and the risk deployment module 160 applied during the deployment phase can be performed by a second party (e.g., to deploy the risk model).
  • the imputation module 150 can be applied to train an imputation model by a first party and the imputation module 150 can be applied to deploy an imputation model by a second party.
  • Embodiments described herein include methods for cancer staging by analyzing one or more images captured from a subject and applying one or more models to determine a subject level of metastatic cancer risk involving the lymph nodes. Such methods can be performed by the cancer staging system 130 described in FIG. IB.
  • the lymph node feature module 145 analyzes one or more images and extracts features from the one or more images.
  • the lymph node feature module 145 extracts features from training images obtained from reference individuals. These features extracted from training images can be used to train one or more models, such as an imputation model, as is described in further detail below.
  • the lymph node feature module 145 extracts features images obtained from a subject (e.g., subject 110 in FIG. 1A). These features extracted from images obtained from the subject can be analyzed by deploying a model, such as a risk model, as is described in further detail below.
  • the one or more images or training images includes one or more lymph nodes in the image.
  • the image or training image can be an image of a thoracic cavity including one or more lymph nodes.
  • an image includes at least 2 lymph nodes.
  • an image includes 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 lymph nodes.
  • an image includes between 1-10 lymph nodes, 1-15 lymph nodes, 2-10 lymph nodes, 2-15 lymph nodes, 2-20 lymph nodes, 4-10 lymph nodes, 4-15 lymph nodes, 4-20 lymph nodes, 5- 10 lymph nodes, 5-15 lymph nodes, 5-20 lymph nodes, 8-15 lymph nodes, 8-20 lymph nodes, 10-15 lymph nodes, 10-20 lymph nodes, 12-15 lymph nodes, 12-20 lymph nodes, or 15-20 lymph nodes.
  • multiple images are captured from a single subject or a reference individual. The multiple images can capture different lymph nodes of a subject or reference individual. In various embodiments, the multiple images include at least 2 lymph nodes.
  • the multiple images include 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 lymph nodes. In various embodiments, the multiple images include between 8 to 20 lymph nodes. In various embodiments, the multiple images include between 10 to 19 lymph nodes, between 12 to 18 lymph nodes, or between 14 and 16 lymph nodes.
  • FIG. 2A depicts an example lymph node map identifying lymph node stations, in accordance with an embodiment. More specifically, FIG. 2A is a depiction of regional lymph node classifications adapted from the American Thoracic Society mapping scheme. Here, lymph node stations are annotated to depict the location and categorization of each lymph node.
  • the lymph node feature module 145 annotates one or more lymph nodes in the image.
  • the lymph node feature module 145 annotates one or more lymph nodes by defining a region around the lymph node.
  • the defined region around the lymph node is a spherical region.
  • the spherical region has a radius of X mm.
  • X is 5 mm.
  • X is 7.5 mm.
  • X is 0.5 mm, 1 mm, 1.5 mm, 2 mm, 2.5 mm, 3 mm, 3.5 mm, 4 mm, 4.5 mm, 5.5 mm, 6 mm, 6.5 mm, 7 mm, 8 mm, 8.5 mm, 9 mm, 9.5 mm, 10 mm, 10.5 mm, 11 mm, 11.5 mm, 12 mm, 12.5 mm, 13 mm, 13.5 mm, 14 mm, 14.5 mm, or 15 mm.
  • X is between 1 and 15 mm.
  • X is between 2 and 12 mm.
  • X is between 3 and 10 mm.
  • X is between 4 and 8 mm. In various embodiments, X is between 5 and 7.5 mm. In various embodiments, X is between 4 and 6 mm. In various embodiments, X is between 6 and 8 mm. In various embodiments, X is between 7 and 8 mm. In various embodiments, is between 10 and 15 mm.
  • the lymph node feature module 145 annotates one or more lymph nodes by defining two or more spherical regions around lymph nodes.
  • the lymph node feature module 145 annotates one or more lymph nodes by defining a first spherical region of 5 mm around the lymph node and a second spherical region of 7.5 mm around the lymph node.
  • FIG. 2B is an example lymph node annotation, in accordance with an embodiment. Specifically, FIG. 2B depicts two defined spherical regions around two 2R upper paratracheal nodes.
  • the lymph node feature module 145 extracts features from each defined region around a lymph node. In various embodiments, the lymph node feature module 145 performs a transformation of the original input to generate a transformed image from which additional features are then extracted. In one embodiment, the lymph node feature module 145 performs a wavelet transform of the original image. In one embodiment, the lymph node feature module 145 performs a Laplacian of Gaussians (LoG) of the original image. Therefore, the lymph node feature module 145 can extract features from defined regions around lymph nodes in the transformed image. In various embodiments, the lymph node feature module 145 extracts features from defined regions around lymph nodes in the original image and the transformed image.
  • LiG Laplacian of Gaussians
  • the lymph node feature module 145 performs two transforms of the original image (e.g., both a wavelet transform of the original image and a Laplacian of Gaussians (LoG) of the original image).
  • the lymph node feature module 145 can extract features from defined regions around lymph nodes in both the first transformed image and the second transformed image.
  • the lymph node feature module 145 extracts features from defined regions around lymph nodes in the original image, the first transformed image, and the second transformed image.
  • Example feature categories from the original image, the first transformed image, and/or the second transformed image can include first order features, shape features, gray level co-occurrence matrix 1 (GLCM) features, gray level run length matrix (GLRLM) features, gray level size zone matrix (GLSZM) features, and neighborhood gray tone difference matrix (NGTDM) features.
  • First order features refer to features of the first-order statistics of the intensity values.
  • GLCM features derive from a histogram of co-occurring greyscale values at a given offset over an image.
  • GLRLM features involve statistics of the gray level run length matrix.
  • GLSZM features involve statistics of the gray level size zone matrix.
  • NGTDM features involve statistics of the neighboring gray tone difference matrix.
  • features in the different feature categories include correlation features, LMC2, MCC, Difference in Entropy, Joint Entropy, Sum of Entropy, Cluster prominence, average difference joint average, sum of average, autocorrelation, difference in variance, contrast, cluster tendency, sum of squares, LDMN, LDN, Cluster shade, LMC1, Inverse variance, LD, LDM, joint energy, and maximum probability.
  • the lymph node feature module 145 extracts at least 50 features from an image. In various embodiments, the lymph node feature module 145 extracts at least 100 features, at least 150 features, at least 200 features, at least 250 features, at least 300 features, at least 350 features, at least 400 features, at least 450 features, at least 500 features, at least 550 features, at least 600 features, at least 650 features, at least 700 features, at least 750 features, at least 800 features, at least 850 features, at least 900 features, at least 950 features, at least 1000 features, at least 1100 features, at least 1200 features, at least 1300 features, at least 1400 features, at least 1500 features, at least 1600 features, at least 1700 features, at least 1800 features, at least 1900 features, at least 2000 features, at least 3000 features, at least 4000 features, at least 5000 features, at least 6000 features, at least 7000 features, at least 8000 features, at least 9000 features, or at least 10,000 features from an image.
  • the lymph node feature module 145 extracts between 10 features and 100,00 features. In various embodiments, the lymph node feature module 145 extracts between 100 features and 1000 features. In various embodiments, the lymph node feature module 145 extracts between 300 features and 900 features. In various embodiments, the lymph node feature module 145 extracts between 500 features and 1000 features. In various embodiments, the lymph node feature module 145 extracts between 500 features and 900 features. In various embodiments, the lymph node feature module 145 extracts between 700 features and 1000 features. In various embodiments, the lymph node feature module 145 extracts between 700 features and 900 features. In various embodiments, the lymph node feature module 145 extracts between 800 features and 1000 features. In various embodiments, the lymph node feature module 145 extracts between 800 features and 900 features.
  • the lymph node feature module 145 provides the extracted features to the imputation module 150 for training or deploying the imputation model. For example, if the extracted features are derived from training images captured from reference individuals, the lymph node feature module 145 provides the extracted features to the imputation module 150. In various embodiments, the lymph node feature module 145 provides the extracted features to the risk training module 155 for training the risk model. For example, if the extracted features are derived from training images captured from reference individuals, the lymph node feature module 145 provides the extracted features to the risk training module 155 for training the risk model. In various embodiments, the lymph node feature module 145 provides the extracted features to the risk deployment module 160 for deploying the risk model. For example, if the extracted features are derived from images captured from a subject (e.g., subject 110 in FIG. 1A), the lymph node feature module 145 provides the extracted features to the risk deployment module 160.
  • the lymph node feature module 145 provides the extracted features to the risk deployment module 160.
  • the imputation module 150 trains and deploys an imputation model.
  • the imputation model is structured such that it analyzes features extracted from an image including one or more lymph nodes and predicts probabilities for the one or more lymph nodes that the lymph nodes are cancerous.
  • the imputation model based on the extracted features from an image including one or more lymph nodes, the imputation model imputes a probability for each of the lymph nodes, an imputed probability indicating a likelihood that the lymph node is cancerous (e.g., metastatic cancer).
  • the imputation model discriminates between cancerous and non-cancerous lymph nodes by determining risks of metastatic cancer at the lymph-node level.
  • the imputation model is any one of a regression model (e.g., linear regression, logistic regression, or polynomial regression), decision tree, random forest, support vector machine, Naive Bayes model, k-means cluster, or neural network (e.g., feedforward networks, convolutional neural networks (CNN), deep neural networks (DNN), autoencoder neural networks, generative adversarial networks, or recurrent networks (e.g., long short-term memory networks (LSTM), bi-directional recurrent networks, deep bidirectional recurrent networks), or any combination thereof.
  • the imputation model is a random forest classifier or a least absolute shrinkage and selection operator (LASSO) classifier.
  • the imputation model is a machine learned model.
  • the imputation model is a boosted gradient machine learning model (e.g., xgboost).
  • the imputation model can be trained using a machine learning implemented method, such as any one of a linear regression algorithm, logistic regression algorithm, decision tree algorithm, support vector machine classification, Naive Bayes classification, K-Nearest Neighbor classification, random forest algorithm, deep learning algorithm, gradient boosting algorithm, and dimensionality reduction techniques such as manifold learning, principal component analysis, factor analysis, autoencoder regularization, and independent component analysis, or combinations thereof.
  • the imputation model is trained using supervised learning algorithms, unsupervised learning algorithms, semi-supervised learning algorithms (e.g., partial supervision), weak supervision, transfer, multi-task learning, or any combination thereof.
  • the imputation model has one or more parameters, such as hyperparameters or model parameters.
  • Hyperparameters are generally established prior to training. Examples of hyperparameters include the learning rate, depth or leaves of a decision tree, number of hidden layers in a deep neural network, number of clusters in a k- means cluster, penalty in a regression model, and a regularization parameter associated with a cost function.
  • Model parameters are generally adjusted during training. Examples of model parameters include weights associated with nodes in layers of neural network, support vectors in a support vector machine, and coefficients in a regression model.
  • the model parameters of the imputation model are trained (e.g. , adjusted) using the training data to improve the predictive capacity of the imputation model.
  • the imputation module 150 trains the imputation model using training data.
  • the training data can be stored and/or retrieved from training data store 170.
  • the training data includes extracted features from training images including one or more lymph nodes obtained from reference individuals.
  • reference individuals includes healthy individuals (e.g., control individuals).
  • reference individuals include individuals previously diagnosed with cancer.
  • reference individuals include individuals previously diagnosed with cancer that has metastasized to the lymph nodes. Therefore, images captured from such individuals include one or more cancerous lymph nodes.
  • the training data used for training the imputation model includes reference ground truths that indicate whether a lymph node is cancerous (hereafter also referred to as “positive” or “+”) or whether a lymph node is non-cancerous (hereafter also referred to as “negative” or “-“).
  • the reference ground truths in the training data are binary values, such as “1” or “0.”
  • a positive lymph node that is cancerous can be identified in the training data with a value of “1” whereas a negative lymph node that is non-cancerous can be identified in the training data with a value of “0.”
  • the imputation module 150 trains the imputation model using the training data to minimize a loss function such that the imputation model can better predict an outcome (e.g., a probability of lymph node metastatic cancer for a lymph node) based on the input (e.g., extracted features of the lymph node).
  • the loss function is constructed for any of a least absolute shrinkage and selection operator (LASSO) regression, Ridge regression, or ElasticNet regression.
  • LASSO least absolute shrinkage and selection operator
  • the training data can be obtained and/or derived from a publicly available database.
  • the training data can be obtained and/or derived from the National Lung Screening Trial (NLST) which includes nodal stage (N stage) cancer information for reference individuals.
  • NLST National Lung Screening Trial
  • the training data can be obtained and collected independent of publicly available databases e.g., by capturing images of one or more lymph nodes from a plurality of reference individuals.
  • such training data can be a custom dataset. Training data of a custom dataset can be acquired during patient care and therefore, can have corresponding surgical sampling information of the lymph nodes.
  • the training data can include N stage cancer information for the reference individuals e.g., N stage information previously diagnosed by a clinician based on imaging and/or biopsy results for the reference individuals.
  • the reference ground truths that indicate whether a lymph node is cancerous is nodal stage (N stage) information that is derived from clinical stage data.
  • N stage nodal stage
  • the publicly available NLST dataset includes clinical stage information (e.g., stages la, lb, 2a, 2b, 3a, etc.) for reference individuals but does not include lymph node specific information.
  • the NLST dataset does not distinguish whether individual lymph nodes are cancerous or non-cancerous.
  • the reference ground truths that indicate whether a lymph node is cancerous can be back calculated based on the N stage information and other characteristics of the reference individual’s cancer (e.g., tumor size, tumor location, tumor shape, and/or tumor density) and/or the lymph node (e.g., lymph node size, lymph node density, and/or lymph node shape).
  • cancer e.g., tumor size, tumor location, tumor shape, and/or tumor density
  • the lymph node e.g., lymph node size, lymph node density, and/or lymph node shape
  • a reference individual with a N stage of 0 or 1 has not yet encountered cancer metastasis and therefore, the lymph nodes of this reference individual are assigned a reference ground truth of “negative” or non-cancerous.
  • a reference individual with a N stage of 2 or 3 has encountered cancer metastasis. Therefore, based on the characteristics of the tumor and/or the lymph node, the lymph node is assigned a reference ground truth of “positive” or “negative
  • the imputation module 150 deploys the trained imputation model to analyze features extracted from an image including one or more lymph nodes and predicts probabilities for the one or more lymph nodes that the lymph nodes are cancerous.
  • the trained imputation model is deployed against a test dataset.
  • the test dataset and the training dataset used to train the imputation model may be derived from a common dataset.
  • the common dataset may be split into two subsets for training and testing the imputation model.
  • the common dataset undergoes a 50:50 training testing dataset split.
  • the common dataset undergoes a 60:40 training testing dataset split.
  • the common dataset undergoes a 80:20 training testing dataset split.
  • the deployment of the trained imputation model yields probabilities (e.g., continuous probabilities between 0 and 1) for lymph nodes, each probability indicating whether the corresponding lymph node is likely to be cancerous.
  • the imputation module 150 provides one or more of the lymph node metastatic risks for the lymph nodes predicted by the imputation model to be used to train the risk model.
  • the imputation module 150 provides all of the lymph node metastatic risks for the lymph nodes predicted by the imputation model to be used to train the risk model.
  • the imputation module 150 performs a selection process to identify a subset of the lymph node metastatic risks for the lymph nodes predicted by the imputation model.
  • the imputation module 150 selects the subset of lymph node metastatic risks by comparing the lymph node metastatic risks to a threshold value. If a lymph node metastatic risk for a lymph node is above a threshold value, the lymph node metastatic risk for the lymph node is selected for inclusion in the subset for training the risk model. If a lymph node metastatic risk for a lymph node is below a threshold value, the lymph node metastatic risk for the lymph node is excluded from the subset and therefore, not used fortraining the risk model.
  • the threshold value is a fixed value. In various embodiments, the threshold value is a 50% risk of metastatic cancer. In various embodiments, the threshold value is 30%, 35%, 40%, 45%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, or 99% risk of metastatic cancer. In some embodiments, the threshold value is between 30 and 70%. In some embodiments, the threshold value is between 40 and 60%. In some embodiments, the threshold value is between 45 and 55%. In some embodiments, the threshold value is between 48 and 52%. In some embodiments, the threshold value is between 49 and 51%. In various embodiments, the threshold value is variable. In various embodiments, the threshold value is a median intra-subject probability for having cancer.
  • the imputation module 150 provides the subset of lymph node metastatic risks to the risk training module 155 for training the risk model.
  • the risk training module 155 trains the risk model using training data including at least one or more of the lymph node metastatic risks for the lymph nodes predicted by the imputation model.
  • the one or more lymph node metastatic risks may be the subset of lymph node metastatic risks selected by the imputation module 150, as is described above.
  • the subset of lymph node metastatic risks for training the risk model represents a purified training dataset in comparison to the training dataset that was used to train the imputation model.
  • the purified training dataset includes the subset of lymph nodes that are more likely to be cancerous whereas lymph nodes that are less likely to be cancerous (but otherwise may represent false positives or be confounding for training the risk model) are removed from the purified training dataset.
  • the risk model predicts a greater range of probabilities for all of the lymph nodes in those subjects with metastatic disease. More specifically, the risk model is more likely to assign lymph nodes likely to have cancer with a higher lymph node metastatic risk and lymph nodes unlikely to have cancer with a lower lymph node metastatic risk.
  • the risk model analyzes features extracted from an image including one or more lymph nodes and predicts a continuous probability for each of the one or more lymph nodes that the lymph node is cancerous. Thus, based on the extracted features from an image including one or more lymph nodes, the risk model determines a probability for each of the lymph nodes, the determined probability indicating a likelihood that the lymph is cancerous (e.g., metastatic cancer). Here, the risk model determines risks of metastatic cancer at the lymph-node level. In various embodiments, the risk model is similar to the imputation model in that both models analyze features extracted from images and predict risks of metastatic cancer at the lymph-node level.
  • the risk model is trained using predictions of the imputation model e.g.., one or more lymph-node level risks predicted by the imputation model. Altogether, this enables the risk model to predict LN-level risks of metastatic cancer with a greater range of probabilities than the imputed LN-level risks predicted by the imputation model.
  • the risk model is trained using training data that includes at least one or more lymph node metastatic risks predicted by the imputation model.
  • the one or more lymph node metastatic risks can serve as reference ground truths fortraining the risk model.
  • the training data can additionally include extracted features from training images including one or more lymph nodes obtained from reference individuals.
  • the training images used to train the risk model can be the same training images previously used to train the imputation model. In some embodiments, the training images used to train the risk model are different from the training images previously used to train the imputation model.
  • the training data includes individual training examples, such that the risk model is trained using each training example.
  • each training example can include extracted features of one or more lymph nodes of a particular training image as well as a reference ground truth (e.g., lymph node metastatic risk for the same one or more lymph nodes of the particular training image).
  • the risk model is trained to minimize a loss function (e.g., a loss function for any of a least absolute shrinkage and selection operator (LASSO) regression, Ridge regression, or ElasticNet regression).
  • LASSO least absolute shrinkage and selection operator
  • the risk model is any one of a regression model (e.g., linear regression, logistic regression, or polynomial regression), decision tree, random forest, support vector machine, Naive Bayes model, k-means cluster, or neural network (e.g., feedforward networks, convolutional neural networks (CNN), deep neural networks (DNN), autoencoder neural networks, generative adversarial networks, or recurrent networks (e.g., long short-term memory networks (LSTM), bi-directional recurrent networks, deep bidirectional recurrent networks), or any combination thereof.
  • the risk model is a random forest classifier or a least absolute shrinkage and selection operator (LASSO) classifier.
  • LASSO least absolute shrinkage and selection operator
  • the risk model is a machine learned model.
  • the risk model can be trained using a machine learning implemented method, such as any one of a linear regression algorithm, logistic regression algorithm, decision tree algorithm, support vector machine classification, Naive Bayes classification, K-Nearest Neighbor classification, random forest algorithm, deep learning algorithm, gradient boosting algorithm, and dimensionality reduction techniques such as manifold learning, principal component analysis, factor analysis, autoencoder regularization, and independent component analysis, or combinations thereof.
  • the risk model is trained using supervised learning algorithms, unsupervised learning algorithms, semi-supervised learning algorithms (e.g., partial supervision), weak supervision, transfer, multi-task learning, or any combination thereof.
  • the risk model has one or more parameters, such as hyperparameters or model parameters.
  • Hyperparameters are generally established prior to training. Examples of hyperparameters include the learning rate, depth or leaves of a decision tree, number of hidden layers in a deep neural network, number of clusters in a k- means cluster, penalty in a regression model, and a regularization parameter associated with a cost function.
  • Model parameters are generally adjusted during training. Examples of model parameters include weights associated with nodes in layers of neural network, support vectors in a support vector machine, and coefficients in a regression model. The model parameters of the risk model are trained (e.g., adjusted) using the training data to improve the predictive capacity of the risk model.
  • the risk deployment module 160 deploys the risk model to determine a subject level risk of metastatic cancer for a subject (e.g., subject 110 described above in FIG. 1A).
  • the risk deployment module 160 provides the extracted features of images including one or more lymph nodes captured from the subject as input to the trained risk model.
  • the trained risk model analyzes the extracted features and outputs probabilities for the lymph nodes that indicate whether individual lymph nodes are likely to be cancerous.
  • the risk deployment module 160 transitions from individual lymph node level risks of metastatic cancer predicted by the risk model to a subject level risk of metastatic cancer for the subject.
  • the risk deployment module 160 combines the individual lymph node level risks of metastatic cancer to determine the subject level risk of metastatic cancer.
  • the risk deployment module 160 performs a statistical operation across the lymph node level risks of metastatic cancer. The statistical operation can identify a maximum value, minimum value, average value, median value, or mode value across the individual lymph node level risks of metastatic cancer. In particular embodiments, the statistical operation is a maximum value across the individual lymph node level risks of metastatic cancer.
  • the cancer staging module 165 determines a cancer stage for the subject based on the subject level risk of metastatic disease for the subject. In various embodiments, the cancer staging module 165 determines a cancer stage for the subject by comparing the subject level risk of metastatic disease for the subject to one or more threshold risk values. These threshold risk values can serve as cutoffs for determining whether a cancer in the subject is likely in any of stage 0, stage 1, stage 2, or stage 3.
  • the subject level risk is greater than 30%, the subject is categorized as having at least stage 3 cancer. In some embodiments, if the subject level risk is greater than 40%, greater than 50%, greater than 60%, greater than 70%, greater than 80%, or greater than 90% risk, the subject is categorized as having at least stage 3 cancer.
  • the threshold risk values define risk ranges corresponding to one or more cancer stages. Thus, threshold risk values set the upper and lower bounds of the risk ranges. In some embodiments, if the subject level risk is between 30% and 100%, the subject is categorized as having nodal stage 3 cancer. In some embodiments, if the subject level risk is between 40% and 100%, the subject is categorized as having nodal stage 3 cancer. In some embodiments, if the subject level risk is between 50% and 100%, the subject is categorized as having nodal stage 3 cancer.
  • the subject level risk is between 10% and 40%, the subject is categorized as having nodal stage 2 cancer. In some embodiments, if the subject level risk is between 15% and 40%, the subject is categorized as having nodal stage 2 cancer. In some embodiments, if the subject level risk is between 20% and 40%, the subject is categorized as having nodal stage 2 cancer. In some embodiments, if the subject level risk is between 30% and 40%, the subject is categorized as having nodal stage 2 cancer. In some embodiments, if the subject level risk is between 15% and 30%, the subject is categorized as having nodal stage 2 cancer. In some embodiments, if the subject level risk is between 20% and 30%, the subject is categorized as having nodal stage 2 cancer.
  • the subject level risk is between 5% and 20%, the subject is categorized as having either nodal stage 0 or nodal stage 1 cancer. In some embodiments, if the subject level risk is between 10% and 20%, the subject is categorized as having either nodal stage 0 or nodal stage 1 cancer. In some embodiments, if the subject level risk is between 5% and 15%, the subject is categorized as having either nodal stage 0 or nodal stage 1 cancer. In some embodiments, if the subject level risk is between 8% and 12%, the subject is categorized as having either nodal stage 0 or nodal stage 1 cancer.
  • FIG. 3A depicts an example implementation of the imputation model and the risk model, in accordance with an embodiment.
  • reference individuals 305A and 305B may be identified with a particular N stage.
  • reference individual 305 A is identified as N stage negative, indicating that reference individual 305A does not exhibit lymph node cancer involvement.
  • Reference individual 305B is identified as N stage positive, indicating that reference individual 305B exhibits lymph node cancer involvement.
  • the subject level N stage does not distinguish between particular lymph nodes and whether individual lymph nodes are cancerous or non-cancerous.
  • the imputation model 310 imputes metastatic risk for each lymph node (e.g., lymph node metastatic risk), thereby enabling the differentiation between lymph nodes with metastatic cancer and lymph nodes that are non-cancerous.
  • the imputation model 310 predicts a probability for each lymph node that represents a likelihood that the lymph node is cancerous.
  • the output of the imputation model 310 can identify that individual 305C (corresponding to reference individual 305A) does not exhibit lymph node involvement. In other words, none of the lymph nodes of individual 305C are likely to be cancerous. This confirms the N stage negative identification. Additionally, the output of imputation model 310 can identify that individual 305D (corresponding to reference individual 305B) exhibits lymph node involvement for one or more lymph nodes. For example, the imputation model 310 can impute probabilities such that exactly one lymph node in individual 305D is identified as cancerous whereas other lymph nodes in individual 305D are non-cancerous.
  • one or more of the lymph node metastatic risks for the lymph nodes predicted by the imputation model 310 are used to train the risk model 320.
  • all of the lymph node metastatic risks for the lymph nodes predicted by the imputation model 310 are used to train the risk model 320.
  • a subset of the lymph node metastatic risk corresponding to a subset of the lymph nodes are used to train the risk model 320, thereby enabling the risk model to predict a greater range of probabilities for all of the lymph nodes in those subjects with metastatic disease.
  • the risk model 320 is deployed to predict subject level risk of metastatic cancer for one or more subjects 315.
  • the risk model 320 is deployed to transition from lymph node level risks of metastatic disease to a subject level risk of metastatic disease.
  • the risk model 320 is deployed to analyze features extracted from images captured from individuals 315.
  • the risk model 320 predicts subject level risks of metastatic cancer for reach of the individuals 315.
  • the risk model 320 predicts a 20% risk of metastatic cancer for individual 315A, a 30% risk of metastatic cancer for individual 315B, a 50% risk of metastatic cancer for individual 315C, and a 60% risk of metastatic cancer for individual 315D.
  • individuals 315A and 315B can be categorized with an earlier stage cancer (e.g., stage 0 or 1) whereas individuals 315C and 315D can be categorized with a later stage cancer (e.g., stage 2, 3, or 4).
  • Individuals 315 can receive different treatments according to their subject level risk of metastatic disease. For example, individuals 315A and 315B who are at low risk of having lymph node metastasis could proceed directly to immediate therapy such as surgical resection or combined bronchoscopic diagnosis and treatment. As another example, individuals 315C and 315D who are at higher risk of having lymph node metastasis can receive additional testing such as PET scanning and lymph node biopsy.
  • FIG. 3B depicts the implementation of the risk model 320 in further detail, in accordance with an embodiment.
  • the risk model 320 analyzes extracted features from an image captured from subject 350 and outputs probabilities for each of one or more lymph nodes for a subject. As shown in FIG. 3B, the risk model 320 outputs probabilities for three lymph nodes. In other embodiments, the risk model 320 can output additional or fewer lymph node probabilities for the subject 350.
  • the risk model 320 predicts that a first lymph node has a 20% probability of being cancerous and conversely, a 80% probability of being non-cancerous.
  • the risk model 320 predicts that a second lymph node has a 30% probability of being cancerous and conversely, a 70% probability of being non-cancerous. Additionally, the risk model 320 predicts that a third lymph node has a 60% probability of being cancerous and conversely, a 40% probability of being non-cancerous.
  • the individual lymph node probabilities are analyzed to identify a subject level risk of metastatic cancer. In this example, the maximum lymph node probability of metastatic cancer is taken as the subject level probability of metastatic cancer.
  • the third lymph node with a 60% probability of being cancerous is the highest probability amongst the lymph nodes. Thus, the subject level probability of metastatic cancer for the subject 350 is 60%.
  • the cancer in the subject can include one or more of: lymphoma, B cell lymphoma, T cell lymphoma, mycosis fungoides, Hodgkin's Disease, myeloid leukemia, bladder cancer, brain cancer, nervous system cancer, head and neck cancer, squamous cell carcinoma of head and neck, kidney cancer, lung cancer, neuroblastoma/glioblastoma, ovarian cancer, pancreatic cancer, prostate cancer, skin cancer, liver cancer, melanoma, squamous cell carcinomas of the mouth, throat, larynx, and lung, colon cancer, cervical cancer, cervical carcinoma, breast cancer, epithelial cancer, gastrointestinal cancer, renal cancer, genitourinary cancer, pulmonary cancer, esophageal carcinoma, stomach cancer, thyroid cancer, head and neck carcinoma, large bowel cancer, hematopoietic cancer, testicular cancer, colon
  • the cancer can be any one of lung bronchioloalveolar carcinoma (BAC), bladder cancer, a female genital tract malignancy (e.g., uterine serous carcinoma, endometrial carcinoma, vulvar squamous cell carcinoma, and uterine sarcoma), an ovarian surface epithelial carcinoma (e.g., clear cell carcinoma of the ovary, epithelial ovarian cancer, fallopian tube cancer, and primary peritoneal cancer), breast carcinoma, nonsmall cell lung cancer (NSCLC), a male genital tract malignancy (e.g., testicular cancer), retroperitoneal or peritoneal carcinoma, gastroesophageal adenocarcinoma, esophagogastric junction carcinoma, liver hepatocellular carcinoma, esophageal and esophagogastric junction carcinoma, cervical cancer, cholangiocarcinoma, pancreatic adenocarcinoma, extrahepatic bile duct
  • BAC lung
  • the cancer is breast cancer, non-small cell lung cancer, bladder cancer, kidney cancer, colon cancer, and melanoma.
  • the cancer in the subject can be a metastatic cancer, including any one of bladder cancer, breast cancer, colon cancer, kidney cancer, lung cancer, melanoma, ovarian cancer, pancreatic cancer, prostatic cancer, rectal cancer, stomach cancer, thyroid cancer, or uterine cancer.
  • the cancer is a lung cancer.
  • the cancer is a type of lung cancer, including any one of small cell lung cancer, non-small cell lung cancer, adenocarcinoma, squamous cell cancer, large cell carcinoma, small cell carcinoma, combined small cell carcinoma, lung sarcoma, lung lymphoma, bronchial carcinoids.
  • Embodiments described herein involve the determination of a cancer staging by determining a subject level risk of metastatic cancer in a subject.
  • an intervention is provided to a subject based on the cancer staging and/or the subject level risk of metastatic cancer predicted for the subject.
  • Such interventions can be any one of: application of a diagnostic, application of a therapeutic agent, subsequent testing of the subject, subsequent biopsy (e.g., cancer biopsy or lymph node biopsy, bronchoscopic sampling of lymph nodes), subsequent image scanning (e.g., PET scanning, MRI scanning, ultrasound imaging, or X-ray imaging), tumor resection, bronchoscopic diagnosis, selection and/or administration of therapeutic(s), selection/administration of pharmaceutical composition, enrollment in a clinical trial, nonenrollment in a clinical trial, or any combination thereof.
  • subsequent biopsy e.g., cancer biopsy or lymph node biopsy, bronchoscopic sampling of lymph nodes
  • subsequent image scanning e.g., PET scanning, MRI scanning, ultrasound imaging, or X-ray imaging
  • tumor resection e.g., bronchoscopic diagnosis, selection and/or administration of therapeutic(s)
  • selection/administration of pharmaceutical composition e.g., enrollment in a clinical trial, nonenrollment in a clinical trial, or any combination thereof.
  • the subject level risk of metastatic cancer predicted for the subject or the predicted cancer stage indicates that the subject is at high risk of metastatic cancer.
  • a subject at high risk of metastatic cancer has a predicted subject level risk of metastatic cancer that is above a threshold score.
  • the threshold score is above 20% risk of metastatic cancer.
  • the threshold score is above 25% risk of metastatic cancer.
  • the threshold score is above 30%, above 35%, above 40%, above 45%, above 50%, above 55%, above 60%, above 65%, above 70%, above 75%, above 80%, above 85%, above 9%, above 95%, or above 99% risk of metastatic cancer.
  • the threshold score is between 20% and 80%.
  • the threshold value is between 30 and 70%. In some embodiments, the threshold value is between 40 and 60%. In some embodiments, the threshold value is between 45 and 55%. In some embodiments, the threshold value is between 20 and 60%. In some embodiments, the threshold value is between 30 and 55%. In some embodiments, the threshold value is between 35 and 50%. In some embodiments, the threshold value is between 40 and 45%.
  • a subject is at high risk of metastatic cancer if the predicted cancer stage for the subject is at least stage 2 (e.g., stage 2 or stage 3).
  • the guided intervention involves one of subsequent testing of the subject, subsequent biopsy of the subject, or subsequent image scanning of the subject. This allows clinical providers to prioritize the subsequent testing, biopsy, or image scanning of subjects that are at highest risk.
  • procedures that are inappropriate for late-stage metastatic cancer can be ruled out for the subject.
  • procedures that are ruled out include tumor resection, lung resection, radiation (e.g., external beam radiation), chemotherapy (e.g., local or systemic chemotherapy), immunomodulatory therapy, or even general anesthesia.
  • the subject if the subject is predicted to be at a high risk of metastatic cancer, the subject undergoes a subsequent image scanning to confirm the prediction. For example, if the subject is predicted to be at a high risk of metastatic cancer, the subject further undergoes PET scanning. In such embodiments, if the PET scanning returns a negative result (e.g., no lymph node involvement), the subject can undergo further testing. For example, the subject can further undergo a subsequent biopsy to definitively sample the lymph node. In some embodiments, the subsequent biopsy is a bronchoscopic sampling of lymph nodes.
  • the subject directly undergoes a subsequent biopsy to confirm the prediction.
  • the subject need not undergo subsequent image scanning to confirm the prediction and instead, directly undergoes a subsequent biopsy.
  • the subject further undergoes bronchoscopic sampling of lymph nodes.
  • the subject level risk of metastatic cancer can be determined from individual lymph node level risks across the lymph nodes. Thus, for a subject who is predicted to have high risk of metastatic cancer based on the subject level of risk, the particular lymph node(s) that led to the high subject level of risk can be identified.
  • those particular lymph nodes can be targeted for subsequent testing or treatment.
  • those particular lymph nodes can undergo subsequent image scanning (e.g., PET scanning) to confirm the prediction.
  • those particular lymph nodes can undergo a subsequent biopsy to definitively sample the particular lymph nodes.
  • the subsequent biopsy can confirm the presence or absence of metastatic cancer in those particular lymph nodes.
  • the subject can be enrolled in a clinical trial.
  • a clinical trial may be testing therapeutic efficacy of a treatment against metastatic cancer and therefore, it may specify in its eligibility criteria that enrollees are to be exhibiting metastatic cancer.
  • the subject is withheld from enrollment in a clinical trial.
  • such a clinical trial may be testing therapeutic efficacy of a treatment against a non-metastatic form of cancer.
  • the subject level risk of metastatic cancer predicted for the subject indicates that the subject is at low risk of metastatic cancer.
  • a subject at low risk of metastatic cancer has a predicted subject level risk that is below a threshold score.
  • the threshold score is below 20% risk of metastatic cancer.
  • the threshold score is below 25% risk of metastatic cancer.
  • the threshold score is below 30%, below 35%, below 40%, below 45%, below 50%, below 55%, below 60%, below 65%, below 70%, below 75%, below 80%, below 85%, below 90%, below 95%, or below 99% risk of metastatic cancer.
  • the threshold value is between 5% and 40%.
  • the threshold value is between 10 and 30%. In some embodiments, the threshold value is between 15 and 25%. In some embodiments, the threshold value is between 18 and 22%.
  • a subject is at low risk of metastatic cancer if the predicted cancer stage for the subject is below stage 2 (e.g., stage 0 or stage 1).
  • the guided intervention involves one of tumor resection, bronchoscopic diagnosis, selection and/or administration of therapeutic(s), and/or selection/administration of pharmaceutical composition. This enables the rapid identification and treatment of subjects where metastasis has not yet occurred.
  • a therapeutic agent can be selected and/or administered to the subject based on the subject level risk of metastatic cancer, the selected therapeutic agent likely to exhibit efficacy against the cancer.
  • exemplary therapeutic agents include chemotherapies, radiation, antigen-specific monoclonal antibodies, or immunotherapies.
  • a subject at high risk of metastatic cancer is provided a therapeutic agent.
  • a subject at high risk of metastatic cancer has a predicted subject level risk of metastatic cancer that is above a threshold score.
  • the threshold score is above 20% risk of metastatic cancer.
  • the threshold score is above 25% risk of metastatic cancer.
  • the threshold score is above 30%, above 35%, above 40%, above 45%, above 50%, above 55%, above 60%, above 65%, above 70%, above 75%, above 80%, above 85%, above 90%, above 95%, or above 99% risk of metastatic cancer.
  • a subject is at high risk of metastatic cancer if the predicted cancer stage for the subject is at least stage 2 (e.g., stage 2, stage 3, or stage 4).
  • a subject at low risk of metastatic cancer is provided a therapeutic agent.
  • a subject at low risk of metastatic cancer has a predicted subject level risk that is below a threshold score.
  • the threshold score is below 20% risk of metastatic cancer.
  • the threshold score is below 25% risk of metastatic cancer.
  • the threshold score is below 30%, below 35%, below 40%, below 45%, below 50%, below 55%, below 60%, below 65%, below 70%, below 75%, below 80%, below 85%, below 90%, below 95%, or below 99% risk of metastatic cancer.
  • a subject is at low risk of metastatic cancer if the predicted cancer stage for the subject is below stage 2 (e.g., stage 0 or stage 1).
  • the subject can be directed to early-stage cancer treatments.
  • early-stage cancer treatments include tumor resection, lung resection, chemotherapy (e.g., local or systemic chemotherapy), locally delivered therapies (e.g., transbronchial needle instillation or catheter based ablation), or radiation.
  • the subject undergoes a subsequent image scanning to confirm the prediction. For example, if the subject is predicted to be at a low risk of metastatic cancer, the subject further undergoes PET scanning. In such embodiments, if the PET scanning returns a negative result (e.g., no lymph node involvement), then the PET scan results align with the prediction. However, if the PET scanning returns a positive result (e.g., lymph node involvement), the subject can undergo further testing. For example, the subject can further undergo a subsequent biopsy to definitively sample the lymph node. In some embodiments, the subsequent biopsy is a bronchoscopic sampling of lymph nodes.
  • the therapeutic agent is a biologic, e.g. a cytokine, antibody, soluble cytokine receptor, anti-sense oligonucleotide, siRNA, etc.
  • biologic agents encompass muteins and derivatives of the biological agent, which derivatives can include, for example, fusion proteins, PEGylated derivatives, cholesterol conjugated derivatives, and the like as known in the art.
  • antagonists of cytokines and cytokine receptors e.g. traps and monoclonal antagonists.
  • biosimilar or bioequivalent drugs to the active agents set forth herein.
  • Therapeutic agents for lung cancer can include chemotherapeutics such as docetaxel, cisplatin, carboplatin, gemcitabine, Nab-paclitaxel, paclitaxel, pemetrexed, gefitinib, erlotinib, brigatinib (Alunbrig®), capmatinib (Tabrecta®), selpercatinib (Retevmo®), entrectinib (Rozlytrek®), lorlatinib (Lorbrena®), larotrectinib (Vitrakvi®), dacomitinib (Vizimpro®), and vinorelbine.
  • chemotherapeutics such as docetaxel, cisplatin, carboplatin, gemcitabine, Nab-paclitaxel, paclitaxel, pemetrexed, gefitinib, erlotinib, brigatinib (Alunbrig®), capmatinib
  • Therapeutic agents for lung cancer can include antibody therapies such as durvalumab (Imfinzi®), nivolumab (Opdivo®), pembrolizumab (Keytruda®), atezolizumab (Tecentriq®), and ramucirumab.
  • antibody therapies such as durvalumab (Imfinzi®), nivolumab (Opdivo®), pembrolizumab (Keytruda®), atezolizumab (Tecentriq®), and ramucirumab.
  • a pharmaceutical composition can be selected and/or administered to the subject based on the subject level risk of metastatic cancer , the selected therapeutic agent likely to exhibit efficacy against the cancer.
  • a pharmaceutical composition administered to an individual includes an active agent such as the therapeutic agent described above.
  • the active ingredient is present in a therapeutically effective amount, i.e., an amount sufficient when administered to treat a disease or medical condition mediated thereby.
  • the compositions can also include various other agents to enhance delivery and efficacy, e.g. to enhance delivery and stability of the active ingredients.
  • the compositions can also include, depending on the formulation desired, pharmaceutically acceptable, nontoxic carriers or diluents, which are defined as vehicles commonly used to formulate pharmaceutical compositions for animal or human administration.
  • the diluent is selected so as not to affect the biological activity of the combination.
  • examples of such diluents are distilled water, buffered water, physiological saline, PBS, Ringer’s solution, dextrose solution, and Hank’s solution.
  • the pharmaceutical composition or formulation can include other carriers, adjuvants, or non-toxic, nontherapeutic, nonimmunogenic stabilizers, excipients and the like.
  • the compositions can also include additional substances to approximate physiological conditions, such as pH adjusting and buffering agents, toxicity adjusting agents, wetting agents and detergents.
  • the composition can also include any of a variety of stabilizing agents, such as an antioxidant.
  • compositions described herein can be administered in a variety of different ways. Examples include administering a composition containing a pharmaceutically acceptable carrier via oral, intranasal, rectal, topical, intraperitoneal, intravenous, intramuscular, subcutaneous, subdermal, transdermal, intrathecal, or intracranial method. In various embodiments, pharmaceutical compositions can be administered via catheter based deployment (e.g., using flexible or rigid robotic based bronchoscopy).
  • Such a pharmaceutical composition may be administered for treatment (e.g., after diagnosis of a patient with lung cancer) purposes.
  • Preventing, prophylaxis or prevention of a disease or disorder as used in the context of this invention refers to the administration of a composition to prevent the occurrence, onset, progression, or recurrence of lung cancer some or all of the symptoms of lung cancer or to lessen the likelihood of the onset of lung cancer.
  • Treating, treatment, or therapy of lung cancer shall mean slowing, stopping or reversing the cancer’s progression by administration of treatment according to the present invention.
  • treating lung cancer means reversing the cancer’s progression, ideally to the point of eliminating the cancer itself.
  • the methods of the invention including the methods of for cancer staging by determining a subject level risk of metastatic cancer of a subject, are, in some embodiments, performed on one or more computers.
  • a machine -readable storage medium comprising a data storage material encoded with machine readable data which, when using a machine programmed with instructions for using said data, is capable of displaying any of the datasets and execution and results of the models described herein.
  • the invention can be implemented in computer programs executing on programmable computers, comprising a processor, a data storage system (including volatile and non-volatile memory and/or storage elements), a graphics adapter, a pointing device, a network adapter, at least one input device, and at least one output device.
  • a display is coupled to the graphics adapter.
  • Program code is applied to input data to perform the functions described above and generate output information.
  • the output information is applied to one or more output devices, in known fashion.
  • the computer can be, for example, a personal computer, microcomputer, or workstation of conventional design.
  • Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system.
  • the programs can be implemented in assembly or machine language, if desired. In any case, the language can be a compiled or interpreted language.
  • Each such computer program is preferably stored on a storage media or device (e.g., ROM or magnetic diskette) readable by a general or special purpose programmable computer, for configuring and operating the computer when the storage media or device is read by the computer to perform the procedures described herein.
  • the system can also be considered to be implemented as a computer-readable storage medium, configured with a computer program, where the storage medium so configured causes a computer to operate in a specific and predefined manner to perform the functions described herein.
  • the signature patterns and databases thereof can be provided in a variety of media to facilitate their use.
  • Media refers to a manufacture that contains the signature pattern information of the present invention.
  • the databases of the present invention can be recorded on computer readable media, e.g. any medium that can be read and accessed directly by a computer.
  • Such media include, but are not limited to: magnetic storage media, such as floppy discs, hard disc storage medium, and magnetic tape; optical storage media such as CD-ROM; electrical storage media such as RAM and ROM; and hybrids of these categories such as magnetic/optical storage media.
  • magnetic storage media such as floppy discs, hard disc storage medium, and magnetic tape
  • optical storage media such as CD-ROM
  • electrical storage media such as RAM and ROM
  • hybrids of these categories such as magnetic/optical storage media.
  • Recorded refers to a process for storing information on computer readable medium, using any such methods as known in the art. Any convenient data storage structure can be chosen, based on the means used to access the stored information. A variety of data processor programs and formats can be used for storage, e.g. word processing text file, database format, etc.
  • the methods of the invention are performed on one or more computers in a distributed computing system environment (e.g., in a cloud computing environment).
  • a distributed computing system environment e.g., in a cloud computing environment.
  • cloud computing is defined as a model for enabling on-demand network access to a shared set of configurable computing resources. Cloud computing can be employed to offer on-demand access to the shared set of configurable computing resources.
  • the shared set of configurable computing resources can be rapidly provisioned via virtualization and released with low management effort or service provider interaction, and then scaled accordingly.
  • a cloud-computing model can be composed of various characteristics such as, for example, on-demand self-service, broad network access, resource pooling, rapid elasticity, measured service, and so forth.
  • a cloudcomputing model can also expose various service models, such as, for example, Software as a Service (“SaaS”), Platform as a Service (“PaaS”), and Infrastructure as a Service (“laaS”).
  • SaaS Software as a Service
  • PaaS Platform as a Service
  • laaS Infrastructure as a Service
  • a cloud-computing model can also be deployed using different deployment models such as private cloud, community cloud, public cloud, hybrid cloud, and so forth.
  • a “cloud-computing environment” is an environment in which cloud computing is employed.
  • FIG. 4 illustrates an example computer for implementing the entities shown in FIG. 1A, IB, 2A, 2B, 3A, and 3B.
  • the computer 400 includes at least one processor 402 coupled to a chipset 404.
  • the chipset 404 includes a memory controller hub 420 and an input/output (I/O) controller hub 422.
  • a memory 406 and a graphics adapter 412 are coupled to the memory controller hub 420, and a display 418 is coupled to the graphics adapter 412.
  • a storage device 408, an input device 414, and network adapter 416 are coupled to the I/O controller hub 422.
  • Other embodiments of the computer 400 have different architectures.
  • the storage device 408 is a non-transitory computer-readable storage medium such as a hard drive, compact disk read-only memory (CD-ROM), DVD, or a solid-state memory device.
  • the memory 406 holds instructions and data used by the processor 402.
  • the input interface 414 is a touch-screen interface, a mouse, track ball, or other type of pointing device, a keyboard, or some combination thereof, and is used to input data into the computer 400.
  • the computer 400 may be configured to receive input (e.g., commands) from the input interface 414 via gestures from the user.
  • the graphics adapter 412 displays images and other information on the display 418.
  • the network adapter 416 couples the computer 400 to one or more computer networks.
  • the computer 400 is adapted to execute computer program modules for providing functionality described herein.
  • module refers to computer program logic used to provide the specified functionality.
  • program modules are stored on the storage device 408, loaded into the memory 406, and executed by the processor 402.
  • the types of computers 400 used by the entities of FIGs. 1A or IB can vary depending upon the embodiment and the processing power required by the entity.
  • the cancer staging system 130 can run in a single computer 400 or multiple computers 400 communicating with each other through a network such as in a server farm.
  • the computers 400 can lack some of the components described above, such as graphics adapters 412, and displays 418.
  • cancer staging system 130 can include at least the cancer staging system 130 described above in FIG. 1A.
  • the cancer staging system 130 is embodied as a computer system, such as a computer system with example computer 400 described in FIG. 4.
  • the system includes an imaging device, such as imaging device 120 described above in FIG. 1A.
  • the system includes both the cancer staging system 130 (e.g., a computer system) and an imaging device.
  • the cancer staging system 130 can be communicatively coupled with the imaging device 120 to receive images (e.g., CT scans) captured from a subject.
  • the computer system implements, in silico, one or both of the imputation model and the risk model to analyze the images and to determine a subject level risk of nodal disease for the subject.
  • Example 1 Study Population and Feature Extraction
  • Feature categories included first order features, shape features, gray level co-occurrence matrix 1 (GLCM) features, gray level run length matrix (GLRLM) features, gray level size zone matrix (GLSZM) features, and neighborhood gray tone difference matrix (NGTDM) features).
  • GLCM gray level co-occurrence matrix 1
  • GLRLM gray level run length matrix
  • GLSZM gray level size zone matrix
  • NTTDM neighborhood gray tone difference matrix
  • FIGs. 5A-5R depict example correlation heat maps of different lymph node radiomic features across the different panels for the GLCM feature category. Additionally, FIGs. 5A- 5R each show specific features. Examples of such features shown in FIGs. 5A-5R include correlation features, LMC2, MCC, Difference in Entropy, Joint Entropy, Sum of Entropy, Cluster prominence, average difference, joint average, sum of average, autocorrelation, difference in variance, contrast, cluster tendency, sum of squares, LDMN, LDN, Cluster shade, LMC1, Inverse variance, LD, LDM, joint energy, and maximum probability.
  • FIGs. 5C-5R show the features with respect to different filters. Specifically, FIGs.
  • 5C-5R provide designations of “XYZ” where “X”, “Y”, and “Z” each refer to a filter applied to the X, Y, or Z directions of the image.
  • a “H” refers to a high pass filter whereas a “L” refers to a low pass filter. Therefore, a designation of “HHH” refers to features derived from a 3D image that has undergone high pass filtering in each of the X, Y, and Z directions.
  • a designation of “HHL” refers to features derived from a 3D image that has undergone high pass filtering in the X and Y directions, and has undergone low pass filtering in the Z direction.
  • the metrics were normalized (z-score) and Random Forest and LASSO Classifiers were trained and tested. For each LN, a continuous probability that it was cancerous was generated. For example, there could be a LN with a 20% probability of being cancerous (by subtraction means it was assigned an 80% probability of being free of cancer). All LN within and across study participants were considered to be independent data.
  • the distribution of scores e.g., max probabilities
  • a two-step bootstrap approach was performed to transition from LN level risk of metastatic disease to subject level risk of metastatic disease.
  • the two-step bootstrap approach included the implementation of an imputation model and a risk model, as described above in reference to FIG. 3A.
  • the challenge was that a subject’s N stage is imputed to all the LN stations to train a classifier. This imputation introduces error at the LN station given that most LN are cancer free.
  • IMPUTATION The lymph nodes most likely to be involved by cancer was identified for each subject with known metastatic disease. This was done for each subject by selecting the LNs with probabilities of being cancerous (output from last model) above the median intrasubject value (Imputation process involving the imputation model shown in FIG. 3A). LN below the median intra-subject probability for having cancer were classified as negative.
  • CLASSIFICATION The cohort was split 50:50 training:testing. A second classifier (e.g., risk model 320 in FIG. 3A) was then trained using the imputed pool of LN outcomes to predict the presence nodal disease for each subject.
  • a second classifier e.g., risk model 320 in FIG. 3A
  • the performance of the two-step bootstrap model was an improvement in comparison to the approach described in Example 2. Similar performance was achieved using a LASSO Classifier.
  • the model can be used to plot the distribution of scores (or max probabilities) across N stage (FIG. 9B). It is interesting to note that the model yields a higher max probability for those cases with higher N stage. This probability score could be used in different clinical contexts to enable different triage decision making tools. Additional data fortraining and testing could provide a model that predicts N as a categorical rather than binary covariable.
  • Example 4 Example Guided Interventions using CT Image-Based Prediction Device
  • An image-based device that implements the methods described herein (e.g., deploying a risk model for predicting lymph node (LN) level risks of metastatic cancer for lymph nodes) will impact clinical care because of the poor performance of conventional methods, which involve PET scanning.
  • PLoS ONE 8(10):e78552 the entirety of which is incorporated by reference in its entirety).
  • the sensitivity, specificity, PPV and NPV for PET for detecting hilar and mediastinal lymph node metastases was 74.2%, 54.4%, 86.8% and 73.5% with a false negative rate of 13.2% and false positive rate of 45.5%. Therefore, the methods described herein represent non-invasive techniques for detecting metastatic disease that has resulted in lymph node involvement, which will improve decision making and clinical care. Examples of the impact this device will have on decision making and clinical care include:

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Public Health (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Epidemiology (AREA)
  • Biomedical Technology (AREA)
  • Primary Health Care (AREA)
  • Data Mining & Analysis (AREA)
  • Multimedia (AREA)
  • Radiology & Medical Imaging (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • Pathology (AREA)
  • Quality & Reliability (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Biophysics (AREA)
  • Apparatus For Radiation Diagnosis (AREA)
  • Image Analysis (AREA)
  • Measuring Or Testing Involving Enzymes Or Micro-Organisms (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)
  • Steroid Compounds (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)
  • Nuclear Medicine (AREA)
  • Ultra Sonic Daignosis Equipment (AREA)
PCT/IB2021/061125 2020-12-01 2021-11-30 Machine learning to predict cancer involvement of lymph nodes Ceased WO2022118190A1 (en)

Priority Applications (8)

Application Number Priority Date Filing Date Title
CN202180092579.7A CN116848548A (zh) 2020-12-01 2021-11-30 预测淋巴结的癌累及的机器学习
AU2021390184A AU2021390184A1 (en) 2020-12-01 2021-11-30 Machine learning to predict cancer involvement of lymph nodes
JP2023533338A JP2023551335A (ja) 2020-12-01 2021-11-30 リンパ節のがん侵襲を予測するための機械学習
MX2023006446A MX2023006446A (es) 2020-12-01 2021-11-30 Aprendizaje automático para predecir afectación de cáncer en ganglios linfáticos.
EP21820328.9A EP4256511A1 (en) 2020-12-01 2021-11-30 Machine learning to predict cancer involvement of lymph nodes
US18/255,432 US12608805B2 (en) 2020-12-01 2021-11-30 Machine learning to predict lung cancer involvement of lymph nodes
CA3203664A CA3203664A1 (en) 2020-12-01 2021-11-30 Machine learning to predict cancer involvement of lymph nodes
KR1020237022202A KR20230117391A (ko) 2020-12-01 2021-11-30 림프절의 암 침범을 예측하기 위한 기계 학습

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US202063120102P 2020-12-01 2020-12-01
US63/120,102 2020-12-01

Publications (1)

Publication Number Publication Date
WO2022118190A1 true WO2022118190A1 (en) 2022-06-09

Family

ID=78822446

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/IB2021/061125 Ceased WO2022118190A1 (en) 2020-12-01 2021-11-30 Machine learning to predict cancer involvement of lymph nodes

Country Status (10)

Country Link
US (1) US12608805B2 (https=)
EP (1) EP4256511A1 (https=)
JP (1) JP2023551335A (https=)
KR (1) KR20230117391A (https=)
CN (1) CN116848548A (https=)
AU (1) AU2021390184A1 (https=)
CA (1) CA3203664A1 (https=)
MX (1) MX2023006446A (https=)
TW (1) TW202238515A (https=)
WO (1) WO2022118190A1 (https=)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US12333729B2 (en) * 2022-08-11 2025-06-17 Siemens Medical Solutions Usa, Inc. Automatic staging of non-small cell lung cancer from medical imaging and biopsy reports
KR20250138867A (ko) * 2024-03-13 2025-09-23 연세대학교 산학협력단 인공지능 모델을 기반으로 조기위암에서 절제 전 림프절 전이 위험도를 예측하기 위한 진단 보조 장치 및 방법
CN118334440B (zh) * 2024-04-26 2025-09-26 北京安德医智科技有限公司 一种基于ct影像进行m期分类预测的处理方法和装置
CN119517426A (zh) * 2024-11-19 2025-02-25 中南大学湘雅二医院 一种乳腺癌腋窝淋巴结转移预测半监督模型的构建方法

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190254611A1 (en) * 2018-02-21 2019-08-22 Case Western Reserve University Predicting disease recurrence following trimodality therapy in non-small cell lung cancer using computed tomography derived radiomic features and clinico-patholigic features

Family Cites Families (24)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1356421B1 (en) * 2001-01-23 2009-12-02 Health Discovery Corporation Computer-aided image analysis
KR101038729B1 (ko) * 2008-08-08 2011-06-03 울산대학교 산학협력단 확산강조영상 기반의 임파절 추출 방법
US8078554B2 (en) * 2008-09-03 2011-12-13 Siemens Medical Solutions Usa, Inc. Knowledge-based interpretable predictive model for survival analysis
US8391579B2 (en) * 2010-03-11 2013-03-05 Siemens Corporation Method and system for automatic detection and segmentation of axillary lymph nodes
WO2016191567A1 (en) * 2015-05-26 2016-12-01 Memorial Sloan-Kettering Cancer Center System, method and computer-accessible medium for texture analysis of hepatopancreatobiliary diseases
EP3629898A4 (en) * 2017-05-30 2021-01-20 Arterys Inc. AUTOMATED LESION DETECTION, SEGMENTATION AND LENGTH IDENTIFICATION
CA3069612A1 (en) * 2017-07-13 2019-01-17 Institut Gustave-Roussy A radiomics-based imaging tool to monitor tumor-lymphocyte infiltration and outcome in cancer patients treated by anti-pd-1/pd-l1
WO2020014477A1 (en) * 2018-07-11 2020-01-16 The University Of North Carolina At Chapel Hill Methods, systems, and computer readable media for image analysis with deep learning to predict breast cancer classes
US12369842B2 (en) * 2018-08-31 2025-07-29 Seno Medical Instruments, Inc. Optoacoustic feature score correlation to ipsilateral axillary lymph node status
US11514571B2 (en) 2018-12-17 2022-11-29 Siemens Healthcare Gmbh Hierarchical analysis of medical images for identifying and assessing lymph nodes
EP4597424A3 (en) * 2019-01-07 2025-10-29 Exini Diagnostics AB Systems and methods for platform agnostic whole body image segmentation
CN109785310B (zh) * 2019-01-11 2023-06-06 南京信息工程大学 一种基于乳腺淋巴结全景图像计算的自动分期系统
JP7381590B2 (ja) * 2019-02-04 2023-11-15 マサチューセッツ インスティテュート オブ テクノロジー リンパ節及びリンパ管イメージングのためのシステム及び方法
KR102258756B1 (ko) * 2019-04-12 2021-05-28 계명대학교 산학협력단 의료 영상을 이용한 암의 병기 결정 방법 및 의료 영상 분석 장치
CN111340128A (zh) * 2020-03-05 2020-06-26 上海市肺科医院（上海市职业病防治院） 一种肺癌转移性淋巴结病理图像识别系统及方法
CN111445946B (zh) * 2020-03-26 2021-07-30 山东省肿瘤防治研究院（山东省肿瘤医院） 一种利用pet/ct图像推算肺癌基因分型的演算方法
CN111539918B (zh) * 2020-04-15 2023-05-02 复旦大学附属肿瘤医院 基于深度学习的磨玻璃肺结节风险分层预测系统
GB202007256D0 (en) * 2020-05-15 2020-07-01 Univ Oxford Innovation Ltd Functional imaging features from computed tomography images
CN111584046B (zh) * 2020-05-15 2023-10-27 周凌霄 一种医学影像数据ai处理方法
JP7702477B2 (ja) * 2020-07-24 2025-07-03 オンク エーアイ インコーポレイテッド 画像データ及び臨床データの深層学習解析を用いた免疫療法治療に対する反応予測
AU2020101581A4 (en) * 2020-07-31 2020-09-17 Ampavathi, Anusha MS Lymph node metastases detection from ct images using deep learning
CN111862085A (zh) * 2020-08-03 2020-10-30 徐州市肿瘤医院 一种周围型nsclc的隐匿性n2淋巴结转移的预测方法及系统
CN112530592A (zh) * 2020-12-14 2021-03-19 青岛大学 一种基于机器学习的非小细胞肺癌风险预测方法
US12094107B2 (en) * 2021-04-07 2024-09-17 Optellum Limited CAD device and method for analyzing medical images

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190254611A1 (en) * 2018-02-21 2019-08-22 Case Western Reserve University Predicting disease recurrence following trimodality therapy in non-small cell lung cancer using computed tomography derived radiomic features and clinico-patholigic features

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
WANG XIANG ET AL: "Can peritumoral radiomics increase the efficiency of the prediction for lymph node metastasis in clinical stage T1 lung adenocarcinoma on CT?", EUROPEAN RADIOLOGY, SPRINGER BERLIN HEIDELBERG, BERLIN/HEIDELBERG, vol. 29, no. 11, 18 March 2019 (2019-03-18), pages 6049 - 6058, XP036906601, ISSN: 0938-7994, [retrieved on 20190318], DOI: 10.1007/S00330-019-06084-0 *

Also Published As

Publication number Publication date
US12608805B2 (en) 2026-04-21
JP2023551335A (ja) 2023-12-07
EP4256511A1 (en) 2023-10-11
AU2021390184A9 (en) 2024-07-11
MX2023006446A (es) 2023-08-11
CA3203664A1 (en) 2022-06-09
US20240005502A1 (en) 2024-01-04
KR20230117391A (ko) 2023-08-08
TW202238515A (zh) 2022-10-01
CN116848548A (zh) 2023-10-03
AU2021390184A1 (en) 2023-07-20

Similar Documents

Publication Publication Date Title
US11869187B2 (en) System and method for predicting the risk of future lung cancer
US12608805B2 (en) Machine learning to predict lung cancer involvement of lymph nodes
US12293832B2 (en) System and method for predicting the risk of future lung cancer
Goyal et al. Application of artificial intelligence in pancreaticobiliary diseases
Amparore et al. Computer vision and machine-learning techniques for automatic 3D virtual images overlapping during augmented reality guided robotic partial nephrectomy
Chen et al. AI-based chest CT semantic segmentation algorithm enables semi-automated lung cancer surgery planning by recognizing anatomical variants of pulmonary vessels
Wang et al. Development and validation of artificial intelligence models for preoperative prediction of inferior mesenteric artery lymph nodes metastasis in left colon and rectal cancer
US20240379237A1 (en) Multimodal System and Method for Predicting Cancer
Thompson et al. Deep learning-based model for identifying tumors in endoscopic images from patients with locally advanced rectal cancer treated with total neoadjuvant therapy
Ma et al. An integrated model combined intra-and peritumoral regions for predicting chemoradiation response of non small cell lung cancers based on radiomics and deep learning
Chhikara et al. Deep learning techniques for pancreatic cancer analysis: A systematic review and implantation prerequisites
Rajalaxmi et al. A systematic review of lung cancer prediction using machine learning algorithm
WO2024100632A1 (en) Systems and methods for prioritizing medical resources for cancer screening
EA048045B1 (ru) Машинное обучение для прогнозирования ракового поражения лимфатических узлов
Sandhu et al. Improving breast cancer detection with deep learning techniques: a study using CBIS-DDMS dataset
Wang et al. Mut-Detecter: An EGFR activating mutation type classification method with a deep convolutional neural network
Rao et al. Prostate Cancer Detection Using Integrated Multi Modal Approaches
Huang et al. Exempting axillary staging surgery in breast cancer using multimodal ultrasound imaging and radiomics of sentinel lymph nodes
Kirtane et al. The role of artificial intelligence in surveillance in Barrett’s esophagus and gastric intestinal metaplasia
Wu et al. Deep learning-based prediction of enhanced CT scans for lymph node metastasis in esophageal squamous cell carcinoma
Wang et al. Preoperative prediction of lymph node metastasis in adenocarcinoma of esophagogastric junction using CT texture analysis combined with machine learning
Mareddy et al. An overview of imaging in pancreatic cancer
Holt Navigating a New Path to Test Peripheral Lung Nodules
Pearce et al. BLUE LIGHT CYSTOSCOPY IMPROVES DETECTION RATES FOR UROTHELIAL BLADDER CANCER COMPARED TO WHITE LIGHT CYSTOSCOPY: UPDATED RESULTS FROM A PROSPECTIVE MULTICENTER REGISTRY: MP08-17

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21820328

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 3203664

Country of ref document: CA

WWE Wipo information: entry into national phase

Ref document number: 2023533338

Country of ref document: JP

REG Reference to national code

Ref country code: BR

Ref legal event code: B01A

Ref document number: 112023010697

Country of ref document: BR

ENP Entry into the national phase

Ref document number: 20237022202

Country of ref document: KR

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2021820328

Country of ref document: EP

Effective date: 20230703

ENP Entry into the national phase

Ref document number: 2021390184

Country of ref document: AU

Date of ref document: 20211130

Kind code of ref document: A

ENP Entry into the national phase

Ref document number: 112023010697

Country of ref document: BR

Kind code of ref document: A2

Effective date: 20230531

WWE Wipo information: entry into national phase

Ref document number: 202180092579.7

Country of ref document: CN