google ads

Learning Modules ❯ Deep Learning and AI

Application of Radiomics in Pancreatic Imaging – Current Status and Future Directions

 

 

Application of Radiomics in Pancreatic Imaging - Current Status and Future Directions

Linda C. Chu

The Russel H. Morgan Department of Radiology and Radiological Science, The Department of Pathology, The Department of Cancer Research, and the Department of Computer Science, Johns Hopkins University, Baltimore

 

Disclosure

  • Research support from The Lustgarten Foundation:
    • Linda C. Chu
    • Seyoun Park
    • Satomi Kawamoto
    • Daniel F. Fouladi
    • Shahab Shayesteh
    • Alan L. Yuille
    • Bert Vogelstein
    • Elliot K. Fishman

 

Learning Objectives

  • Review basic principles of radiomics
  • Review current applications of radiomics in pancreatic imaging:
    • Detection
    • Classification
    • Prognostication
  • Discuss technical challenges and future directions

 

Introduction

  • Tumors are spatially heterogeneous structures and this heterogeneity can be quantified on imaging data
  • Radiomics converts this imaging data into high dimensional mineable feature space
  • Radiomics has the potential of providing imaging biomarkers that can be useful in tumor detection, diagnosis, and prediction of treatment response
Kumar V et al. Magn Reson Imaging. 2012;30(9):4-48. Aerts HJ et al. Nat Commun. 2014;5:4006.
Gillies RJ et al. Radiology. 2016;278(2):563-577.

 

Radiomics

  • Radiomics features can be classified into first-order, shape, second-order, and high-order statistical outputs:
  • First-order statistics:
    • Distribution of individual voxel values
    • Histogram-based methods
    • Mean, median, maximum, minimum
    • Uniformity, entropy, skewness, kurtosis, etc.
Radiomics
First-Order Statistics

Kumar V et al. Magn Reson Imaging. 2012;30(9):4-48. Aerts HJ et al. Nat Commun. 2014;5:4006.
Gillies RJ et al. Radiology. 2016;278(2):563-577.

 

Radiomics

  • Shape can be quantified through a combination of shape features such as:
    • Compactness
    • Max 3D diameter
    • Spherical disproportion
    • Sphericity
    • Surface area
    • Surface/volume ratio
    • Volume
Radiomics
Extracted Shape Representations

Kumar V et al. Magn Reson Imaging. 2012;30(9):4-48. Aerts HJ et al. Nat Commun. 2014;5:4006.
Gillies RJ et al. Radiology. 2016;278(2):563-577.

 

Radiomics

  • Second-order statistics (texture features):
    • Statistical interrelationships between voxels of similar contrast values
    • Gray-level co-occurrence matrix (GLCM)
    • Gray-level run length matrix (GLRM)
Radiomics

Kumar V et al. Magn Reson Imaging. 2012;30(9):4-48. Aerts HJ et al. Nat Commun. 2014;5:4006.
Gillies RJ et al. Radiology. 2016;278(2):563-577.

 

Radiomics

  • Higher-order statistics:
    • Impose filter grids to extract repetitive or nonrepetitive patterns
    • Wavelet: Passes image through low pass or high pass filters in x, y, and z direction
    • Laplacian of Gaussian:
      • Gaussian filter smooths the image
      • Laplacian filter detects the edge
Radiomics Radiomics

Kumar V et al. Magn Reson Imaging. 2012;30(9):4-48. Aerts HJ et al. Nat Commun. 2014;5:4006.
Gillies RJ et al. Radiology. 2016;278(2):563-577.

 

Radiomics – Pancreatic Applications

  • Diagnostic and prognostic value of radiomics features have been evaluated in a number of head and neck, thoracic, GI, and GU malignancies
  • This exhibit reviews recent applications of radiomics in pancreatic imaging:
    • Detection
    • Classification
    • Prognostication
Lubner MG et al. Radiographics. 2017;37(5):1483-1503.

 

Detection of PDAC

  • Pancreatic ductal adenocarcinoma (PDAC) has dismal prognosis as most patients are diagnosed at advanced stage of disease
  • Early CT findings of PDAC can be subtle and can be missed by even experienced radiologists
  • Radiomics features extracted from the whole 3D volume of the pancreas has the potential for computer aided detection of PDAC

 

Detection of PDAC

  • Utility of radiomics features to differentiate pancreatic ductal adenocarcinoma vs. normal:
    • 190 patients with PDAC vs. 190 normal controls
    • Whole gland segmentation of pancreas (including tumor region) from preop CT
Detection of PDAC Detection of PDAC

 

Detection of PDAC

  • 40 radiomics features were selected for random forest classifier
  • Validation dataset (n = 125)
    • 60 PDAC + 65 normal controls
  • Overall accuracy: 99.2% (124/125)
  • Sensitivity: 100% (60/60)
  • Specificity: 98.5% (64/65)
Radiomics features can achieve high accuracy in detection of PDAC

 

Detection of PDAC

  • Top 5 maximally relevant features include:
    • Texture – Heterogeneous texture
    • Shape – Less spherical disproportion
    • Wavelets – Textural pattern change at the border of the tumor or dilated pancreatic duct
Detection of PDAC

 

Classification

  • Radiomics features have been applied in the classification of various types of pancreatic pathology:
    • Autoimmune pancreatitis (AIP) vs. PDAC
    • Pancreatic neuroendocrine (PNET) vs. PDAC
    • PNET vs. Intrapancreatic accessory spleen (IPAS)
    • Pancreatic cystic neoplasms
    • IPMNs with high grade vs. low grade dysplasia
    • Pancreatic neuroendocrine tumor (PNET) grades

 

Differentiation of Autoimmune Pancreatitis from PDAC

  • Autoimmune pancreatitis (AIP) shares overlapping clinical and imaging features as PDAC
  • Importantly, treatment and prognosis for these two conditions vary dramatically
Differentiation of Autoimmune Pancreatitis from PDAC

 

Differentiation of Autoimmune Pancreatitis from PDAC

  • Retrospective matched case-control study:
    • 32 patients with AIP
    • 40 patients with PDAC
    • Segmented 3D volume of involved and uninvolved segments of pancreas
    • Compared diagnostic performance of radiomics features vs. diagnosis provided on the CT reports

 

Differentiation of Autoimmune Pancreatitis from PDAC

  • AIP was suspected or included in the differential diagnosis of the CT reports in 47% of cases
  • Radiomics analysis of AIP vs. PDAC:
    • Accuracy 94.4%
    • Sensitivity 95.0%
    • Specificity 93.8%
Differentiation of Autoimmune Pancreatitis from PDAC

Radiomics features can differentiate AIP from PDAC, which has important treatment implications

 

Differentiation of PNET vs. PDAC vs. IPAS

  • A minority of pancreatic neuroendocrine tumors (PNETs) can be relatively hypovascular, and can be difficult to differentiate from PDAC
  • Enhancement pattern of PNETs can also mimic intrapancreatic accessory spleen (IPAS)
  • Radiomics features have the potential to aid their differentiation
Differentiation of PNET vs. PDAC vs. IPAS

Li J et al. Cancer Med. 2018. [Epub ahead of print]. Lin X et al. Acta Radiol. 2018. [Epub ahead of print]

 

Differentiation of PNET vs. PDAC vs. IPAS

Differentiation of PNET vs. PDAC vs. IPAS

Radiomics features can aid in differentiation of pancreatic neuroendocrine tumors from mimickers such as pancreatic ductal adenocarcinoma and intrapancreatic accessory spleen

Li J et al. Cancer Med. 2018. [Epub ahead of print]. Lin X et al. Acta Radiol. 2018. [Epub ahead of print]

 

Classification of Pancreatic Cysts

  • Pancreatic cystic lesions are identified incidentally in approximately 2% of abdominal CTs
  • Malignant potential of pancreatic cystic lesions vary based on underlying pathologic diagnosis
  • These cystic lesions can have overlapping imaging features and can be difficult to differentiate clinically
Laffan TA et al. AJR. 2008;191:802-807. Zanini N et al. Pancreatology. 2015;15:417-422.

 

Classification of Pancreatic Cysts

  • Utility of radiomics in classification of pancreatic cystic neoplasms:
    • Retrospective study of 214 pancreatic cysts:
      • 64 IPMNs
      • 33 Mucinous cystic neoplasms
      • 60 Serous cystadenomas
      • 24 Solid pseudopapillary neoplasms
      • 33 Cystic pancreatic neuroendocrine tumors
    • Compared diagnostic performance of radiomics and random-forest classifier vs. a radiologist

 

Classification of Pancreatic Cysts

  • Academic radiologist with >20 years experience:
    • Accuracy = 77.1%
  • Radiomics features with random-forest:
    • Accuracy = 82%
Classification of Pancreatic Cysts

Radiomics features can achieve improved accuracy in classification of pancreatic cystic neoplasms compared to an experienced radiologist

 

Differentiation of IPMNs with High-Grade vs. Low-Grade Dysplasia

Classification of Pancreatic Cysts

Hanania AN et al. Oncotarget. 2016;7(52):85776-85784. Permuth JB et al. Oncotarget. 2016;7(52):85785-85797. Attiyeh MA et al. HPB (Oxford). 2018 [Epub ahead of print]. Chakraborty J et al. Med Phys. 2018 [Epub ahead of print]

 

Differentiation of IPMNs with High-Grade vs. Low-Grade Dysplasia

Classification of Pancreatic Cysts

Radiomics features can aid in differentiation of IPMNs with high-grade vs. low-grade dysplasia

Hanania AN et al. Oncotarget. 2016;7(52):85776-85784. Permuth JB et al. Oncotarget. 2016;7(52):85785-85797. Attiyeh MA et al. HPB (Oxford). 2018 [Epub ahead of print]. Chakraborty J et al. Med Phys. 2018 [Epub ahead of print]

 

Differentiation of Pancreatic Neuroendocrine Tumor Grades

Differentiation of Pancreatic Neuroendocrine Tumor Grades

G= Grade

Choi TW et al. Acta Radiol. 2018;59(4):383-392. Canellas R et al. AJR. 2018;210(2):341-346. Guo C et al. Abdom Radiol (NY). 2018 [Epub ahead of print]. Chakraborty J et al. SPIE Medical Imaging, Houston, TX. 2018.

 

Differentiation of Pancreatic Neuroendocrine Tumor Grades

Differentiation of Pancreatic Neuroendocrine Tumor Grades

G= Grade

Radiomics features can aid in the differentiation of high-grade vs. low-grade pancreatic neuroendocrine tumors

Choi TW et al. Acta Radiol. 2018;59(4):383-392. Canellas R et al. AJR. 2018;210(2):341-346. Guo C et al. Abdom Radiol (NY). 2018 [Epub ahead of print]. Chakraborty J et al. SPIE Medical Imaging, Houston, TX. 2018.

 

Prognostication – Patient Survival

  • Surgical resection is the only curative treatment for patients with PDAC
  • Pancreatic resection is a major surgery with its associated morbidities
  • Improving patient selection is critical in identifying patients most likely to benefit from surgical resection
  • There are a number of recent publications evaluating the utility of PDAC radiomics from pre-treatment CT in predicting patient survival

 

Prognostication – Patient Survival

Prognostication – Patient Survival

Cassinotto C et al. Eur J Radiol. 2017;90:152-158. Eilaghi A et al. BMC Medical Imaging. 2017;17(1):38. Chakraborty J et al. PLoS One. 2017;12(12):e0188022. Yun G et al. Scientific Reports. 2018;8:7226. Attiyeh MA et al. Ann Surg Oncol. 2018;25:1034-1042. Sandrasegaran K et al. Eur Radiol. 2018 [Epub ahead of print]

 

Tumor hypoattenuation and tumor heterogeneity are associated with poor survival in patients with PDAC

Tumor hypoattenuation and tumor heterogeneity

Cassinotto C et al. Eur J Radiol. 2017;90:152-158. Eilaghi A et al. BMC Medical Imaging. 2017;17(1):38. Chakraborty J et al. PLoS One. 2017;12(12):e0188022. Yun G et al. Scientific Reports. 2018;8:7226. Attiyeh MA et al. Ann Surg Oncol. 2018;25:1034-1042. Sandrasegaran K et al. Eur Radiol. 2018 [Epub ahead of print]

 

Prognostication – Treatment Response

  • Patients with PDAC experience different treatment response to neoadjuvant chemoradiation
  • The ability to predict which patients will respond to neoadjuvant chemoradiation will be helpful in selecting the best treatment regimen for the individual patient

 

Prognostication – Treatment Response

  • Chen et al. explored changes in PDAC radiomics features during neoadjuvant chemoradiation:
    • 20 PDAC patients with pancreatic head tumors
    • Analyzed CT obtained during radiation therapy planning
    • 8 histogram-based radiomics metrics
    • Significant changes in CT radiomics features were present during neoadjuvant chemoradiation
Chan X et al. PLoS One. 2017;12(6):e0178961.

 

Current Challenges in Radiomics

  • Currently, there is no standardization in image acquisition and post-processing protocol
  • Radiomics features can be potentially affected by these technical parameters:
Current Challenges in Radiomics

Kumar V et al. Magn Reson Imaging. 2012;30(9):4-48.

 

Future Directions of Radiomics

  • Standardization of radiomics analysis to facilitate comparison of results from different studies
  • Large quality datasets with enough statistical power for validation of radiomics analysis
  • Multicenter trials on CT scans obtained from different vendors to determine generalizability of the results
  • Integration into clinical workflow

 

Conclusion

  • Radiomics has shown tremendous promise in pancreatic imaging with applications in disease detection, lesion characterization, and patient prognostication
  • It has potential in tailoring individual patient treatment in the era of precision medicine
  • A few technical challenges need to be overcome in the near future to facilitate its integration into our clinical workflow

 

References

Aerts HJ et al. Nat Commun. 2014;5:4006.
Attiyeh MA et al. Ann Surg Oncol. 2018;25:1034-1042.
Attiyeh MA et al. HPB (Oxford). 2018 [Epub ahead of print]
Canellas R et al. AJR. 2018;210(2):341-346.
Cassinotto C et al. Eur J Radiol. 2017;90:152-158.
Chakraborty J et al. PLoS One. 2017;12(12):e0188022.
Chakraborty J et al. Med Phys. 2018 [Epub ahead of print]
Chakraborty J et al. SPIE Medical Imaging, Houston, TX. 2018.
Chan X et al. PLoS One. 2017;12(6):e0178961.
Choi TW et al. Acta Radiol. 2018;59(4):383-392.
Eilaghi A et al. BMC Medical Imaging. 2017;17(1):38.
Gillies RJ et al. Radiology. 2016;278(2):563-577.
Guo C et al. Abdom Radiol (NY). 2018 [Epub ahead of print]
Hanania AN et al. Oncotarget. 2016;7(52):85776-85784.
Kumar V et al. Magn Reson Imaging. 2012;30(9):4-48.
Laffan TA et al. AJR. 2008;191:802-807.
Li J et al. Cancer Med. 2018. [Epub ahead of print]
Lin X et al. Acta Radiol. 2018. [Epub ahead of print]
Lubner MG et al. Radiographics. 2017;37(5):1483-1503.
Permuth JB et al. Oncotarget. 2016;7(52):85785-85797.
Sandrasegaran K et al. Eur Radiol. 2018 [Epub ahead of print]
Yun G et al. Scientific Reports. 2018;8:7226.
Zanini N et al. Pancreatology. 2015;15:417-422.
 
Acknowledgements
  • Linda C. Chu
  • Seyoun Park
  • Satomi Kawamoto
  • Daniel F. Fouladi
  • Shahab Shayesteh
  • Karen M. Horton
  • Alan L. Yuille
  • Ralph H. Hruban
  • Bert Vogelstein
  • Kenneth W. Kinzler
  • Elliot K. Fishman

Privacy Policy

Copyright © 2024 The Johns Hopkins University, The Johns Hopkins Hospital, and The Johns Hopkins Health System Corporation. All rights reserved.