google ads

AI in Radiology: Current Status and What You Need to Know

AI in Radiology: Current Status and What You Need to Know

Elliot K. Fishman M.D.
Johns Hopkins Hospital

Click here to view this module as a video lecture.

 

Questions

  • Do you believe AI will significantly change your practice within the next 2-5 years?
  • Do you believe AI will benefit or hurt Radiology?
  • Do you believe you are prepared for the impact of AI in your practice?

 

“It’s hard to predict the future, and what immensely complicates predictions over seemingly promising technologies like gene therapy or AI is how their complex construction will interface with other equally complex and dynamic technologies, all of which operate in an environment of unceasing economic and institutional flux. It remains anyone’s guess as to how AI applications will be affected by their integration with PACS, how liability trends or regulatory efforts will affect AI, whether reimbursement for AI will justify its use, how mergers and acquisitions will affect AI implementation, and how well AI models will accommodate ethical requirements related to informed consent, privacy, and patient access.”
AI Hype and Radiology: A Plea for Realism and Accuracy
Banja J et al.
Radiology: Artificial Intelligence 2020; 2(4):e190223

 

AI in Radiology: Current Status

 

AI in Radiology: Current Status

 

AI in Radiology: Current Status

 

“In an independent study of six radiologists, the AI system outperformed all of the human readers: the area under the receiver operating characteristic curve (AUC-ROC) for the AI system was greater than the AUC-ROC for the average radiologist by an absolute margin of 11.5%. We ran a simulation in which the AI system participated in the double-reading process that is used in the UK, and found that the AI system maintained non-inferior performance and reduced the workload of the second reader by 88%. This robust assessment of the AI system paves the way for clinical trials to improve the accuracy and efficiency of breast cancer screening.”
International evaluation of an AI system for breast cancer screening
Scott Mayer McKinney et al
Nature Vol577 2January2020

 

• Radiologists improved their cancer detection accuracy in digital breast tomosynthesis (DBT) when using an AI system for support, while simultaneously reducing reading time.
• The stand-alone breast cancer detection performance of an AI system is non-inferior to the average performance of radiologists for reading digital breast tomosynthesis exams.
• The use of an AI support system could make advanced and more reliable imaging techniques more accessible and could allow for more cost-effective breast screening programs with DBT.
Impact of artificial intelligence support on accuracy and reading time in breast tomosynthesis image interpretation: a multi-reader multi-case study.
van Winkel SL et al
Eur Radiol. 2021 Nov;31(11):8682-8691.

 

Godfrey Hounsfield in 1971

Godfrey Hounsfield in 1971

 

How do we read a CT scan today?

1980
  • 4 images on 8 x 10 film
  • 30-40 scan slices per case
  • Acquisition time per study was 40-50 minutes (10 sec scan slides and 60 sec per slice reconstruction time)
  • Limited resolution studies

 

How do we read a CT scan today?

2022
  • Images reviewed on a computer (no film)
  • 2000-4000 scan slices per case
  • Acquisition time per study was 10 seconds or less with real time reconstruction (50 images /sec)
  • High resolution studies

 

Abdominal Pain for 2 Years

Abdominal Pain for 2 Years

 

AI in Radiology: Current Status

 

Where’s Waldo is what I do everyday?

Where’s Waldo is what I do everyday?

 

“ In the daily radiology practice, the rate of interpretation error is between 3% and 4%; however, of the radiology studies that contain abnormalities, the error rate is even higher, averaging in the 30% range.”
Fool Me Twice: Delayed Diagnoses in Radiology With Emphasis on Perpetuated Errors
Kim YW, Mansfield LT
AJR 2014;202:465-470

 

“ In our study, the majority of errors made were errors of underreading (42%), where the finding was simply missed.”
Fool Me Twice: Delayed Diagnoses in Radiology With Emphasis on Perpetuated Errors
Kim YW, Mansfield LT
AJR 2014;202:465-470

 

“Missed findings rather than misinterpretations of detected abnormalities were the most common reason for abdominopelvic CT report addenda. Awareness of the most common misses by anatomic location may help guide quality assurance initiatives. A wide variety of contributing factors were identified. Informatics and workflow optimization may be warranted to facilitate radiologists’ access to all available patient-related data, as well as communication with other physicians, and thereby help reduce diagnostic errors.”
Diagnostic errors in abdominopelvic CT interpretation: characterization based on report addenda
Andrew B. Rosenkrantz, Neil K. Bansal
Abdom Radiol (2016) 41:1793–1799

 

“The purpose of the study was to determine if increasing radiologist reading speed results in more misses and interpretation errors.”
The Effect of Faster Reporting Speed for Imaging Studies on the Number of Misses and Interpretation Errors: A Pilot Study.
Sokolovskaya E et al.
J Am Coll Radiol. 2015 Jul;12(7):683-8.

 

“ Reading at the faster speed resulted in more major misses for 4 of the 5 radiologists. The total number of major misses for the 5 radiologists, when they reported at the faster speed, was 16 of 60 reported cases, versus 6 of 60 reported cases at normal speed; P = .032. The average interpretation error rate of major misses among the 5 radiologists reporting at the faster speed was 26.6%, compared with 10% at normal speed.”
The Effect of Faster Reporting Speed for Imaging Studies on the Number of Misses and Interpretation Errors: A Pilot Study.
Sokolovskaya E et al.
J Am Coll Radiol. 2015 Jul;12(7):683-8.

 

CT of the Pancreas: Scan Analysis

  • Axial CT
  • Multiplanar Reconstruction (MPR)
  • 3D Imaging (Volume Rendering (VRT and MIP)
  • Cinematic Rendering (CR)
  • Radiomics
  • Deep Learning (DL)

 

PNET Tail of Pancreas 1cm

PNET Tail of Pancreas 1cm

 

Neuroendocrine Tumor TOP With Calcifications

Neuroendocrine Tumor TOP With Calcifications

 

Adenocarcinoma Pancreas

Adenocarcinoma Pancreas

 

Texture Changes in the Gland

Texture Changes in the Gland

 

AI in Radiology: Current Status

 

HoloLens 2 from Microsoft

HoloLens 2 from Microsoft

 

AI in Radiology: Current Status

 

AI in Radiology: Current Status

 

”Radiomics refers to the extraction of mineable high-dimensional data from radiologic images and has been applied within oncology to improve diagnosis and prognostication with the aim of delivering precision medicine. The premise is that imaging data convey meaningful information about tumor biology, behavior, and pathophysiology and may reveal information that is not otherwise apparent to current radiologic and clinical interpretation.”
Radiomics in Oncology: A Practical Guide
Joshua D. Shur, et al.
RadioGraphics 2021; 41:1717–1732

 

Radiomics in Oncology: A Practical Guide
Joshua D. Shur, et al.
RadioGraphics 2021; 41:1717–1732

AI in Radiology: Current Status

 

Radiomics in Oncology: A Practical Guide
Joshua D. Shur, et al.
RadioGraphics 2021; 41:1717–1732

AI in Radiology: Current Status

 

Radiomics

  • Tumors are spatially heterogeneous structures
  • Heterogeneity can be quantified on imaging data
  • Radiomics converts this imaging data into high dimensional mineable feature space
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

  • 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.
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

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. Lubner MG et al. RadioGraphics. 2017;37:1483-1503.

Radiomics

 

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
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

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)
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

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 and Gaussian:
    • Gaussian filter smooths the image
    • Laplacian filter detects the edge
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

 

Detection of PDAC

Extracted 478 radiomics features from CTs of whole 3D pancreas volume to differentiate 190 PDAC vs. 190 normal controls

Chu LC et al. AJR

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

 

“Our results showed that, after manual segmentation of pancreas boundaries, radiomics features and the random forest classifier were highly accurate in differentiating PDAC cases from normal control cases (sensitivity, 100%; specificity, 98.5%; accuracy, 99.2%). The radiomics features most relevant to differentiate PDAC from normal pancreas were based on shape and textural heterogeneity.”
Utility of CT Radiomics Features in Differentiation of Pancreatic Ductal Adenocarcinoma From Normal Pancreatic Tissue
Chu LC, Park S, Kawamoto S, Fouladi DF, Shayesteh S, Zinreich ES, Graves JS, Horton KM, Hruban RH, Yuille AL, Kinzler KW, Vogelstein B, Fishman EK.
AJR Am J Roentgenol. 2019 Apr 23:1-9.

 

“CT features of early PDAC can be subtle and missed by even experienced radiologists. Early signs of PDAC such as pancreatic parenchyma inhomogeneity and loss of normal fatty marbling of the pancreas have been described on retrospective CT review up to 34 months before the diagnosis of PDAC. Quantitative analysis of these imaging features offers the potential for computer-aided diagnosis of PDAC.”
Utility of CT Radiomics Features in Differentiation of Pancreatic Ductal Adenocarcinoma From Normal Pancreatic Tissue
Chu LC, Park S, Kawamoto S, Fouladi DF, Shayesteh S, Zinreich ES, Graves JS, Horton KM, Hruban RH, Yuille AL, Kinzler KW, Vogelstein B, Fishman EK.
AJR Am J Roentgenol. 2019 Apr 23:1-9.

 

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