CAD Performance Analysis for CAD Performance Analysis for Pulmonary Nodule Detection: Pulmonary Nodule Detection: Comparison of Thick- and Thin- Comparison of Thick- and Thin- Slice Multi-detector CT Scans Slice Multi-detector CT Scans Randy D Ernst Randy D Ernst 1 , Russell C Hardie , Russell C Hardie 2 , , Metin N Gurcan Metin N Gurcan 3 , Aytekin Oto , Aytekin Oto 1 , , Steve K Rogers Steve K Rogers 3 , Jeffrey W Hoffmeister , Jeffrey W Hoffmeister 3 1. Department of Radiology, The University of Texas 1. Department of Radiology, The University of Texas Medical Branch, Galveston TX Medical Branch, Galveston TX 2. iCAD Inc. and University of Dayton, Dayton OH 2. iCAD Inc. and University of Dayton, Dayton OH 3. iCAD Inc., Beavercreek OH 3. iCAD Inc., Beavercreek OH
35
Embed
Randy D Ernst 1 , Russell C Hardie 2 , Metin N Gurcan 3 , Aytekin Oto 1 ,
Randy D Ernst 1 , Russell C Hardie 2 , Metin N Gurcan 3 , Aytekin Oto 1 , Steve K Rogers 3 , Jeffrey W Hoffmeister 3 1. Department of Radiology, The University of Texas Medical Branch, Galveston TX 2. iCAD Inc. and University of Dayton, Dayton OH 3. iCAD Inc., Beavercreek OH. - PowerPoint PPT Presentation
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
CAD Performance Analysis for CAD Performance Analysis for Pulmonary Nodule Detection: Pulmonary Nodule Detection:
Comparison of Thick- and Thin-Slice Comparison of Thick- and Thin-Slice Multi-detector CT ScansMulti-detector CT Scans
Randy D ErnstRandy D Ernst11, Russell C Hardie, Russell C Hardie22, , Metin N GurcanMetin N Gurcan33, Aytekin Oto, Aytekin Oto11,,
Steve K RogersSteve K Rogers33, Jeffrey W Hoffmeister, Jeffrey W Hoffmeister33
1. Department of Radiology, The University of Texas Medical 1. Department of Radiology, The University of Texas Medical Branch, Galveston TX Branch, Galveston TX
2. iCAD Inc. and University of Dayton, Dayton OH 2. iCAD Inc. and University of Dayton, Dayton OH 3. iCAD Inc., Beavercreek OH3. iCAD Inc., Beavercreek OH
PurposePurpose To compare the performance of To compare the performance of
a CAD (a CAD (QuickCue™QuickCue™, Beavercreek, , Beavercreek, OH) system in detecting lung OH) system in detecting lung nodules from nodules from thick- and thin-slicethick- and thin-slice multi-detector row CT.multi-detector row CT.
To evaluate the potential benefit To evaluate the potential benefit of CAD on radiologist sensitivity.of CAD on radiologist sensitivity.
Methods and MaterialsMethods and Materials 57 studies reviewed retrospectively 57 studies reviewed retrospectively Case selection:Case selection:
Obtained during a 5-month periodObtained during a 5-month period Referred from multiple departmentsReferred from multiple departments Contain at least 1 pulmonary nodule Contain at least 1 pulmonary nodule
but fewer than 10 nodules to localizebut fewer than 10 nodules to localize No significant miss-registration, No significant miss-registration,
kVp)kVp) Images reconstructed at 5-mm Images reconstructed at 5-mm
(thick) and 2.5 mm (thin) slice (thick) and 2.5 mm (thin) slice thicknessesthicknesses
Methods and MaterialsMethods and Materials 140 nodules (3 mm - 25 mm) were 140 nodules (3 mm - 25 mm) were
identifiedidentified pre-CAD by radiologists pre-CAD by radiologists from thick-slice cases onlyfrom thick-slice cases only mean nodule size 7.3 mean nodule size 7.3 ±± 4.2 4.2 mm mm
Truth marks were mapped to the thick-Truth marks were mapped to the thick-slice 5mm data.slice 5mm data.
Gold standards for nodule truth created Gold standards for nodule truth created from post-CAD radiologist review from post-CAD radiologist review One gold standard for thick-sliceOne gold standard for thick-slice Separate gold standard for thin-sliceSeparate gold standard for thin-slice
CAD SystemCAD System (QuickCue™(QuickCue™, iCAD Inc., iCAD Inc.))
3D LungSegmentation
3D CandidateSegmentation
CalculateFeatures
DICOMImages
Classifier
DetectionMask
CAD SystemCAD System Candidates segmented by Candidates segmented by
thresholding and morphological thresholding and morphological processingprocessing
2D and 3D features computed for 2D and 3D features computed for each candidateeach candidate
Anatomical information (hilus, Anatomical information (hilus, airways, aorta, etc.) compared to airways, aorta, etc.) compared to reduce false positivesreduce false positives
A classifier applied for final decisionA classifier applied for final decision
CAD detected 72.1% (101/140) of the CAD detected 72.1% (101/140) of the thick gold standard truth nodulesthick gold standard truth nodules
CAD detected 35 additional CAD detected 35 additional radiologist-confirmed nodules, an radiologist-confirmed nodules, an increase of 25% (35/140) in sensitivityincrease of 25% (35/140) in sensitivity
5.6 (317/57) false-positives per case5.6 (317/57) false-positives per case 55 due to atelectasis55 due to atelectasis 18 due to scarring18 due to scarring
Review of Thick-Slice Review of Thick-Slice CAD ResultsCAD Results
Venn Diagram for ThickVenn Diagram for Thick
3
39 35
317
CADPre-CAD Review
Post-CAD ReviewGold Standard
101
0 0
CAD detected 80.7% (113/140) of CAD detected 80.7% (113/140) of the pre-CAD truth nodules.the pre-CAD truth nodules.
CAD detected 94 additional CAD detected 94 additional radiologist-confirmed nodules, an radiologist-confirmed nodules, an increase of 67.1% (94/140).increase of 67.1% (94/140).
4.6 (262/57) false-positives reported 4.6 (262/57) false-positives reported per case.per case. 70 due to atelectasis70 due to atelectasis 39 due to scarring39 due to scarring
Review of Thin-Slice Review of Thin-Slice CAD ResultsCAD Results
Venn Diagram for ThinVenn Diagram for Thin
0
26 94
262
CAD using thin-slicePre-CAD Review using thick-slice with detections mapped to thin-slice
5 primary lung cancers5 primary lung cancers 24 cases of metastatic cancer including 24 cases of metastatic cancer including
7 lymphomas, 4 breast, 4 head and neck, 2 7 lymphomas, 4 breast, 4 head and neck, 2 colon, 2 pancreas, 1 carcinoid, 1 seminoma, 1 colon, 2 pancreas, 1 carcinoid, 1 seminoma, 1 ovarian, 1 melanoma and 1 tracheal ovarian, 1 melanoma and 1 tracheal papillomatosispapillomatosis
23 cases of infection, including23 cases of infection, including 19 granulomatous disease either calcified, 19 granulomatous disease either calcified,
stable on follow-up or biopsy proven. 4 were stable on follow-up or biopsy proven. 4 were presumed infection that resolved with follow-up presumed infection that resolved with follow-up
1 case proved to be a thrombosed AVM1 case proved to be a thrombosed AVM 4 cases lost to follow up4 cases lost to follow up
Case Follow-upCase Follow-up
Example Example TPsTPs
Examples of Examples of nodules that are nodules that are detected by both detected by both radiologist and radiologist and CADCAD
Example TPsExample TPs Examples of nodules that are initially Examples of nodules that are initially
missed by radiologists then detected missed by radiologists then detected after reviewing CADafter reviewing CAD
Review of CAD ResultsReview of CAD Results Sources of false positivesSources of false positives