SlideShare a Scribd company logo
Bone tumor segmentation on bone scans using context information 
and random forests 
Introduction 
Email: gregory.h.chu@gmail.com 
Motivation: 
• Metrics derived from bone tumor segmentation 
serve as clinical endpoints in metastatic prostate 
cancer clinical drug trials [1] 
• The currently used rule-based method for bone 
tumor segmentation [2] results in a large # of false 
positives (FP). FPs are frequently due to 1) location 
dependent tumor appearance, 2) degenerative 
joint disease, 3) tracer outside of the bone 
Aims: 
• To develop a random forest (RF) classifier that uses 
contextual information (CLFw context), and evaluate 
the significance of the context features by 
comparing to a classifier that does not use context 
features (CLF w/o context) 
• To improve segmentation performance over the 
rule-based method [2] 
Methods 
Experiments and results Conclusion 
Gregory Chu1, Pechin Lo1, Bharath Ramakrishna1, Hyun J Kim1, Darren Morris2, Jonathan G. Goldin1, Matthew S. Brown1 
1 Center for Computer Vision and Imaging Biomarkers, Department of Radiological Sciences, David Geffen School of Medicine, UCLA, Los Angeles, USA 
2 MedQIA Imaging CRO, Los Angeles, USA 
Fig. 2: (a) Example result of landmark detection, (b) offset vectors from 
point p to all landmark points, (c) probability map of regions prone to 
false positives 
References 
1. Scher, H., et al.: An exploratory analysis of bone scan lesion 
area, circulating tumor cell change, pain reduction, and 
overall survival in patients with castration-resistant prostate 
cancer treated with cabozantinib. J Clin. Onc. 31, 15, 5026 
(2013) 
2. Chu, G., et al.: Prelimiary results of automated removal of 
degenerative joint disease in bone scan lesion segmentation. 
Proc. SPIE Medical Imaging, 867007 (2013) 
• 213 pairs of anterior (AP) and posterior (PA) bone 
scans from 56 sites with metastatic prostate cancer, 
with pixel spacing ranging from [2, 3] mm 
• Manual segmentation by a board certified 
radiologist 
• 100 pairs for training; 40 for validation; 73 for 
testing 
• Jaccard index and AUC (AUC is a supplemental 
metric and evaluated on subset of samples) 
Fig 4: Segmentation results of 2 cases (a, b) showing rule-based method 
[2] on the left, and proposed CLF w context on the right. TP, FP, FN. Red 
arrows indicate regions where false positives were reduced. 
• CLFw context improves the Jaccard index by 0.09 over 
the CLFw/o context (Table 1), largely due to the 
discriminative power of the context features (84% 
feature importance, Fig. 3) 
• CLFw context provides incremental improvement 
(Jaccard index increase of 0.08) over the rule-based 
method [2] (Table 1). Fig 4 demonstrates 2 cases 
where the # of false positives were reduced 
compared to method [2] 
Feature type Feature importance 
Symmetry (context) 
Landmark (context) 
Offset (context) 
Intensity + Texture 
Landmark detection 
• ASM search matches HOG descriptors using the 
Mahalanobis distance, searching horizontally, 
vertically, and diagonally, at 3 resolutions (Fig. 2a) 
Features 
• Context features: Offset vector to each landmark 
(Fig. 2b), intensity difference w.r.t. each landmark, 
intensity difference w.r.t. symmetric point across 
the midline 
• Intensity: 0th, 1st, 2nd order scale-space features 
• Texture: Grey level cooccurrence matrix properties 
• Note: all features using intensity are computed at 
3 Gaussian scales (1, 2, 4 pixels) 
Fig. 1: Workflow for proposed tumor segmentation method 
Sampling for training 
• Negatives sampled from a probability map of 
regions prone to false positives (Fig. 2c) 
Classifier Training 
• Axis-aligned RF using the Gini impurity measure, 
with 100 trees and 20 random features at each 
split node. 
Bone scan 
Intensity and 
texture features 
[2] CLFw context 
(a) (b) (c) 
Landmark detection 
Context features 
Classification 
(CLFw context) 
Tumor segmentation 
0 0.25 0.5 
Method Jaccard AUC 
[2] 0.50 ± 0.31 -- 
CLF w/o context 0.49 ± 0.26 0.91 ± 0.05 
CLF w context 0.58 ± 0.27 0.96 ± 0.03 
[2] CLFw context 
(a) (b) 
Fig 3: Gini feature importance from RF training 
Table 1: Jaccard and AUC metrics for rule-based method [2], CLF w/o 
context features, and CLF w/ context features

More Related Content

What's hot

A theoretical study on partially automated method
A theoretical study on partially automated methodA theoretical study on partially automated method
A theoretical study on partially automated method
eSAT Publishing House
 
Breast Cancer Detection and Classification using Ultrasound and Ultrasound El...
Breast Cancer Detection and Classification using Ultrasound and Ultrasound El...Breast Cancer Detection and Classification using Ultrasound and Ultrasound El...
Breast Cancer Detection and Classification using Ultrasound and Ultrasound El...
IRJET Journal
 
Comparison of Three Segmentation Methods for Breast Ultrasound Images Based o...
Comparison of Three Segmentation Methods for Breast Ultrasound Images Based o...Comparison of Three Segmentation Methods for Breast Ultrasound Images Based o...
Comparison of Three Segmentation Methods for Breast Ultrasound Images Based o...
IJECEIAES
 
Determination of Proton Energy and Dosage to Obtain SOBP Curve in the Proton ...
Determination of Proton Energy and Dosage to Obtain SOBP Curve in the Proton ...Determination of Proton Energy and Dosage to Obtain SOBP Curve in the Proton ...
Determination of Proton Energy and Dosage to Obtain SOBP Curve in the Proton ...
IJERA Editor
 
research paper Ijetae 0812 23
research paper Ijetae 0812 23research paper Ijetae 0812 23
research paper Ijetae 0812 23
Punit Karnani
 
MEDICAL IMAGE PROCESSING METHODOLOGY FOR LIVER TUMOUR DIAGNOSIS
MEDICAL IMAGE PROCESSING METHODOLOGY FOR LIVER TUMOUR DIAGNOSISMEDICAL IMAGE PROCESSING METHODOLOGY FOR LIVER TUMOUR DIAGNOSIS
MEDICAL IMAGE PROCESSING METHODOLOGY FOR LIVER TUMOUR DIAGNOSIS
ijsc
 
A magnetic resonance spectroscopy driven initialization scheme for active sha...
A magnetic resonance spectroscopy driven initialization scheme for active sha...A magnetic resonance spectroscopy driven initialization scheme for active sha...
A magnetic resonance spectroscopy driven initialization scheme for active sha...
TRS Telehealth Services
 
A Hybrid Data Analysis And Mesh Refinement Paradigm For Conformal Voxel
A Hybrid Data Analysis And Mesh Refinement Paradigm For Conformal VoxelA Hybrid Data Analysis And Mesh Refinement Paradigm For Conformal Voxel
A Hybrid Data Analysis And Mesh Refinement Paradigm For Conformal Voxel
cartiksharma
 
Comparison of radiomic features between metastatic lesions and degenerative l...
Comparison of radiomic features between metastatic lesions and degenerative l...Comparison of radiomic features between metastatic lesions and degenerative l...
Comparison of radiomic features between metastatic lesions and degenerative l...
Sin-di Lee
 
BIR_Layal_Final
BIR_Layal_FinalBIR_Layal_Final
BIR_Layal_Final
Layal Jambi
 
Optimal fuzzy rule based pulmonary nodule detection
Optimal fuzzy rule based pulmonary nodule detectionOptimal fuzzy rule based pulmonary nodule detection
Optimal fuzzy rule based pulmonary nodule detection
Wookjin Choi
 
Comparative dosimetry of forward and inverse treatment planning for Intensity...
Comparative dosimetry of forward and inverse treatment planning for Intensity...Comparative dosimetry of forward and inverse treatment planning for Intensity...
Comparative dosimetry of forward and inverse treatment planning for Intensity...
iosrjce
 
CT computer aided diagnosis system
CT computer aided diagnosis systemCT computer aided diagnosis system
CT computer aided diagnosis system
Aboul Ella Hassanien
 
Ijciet 10 02_012
Ijciet 10 02_012Ijciet 10 02_012
Ijciet 10 02_012
IAEME Publication
 
50120140506006
5012014050600650120140506006
50120140506006
IAEME Publication
 
Classification with Random Forest Based on Local Tangent Space Alignment and ...
Classification with Random Forest Based on Local Tangent Space Alignment and ...Classification with Random Forest Based on Local Tangent Space Alignment and ...
Classification with Random Forest Based on Local Tangent Space Alignment and ...
International Journal of Modern Research in Engineering and Technology
 
The International Journal of Engineering and Science (The IJES)
The International Journal of Engineering and Science (The IJES)The International Journal of Engineering and Science (The IJES)
The International Journal of Engineering and Science (The IJES)
theijes
 
ENHANCING SEGMENTATION APPROACHES FROM GAUSSIAN MIXTURE MODEL AND EXPECTED MA...
ENHANCING SEGMENTATION APPROACHES FROM GAUSSIAN MIXTURE MODEL AND EXPECTED MA...ENHANCING SEGMENTATION APPROACHES FROM GAUSSIAN MIXTURE MODEL AND EXPECTED MA...
ENHANCING SEGMENTATION APPROACHES FROM GAUSSIAN MIXTURE MODEL AND EXPECTED MA...
Christo Ananth
 

What's hot (18)

A theoretical study on partially automated method
A theoretical study on partially automated methodA theoretical study on partially automated method
A theoretical study on partially automated method
 
Breast Cancer Detection and Classification using Ultrasound and Ultrasound El...
Breast Cancer Detection and Classification using Ultrasound and Ultrasound El...Breast Cancer Detection and Classification using Ultrasound and Ultrasound El...
Breast Cancer Detection and Classification using Ultrasound and Ultrasound El...
 
Comparison of Three Segmentation Methods for Breast Ultrasound Images Based o...
Comparison of Three Segmentation Methods for Breast Ultrasound Images Based o...Comparison of Three Segmentation Methods for Breast Ultrasound Images Based o...
Comparison of Three Segmentation Methods for Breast Ultrasound Images Based o...
 
Determination of Proton Energy and Dosage to Obtain SOBP Curve in the Proton ...
Determination of Proton Energy and Dosage to Obtain SOBP Curve in the Proton ...Determination of Proton Energy and Dosage to Obtain SOBP Curve in the Proton ...
Determination of Proton Energy and Dosage to Obtain SOBP Curve in the Proton ...
 
research paper Ijetae 0812 23
research paper Ijetae 0812 23research paper Ijetae 0812 23
research paper Ijetae 0812 23
 
MEDICAL IMAGE PROCESSING METHODOLOGY FOR LIVER TUMOUR DIAGNOSIS
MEDICAL IMAGE PROCESSING METHODOLOGY FOR LIVER TUMOUR DIAGNOSISMEDICAL IMAGE PROCESSING METHODOLOGY FOR LIVER TUMOUR DIAGNOSIS
MEDICAL IMAGE PROCESSING METHODOLOGY FOR LIVER TUMOUR DIAGNOSIS
 
A magnetic resonance spectroscopy driven initialization scheme for active sha...
A magnetic resonance spectroscopy driven initialization scheme for active sha...A magnetic resonance spectroscopy driven initialization scheme for active sha...
A magnetic resonance spectroscopy driven initialization scheme for active sha...
 
A Hybrid Data Analysis And Mesh Refinement Paradigm For Conformal Voxel
A Hybrid Data Analysis And Mesh Refinement Paradigm For Conformal VoxelA Hybrid Data Analysis And Mesh Refinement Paradigm For Conformal Voxel
A Hybrid Data Analysis And Mesh Refinement Paradigm For Conformal Voxel
 
Comparison of radiomic features between metastatic lesions and degenerative l...
Comparison of radiomic features between metastatic lesions and degenerative l...Comparison of radiomic features between metastatic lesions and degenerative l...
Comparison of radiomic features between metastatic lesions and degenerative l...
 
BIR_Layal_Final
BIR_Layal_FinalBIR_Layal_Final
BIR_Layal_Final
 
Optimal fuzzy rule based pulmonary nodule detection
Optimal fuzzy rule based pulmonary nodule detectionOptimal fuzzy rule based pulmonary nodule detection
Optimal fuzzy rule based pulmonary nodule detection
 
Comparative dosimetry of forward and inverse treatment planning for Intensity...
Comparative dosimetry of forward and inverse treatment planning for Intensity...Comparative dosimetry of forward and inverse treatment planning for Intensity...
Comparative dosimetry of forward and inverse treatment planning for Intensity...
 
CT computer aided diagnosis system
CT computer aided diagnosis systemCT computer aided diagnosis system
CT computer aided diagnosis system
 
Ijciet 10 02_012
Ijciet 10 02_012Ijciet 10 02_012
Ijciet 10 02_012
 
50120140506006
5012014050600650120140506006
50120140506006
 
Classification with Random Forest Based on Local Tangent Space Alignment and ...
Classification with Random Forest Based on Local Tangent Space Alignment and ...Classification with Random Forest Based on Local Tangent Space Alignment and ...
Classification with Random Forest Based on Local Tangent Space Alignment and ...
 
The International Journal of Engineering and Science (The IJES)
The International Journal of Engineering and Science (The IJES)The International Journal of Engineering and Science (The IJES)
The International Journal of Engineering and Science (The IJES)
 
ENHANCING SEGMENTATION APPROACHES FROM GAUSSIAN MIXTURE MODEL AND EXPECTED MA...
ENHANCING SEGMENTATION APPROACHES FROM GAUSSIAN MIXTURE MODEL AND EXPECTED MA...ENHANCING SEGMENTATION APPROACHES FROM GAUSSIAN MIXTURE MODEL AND EXPECTED MA...
ENHANCING SEGMENTATION APPROACHES FROM GAUSSIAN MIXTURE MODEL AND EXPECTED MA...
 

Similar to Bone tumor segmentation on bone scans using context information and random forests

Quantitative Image Analysis for Cancer Diagnosis and Radiation Therapy
Quantitative Image Analysis for Cancer Diagnosis and Radiation TherapyQuantitative Image Analysis for Cancer Diagnosis and Radiation Therapy
Quantitative Image Analysis for Cancer Diagnosis and Radiation Therapy
Wookjin Choi
 
PERFORMANCE EVALUATION OF TUMOR DETECTION TECHNIQUES
PERFORMANCE EVALUATION OF TUMOR DETECTION TECHNIQUES PERFORMANCE EVALUATION OF TUMOR DETECTION TECHNIQUES
PERFORMANCE EVALUATION OF TUMOR DETECTION TECHNIQUES
ijcsa
 
Women in STEM
Women in STEM Women in STEM
Women in STEM
Kritika Lakhotia
 
BFO – AIS: A Framework for Medical Image Classification Using Soft Computing ...
BFO – AIS: A Framework for Medical Image Classification Using Soft Computing ...BFO – AIS: A Framework for Medical Image Classification Using Soft Computing ...
BFO – AIS: A Framework for Medical Image Classification Using Soft Computing ...
ijsc
 
BFO – AIS: A FRAME WORK FOR MEDICAL IMAGE CLASSIFICATION USING SOFT COMPUTING...
BFO – AIS: A FRAME WORK FOR MEDICAL IMAGE CLASSIFICATION USING SOFT COMPUTING...BFO – AIS: A FRAME WORK FOR MEDICAL IMAGE CLASSIFICATION USING SOFT COMPUTING...
BFO – AIS: A FRAME WORK FOR MEDICAL IMAGE CLASSIFICATION USING SOFT COMPUTING...
ijsc
 
Medical Image Processing Methodology for Liver Tumour Diagnosis
Medical Image Processing Methodology for Liver Tumour Diagnosis Medical Image Processing Methodology for Liver Tumour Diagnosis
Medical Image Processing Methodology for Liver Tumour Diagnosis
ijsc
 
Applying Deep Learning Techniques in Automated Analysis of CT scan images for...
Applying Deep Learning Techniques in Automated Analysis of CT scan images for...Applying Deep Learning Techniques in Automated Analysis of CT scan images for...
Applying Deep Learning Techniques in Automated Analysis of CT scan images for...
NEHA Kapoor
 
Texture-Based Computational Models of Tissue in Biomedical Images: Initial Ex...
Texture-Based Computational Models of Tissue in Biomedical Images: Initial Ex...Texture-Based Computational Models of Tissue in Biomedical Images: Initial Ex...
Texture-Based Computational Models of Tissue in Biomedical Images: Initial Ex...
Institute of Information Systems (HES-SO)
 
POSTER_JIANYU_LIU
POSTER_JIANYU_LIUPOSTER_JIANYU_LIU
POSTER_JIANYU_LIU
Jianyu Liu
 
Automated bone metastasis detection
Automated bone metastasis detection Automated bone metastasis detection
Automated bone metastasis detection
Kyuri Kim
 
ENHANCING SEGMENTATION APPROACHES FROM FUZZY K-C-MEANS TO FUZZY-MPSO BASED LI...
ENHANCING SEGMENTATION APPROACHES FROM FUZZY K-C-MEANS TO FUZZY-MPSO BASED LI...ENHANCING SEGMENTATION APPROACHES FROM FUZZY K-C-MEANS TO FUZZY-MPSO BASED LI...
ENHANCING SEGMENTATION APPROACHES FROM FUZZY K-C-MEANS TO FUZZY-MPSO BASED LI...
Christo Ananth
 
01531
0153101531
Design of Modified Bio-Inspired Algorithm for Identification and Segmentation...
Design of Modified Bio-Inspired Algorithm for Identification and Segmentation...Design of Modified Bio-Inspired Algorithm for Identification and Segmentation...
Design of Modified Bio-Inspired Algorithm for Identification and Segmentation...
yudhveersingh18
 
Interpretable Spiculation Quantification for Lung Cancer Screening
Interpretable Spiculation Quantification for Lung Cancer ScreeningInterpretable Spiculation Quantification for Lung Cancer Screening
Interpretable Spiculation Quantification for Lung Cancer Screening
Wookjin Choi
 
Precision Radiotherapy: Tailoring Treatment for Individualised Cancer Care.pptx
Precision Radiotherapy: Tailoring Treatment for Individualised Cancer Care.pptxPrecision Radiotherapy: Tailoring Treatment for Individualised Cancer Care.pptx
Precision Radiotherapy: Tailoring Treatment for Individualised Cancer Care.pptx
Dr. Rituparna Biswas
 
poster
posterposter
Ol3425172524
Ol3425172524Ol3425172524
Ol3425172524
IJERA Editor
 
Automatic Diagnosis of Abnormal Tumor Region from Brain Computed Tomography I...
Automatic Diagnosis of Abnormal Tumor Region from Brain Computed Tomography I...Automatic Diagnosis of Abnormal Tumor Region from Brain Computed Tomography I...
Automatic Diagnosis of Abnormal Tumor Region from Brain Computed Tomography I...
ijcseit
 
Asnr
AsnrAsnr
The International Journal of Engineering and Science (The IJES)
 The International Journal of Engineering and Science (The IJES) The International Journal of Engineering and Science (The IJES)
The International Journal of Engineering and Science (The IJES)
theijes
 

Similar to Bone tumor segmentation on bone scans using context information and random forests (20)

Quantitative Image Analysis for Cancer Diagnosis and Radiation Therapy
Quantitative Image Analysis for Cancer Diagnosis and Radiation TherapyQuantitative Image Analysis for Cancer Diagnosis and Radiation Therapy
Quantitative Image Analysis for Cancer Diagnosis and Radiation Therapy
 
PERFORMANCE EVALUATION OF TUMOR DETECTION TECHNIQUES
PERFORMANCE EVALUATION OF TUMOR DETECTION TECHNIQUES PERFORMANCE EVALUATION OF TUMOR DETECTION TECHNIQUES
PERFORMANCE EVALUATION OF TUMOR DETECTION TECHNIQUES
 
Women in STEM
Women in STEM Women in STEM
Women in STEM
 
BFO – AIS: A Framework for Medical Image Classification Using Soft Computing ...
BFO – AIS: A Framework for Medical Image Classification Using Soft Computing ...BFO – AIS: A Framework for Medical Image Classification Using Soft Computing ...
BFO – AIS: A Framework for Medical Image Classification Using Soft Computing ...
 
BFO – AIS: A FRAME WORK FOR MEDICAL IMAGE CLASSIFICATION USING SOFT COMPUTING...
BFO – AIS: A FRAME WORK FOR MEDICAL IMAGE CLASSIFICATION USING SOFT COMPUTING...BFO – AIS: A FRAME WORK FOR MEDICAL IMAGE CLASSIFICATION USING SOFT COMPUTING...
BFO – AIS: A FRAME WORK FOR MEDICAL IMAGE CLASSIFICATION USING SOFT COMPUTING...
 
Medical Image Processing Methodology for Liver Tumour Diagnosis
Medical Image Processing Methodology for Liver Tumour Diagnosis Medical Image Processing Methodology for Liver Tumour Diagnosis
Medical Image Processing Methodology for Liver Tumour Diagnosis
 
Applying Deep Learning Techniques in Automated Analysis of CT scan images for...
Applying Deep Learning Techniques in Automated Analysis of CT scan images for...Applying Deep Learning Techniques in Automated Analysis of CT scan images for...
Applying Deep Learning Techniques in Automated Analysis of CT scan images for...
 
Texture-Based Computational Models of Tissue in Biomedical Images: Initial Ex...
Texture-Based Computational Models of Tissue in Biomedical Images: Initial Ex...Texture-Based Computational Models of Tissue in Biomedical Images: Initial Ex...
Texture-Based Computational Models of Tissue in Biomedical Images: Initial Ex...
 
POSTER_JIANYU_LIU
POSTER_JIANYU_LIUPOSTER_JIANYU_LIU
POSTER_JIANYU_LIU
 
Automated bone metastasis detection
Automated bone metastasis detection Automated bone metastasis detection
Automated bone metastasis detection
 
ENHANCING SEGMENTATION APPROACHES FROM FUZZY K-C-MEANS TO FUZZY-MPSO BASED LI...
ENHANCING SEGMENTATION APPROACHES FROM FUZZY K-C-MEANS TO FUZZY-MPSO BASED LI...ENHANCING SEGMENTATION APPROACHES FROM FUZZY K-C-MEANS TO FUZZY-MPSO BASED LI...
ENHANCING SEGMENTATION APPROACHES FROM FUZZY K-C-MEANS TO FUZZY-MPSO BASED LI...
 
01531
0153101531
01531
 
Design of Modified Bio-Inspired Algorithm for Identification and Segmentation...
Design of Modified Bio-Inspired Algorithm for Identification and Segmentation...Design of Modified Bio-Inspired Algorithm for Identification and Segmentation...
Design of Modified Bio-Inspired Algorithm for Identification and Segmentation...
 
Interpretable Spiculation Quantification for Lung Cancer Screening
Interpretable Spiculation Quantification for Lung Cancer ScreeningInterpretable Spiculation Quantification for Lung Cancer Screening
Interpretable Spiculation Quantification for Lung Cancer Screening
 
Precision Radiotherapy: Tailoring Treatment for Individualised Cancer Care.pptx
Precision Radiotherapy: Tailoring Treatment for Individualised Cancer Care.pptxPrecision Radiotherapy: Tailoring Treatment for Individualised Cancer Care.pptx
Precision Radiotherapy: Tailoring Treatment for Individualised Cancer Care.pptx
 
poster
posterposter
poster
 
Ol3425172524
Ol3425172524Ol3425172524
Ol3425172524
 
Automatic Diagnosis of Abnormal Tumor Region from Brain Computed Tomography I...
Automatic Diagnosis of Abnormal Tumor Region from Brain Computed Tomography I...Automatic Diagnosis of Abnormal Tumor Region from Brain Computed Tomography I...
Automatic Diagnosis of Abnormal Tumor Region from Brain Computed Tomography I...
 
Asnr
AsnrAsnr
Asnr
 
The International Journal of Engineering and Science (The IJES)
 The International Journal of Engineering and Science (The IJES) The International Journal of Engineering and Science (The IJES)
The International Journal of Engineering and Science (The IJES)
 

Recently uploaded

ISPM 15 Heat Treated Wood Stamps and why your shipping must have one
ISPM 15 Heat Treated Wood Stamps and why your shipping must have oneISPM 15 Heat Treated Wood Stamps and why your shipping must have one
ISPM 15 Heat Treated Wood Stamps and why your shipping must have one
Las Vegas Warehouse
 
Null Bangalore | Pentesters Approach to AWS IAM
Null Bangalore | Pentesters Approach to AWS IAMNull Bangalore | Pentesters Approach to AWS IAM
Null Bangalore | Pentesters Approach to AWS IAM
Divyanshu
 
学校原版美国波士顿大学毕业证学历学位证书原版一模一样
学校原版美国波士顿大学毕业证学历学位证书原版一模一样学校原版美国波士顿大学毕业证学历学位证书原版一模一样
学校原版美国波士顿大学毕业证学历学位证书原版一模一样
171ticu
 
Advanced control scheme of doubly fed induction generator for wind turbine us...
Advanced control scheme of doubly fed induction generator for wind turbine us...Advanced control scheme of doubly fed induction generator for wind turbine us...
Advanced control scheme of doubly fed induction generator for wind turbine us...
IJECEIAES
 
CompEx~Manual~1210 (2).pdf COMPEX GAS AND VAPOURS
CompEx~Manual~1210 (2).pdf COMPEX GAS AND VAPOURSCompEx~Manual~1210 (2).pdf COMPEX GAS AND VAPOURS
CompEx~Manual~1210 (2).pdf COMPEX GAS AND VAPOURS
RamonNovais6
 
john krisinger-the science and history of the alcoholic beverage.pptx
john krisinger-the science and history of the alcoholic beverage.pptxjohn krisinger-the science and history of the alcoholic beverage.pptx
john krisinger-the science and history of the alcoholic beverage.pptx
Madan Karki
 
Software Quality Assurance-se412-v11.ppt
Software Quality Assurance-se412-v11.pptSoftware Quality Assurance-se412-v11.ppt
Software Quality Assurance-se412-v11.ppt
TaghreedAltamimi
 
Curve Fitting in Numerical Methods Regression
Curve Fitting in Numerical Methods RegressionCurve Fitting in Numerical Methods Regression
Curve Fitting in Numerical Methods Regression
Nada Hikmah
 
Material for memory and display system h
Material for memory and display system hMaterial for memory and display system h
Material for memory and display system h
gowrishankartb2005
 
2008 BUILDING CONSTRUCTION Illustrated - Ching Chapter 02 The Building.pdf
2008 BUILDING CONSTRUCTION Illustrated - Ching Chapter 02 The Building.pdf2008 BUILDING CONSTRUCTION Illustrated - Ching Chapter 02 The Building.pdf
2008 BUILDING CONSTRUCTION Illustrated - Ching Chapter 02 The Building.pdf
Yasser Mahgoub
 
Embedded machine learning-based road conditions and driving behavior monitoring
Embedded machine learning-based road conditions and driving behavior monitoringEmbedded machine learning-based road conditions and driving behavior monitoring
Embedded machine learning-based road conditions and driving behavior monitoring
IJECEIAES
 
Generative AI leverages algorithms to create various forms of content
Generative AI leverages algorithms to create various forms of contentGenerative AI leverages algorithms to create various forms of content
Generative AI leverages algorithms to create various forms of content
Hitesh Mohapatra
 
一比一原版(CalArts毕业证)加利福尼亚艺术学院毕业证如何办理
一比一原版(CalArts毕业证)加利福尼亚艺术学院毕业证如何办理一比一原版(CalArts毕业证)加利福尼亚艺术学院毕业证如何办理
一比一原版(CalArts毕业证)加利福尼亚艺术学院毕业证如何办理
ecqow
 
官方认证美国密歇根州立大学毕业证学位证书原版一模一样
官方认证美国密歇根州立大学毕业证学位证书原版一模一样官方认证美国密歇根州立大学毕业证学位证书原版一模一样
官方认证美国密歇根州立大学毕业证学位证书原版一模一样
171ticu
 
22CYT12-Unit-V-E Waste and its Management.ppt
22CYT12-Unit-V-E Waste and its Management.ppt22CYT12-Unit-V-E Waste and its Management.ppt
22CYT12-Unit-V-E Waste and its Management.ppt
KrishnaveniKrishnara1
 
IEEE Aerospace and Electronic Systems Society as a Graduate Student Member
IEEE Aerospace and Electronic Systems Society as a Graduate Student MemberIEEE Aerospace and Electronic Systems Society as a Graduate Student Member
IEEE Aerospace and Electronic Systems Society as a Graduate Student Member
VICTOR MAESTRE RAMIREZ
 
Certificates - Mahmoud Mohamed Moursi Ahmed
Certificates - Mahmoud Mohamed Moursi AhmedCertificates - Mahmoud Mohamed Moursi Ahmed
Certificates - Mahmoud Mohamed Moursi Ahmed
Mahmoud Morsy
 
KuberTENes Birthday Bash Guadalajara - K8sGPT first impressions
KuberTENes Birthday Bash Guadalajara - K8sGPT first impressionsKuberTENes Birthday Bash Guadalajara - K8sGPT first impressions
KuberTENes Birthday Bash Guadalajara - K8sGPT first impressions
Victor Morales
 
Electric vehicle and photovoltaic advanced roles in enhancing the financial p...
Electric vehicle and photovoltaic advanced roles in enhancing the financial p...Electric vehicle and photovoltaic advanced roles in enhancing the financial p...
Electric vehicle and photovoltaic advanced roles in enhancing the financial p...
IJECEIAES
 
Welding Metallurgy Ferrous Materials.pdf
Welding Metallurgy Ferrous Materials.pdfWelding Metallurgy Ferrous Materials.pdf
Welding Metallurgy Ferrous Materials.pdf
AjmalKhan50578
 

Recently uploaded (20)

ISPM 15 Heat Treated Wood Stamps and why your shipping must have one
ISPM 15 Heat Treated Wood Stamps and why your shipping must have oneISPM 15 Heat Treated Wood Stamps and why your shipping must have one
ISPM 15 Heat Treated Wood Stamps and why your shipping must have one
 
Null Bangalore | Pentesters Approach to AWS IAM
Null Bangalore | Pentesters Approach to AWS IAMNull Bangalore | Pentesters Approach to AWS IAM
Null Bangalore | Pentesters Approach to AWS IAM
 
学校原版美国波士顿大学毕业证学历学位证书原版一模一样
学校原版美国波士顿大学毕业证学历学位证书原版一模一样学校原版美国波士顿大学毕业证学历学位证书原版一模一样
学校原版美国波士顿大学毕业证学历学位证书原版一模一样
 
Advanced control scheme of doubly fed induction generator for wind turbine us...
Advanced control scheme of doubly fed induction generator for wind turbine us...Advanced control scheme of doubly fed induction generator for wind turbine us...
Advanced control scheme of doubly fed induction generator for wind turbine us...
 
CompEx~Manual~1210 (2).pdf COMPEX GAS AND VAPOURS
CompEx~Manual~1210 (2).pdf COMPEX GAS AND VAPOURSCompEx~Manual~1210 (2).pdf COMPEX GAS AND VAPOURS
CompEx~Manual~1210 (2).pdf COMPEX GAS AND VAPOURS
 
john krisinger-the science and history of the alcoholic beverage.pptx
john krisinger-the science and history of the alcoholic beverage.pptxjohn krisinger-the science and history of the alcoholic beverage.pptx
john krisinger-the science and history of the alcoholic beverage.pptx
 
Software Quality Assurance-se412-v11.ppt
Software Quality Assurance-se412-v11.pptSoftware Quality Assurance-se412-v11.ppt
Software Quality Assurance-se412-v11.ppt
 
Curve Fitting in Numerical Methods Regression
Curve Fitting in Numerical Methods RegressionCurve Fitting in Numerical Methods Regression
Curve Fitting in Numerical Methods Regression
 
Material for memory and display system h
Material for memory and display system hMaterial for memory and display system h
Material for memory and display system h
 
2008 BUILDING CONSTRUCTION Illustrated - Ching Chapter 02 The Building.pdf
2008 BUILDING CONSTRUCTION Illustrated - Ching Chapter 02 The Building.pdf2008 BUILDING CONSTRUCTION Illustrated - Ching Chapter 02 The Building.pdf
2008 BUILDING CONSTRUCTION Illustrated - Ching Chapter 02 The Building.pdf
 
Embedded machine learning-based road conditions and driving behavior monitoring
Embedded machine learning-based road conditions and driving behavior monitoringEmbedded machine learning-based road conditions and driving behavior monitoring
Embedded machine learning-based road conditions and driving behavior monitoring
 
Generative AI leverages algorithms to create various forms of content
Generative AI leverages algorithms to create various forms of contentGenerative AI leverages algorithms to create various forms of content
Generative AI leverages algorithms to create various forms of content
 
一比一原版(CalArts毕业证)加利福尼亚艺术学院毕业证如何办理
一比一原版(CalArts毕业证)加利福尼亚艺术学院毕业证如何办理一比一原版(CalArts毕业证)加利福尼亚艺术学院毕业证如何办理
一比一原版(CalArts毕业证)加利福尼亚艺术学院毕业证如何办理
 
官方认证美国密歇根州立大学毕业证学位证书原版一模一样
官方认证美国密歇根州立大学毕业证学位证书原版一模一样官方认证美国密歇根州立大学毕业证学位证书原版一模一样
官方认证美国密歇根州立大学毕业证学位证书原版一模一样
 
22CYT12-Unit-V-E Waste and its Management.ppt
22CYT12-Unit-V-E Waste and its Management.ppt22CYT12-Unit-V-E Waste and its Management.ppt
22CYT12-Unit-V-E Waste and its Management.ppt
 
IEEE Aerospace and Electronic Systems Society as a Graduate Student Member
IEEE Aerospace and Electronic Systems Society as a Graduate Student MemberIEEE Aerospace and Electronic Systems Society as a Graduate Student Member
IEEE Aerospace and Electronic Systems Society as a Graduate Student Member
 
Certificates - Mahmoud Mohamed Moursi Ahmed
Certificates - Mahmoud Mohamed Moursi AhmedCertificates - Mahmoud Mohamed Moursi Ahmed
Certificates - Mahmoud Mohamed Moursi Ahmed
 
KuberTENes Birthday Bash Guadalajara - K8sGPT first impressions
KuberTENes Birthday Bash Guadalajara - K8sGPT first impressionsKuberTENes Birthday Bash Guadalajara - K8sGPT first impressions
KuberTENes Birthday Bash Guadalajara - K8sGPT first impressions
 
Electric vehicle and photovoltaic advanced roles in enhancing the financial p...
Electric vehicle and photovoltaic advanced roles in enhancing the financial p...Electric vehicle and photovoltaic advanced roles in enhancing the financial p...
Electric vehicle and photovoltaic advanced roles in enhancing the financial p...
 
Welding Metallurgy Ferrous Materials.pdf
Welding Metallurgy Ferrous Materials.pdfWelding Metallurgy Ferrous Materials.pdf
Welding Metallurgy Ferrous Materials.pdf
 

Bone tumor segmentation on bone scans using context information and random forests

  • 1. Bone tumor segmentation on bone scans using context information and random forests Introduction Email: gregory.h.chu@gmail.com Motivation: • Metrics derived from bone tumor segmentation serve as clinical endpoints in metastatic prostate cancer clinical drug trials [1] • The currently used rule-based method for bone tumor segmentation [2] results in a large # of false positives (FP). FPs are frequently due to 1) location dependent tumor appearance, 2) degenerative joint disease, 3) tracer outside of the bone Aims: • To develop a random forest (RF) classifier that uses contextual information (CLFw context), and evaluate the significance of the context features by comparing to a classifier that does not use context features (CLF w/o context) • To improve segmentation performance over the rule-based method [2] Methods Experiments and results Conclusion Gregory Chu1, Pechin Lo1, Bharath Ramakrishna1, Hyun J Kim1, Darren Morris2, Jonathan G. Goldin1, Matthew S. Brown1 1 Center for Computer Vision and Imaging Biomarkers, Department of Radiological Sciences, David Geffen School of Medicine, UCLA, Los Angeles, USA 2 MedQIA Imaging CRO, Los Angeles, USA Fig. 2: (a) Example result of landmark detection, (b) offset vectors from point p to all landmark points, (c) probability map of regions prone to false positives References 1. Scher, H., et al.: An exploratory analysis of bone scan lesion area, circulating tumor cell change, pain reduction, and overall survival in patients with castration-resistant prostate cancer treated with cabozantinib. J Clin. Onc. 31, 15, 5026 (2013) 2. Chu, G., et al.: Prelimiary results of automated removal of degenerative joint disease in bone scan lesion segmentation. Proc. SPIE Medical Imaging, 867007 (2013) • 213 pairs of anterior (AP) and posterior (PA) bone scans from 56 sites with metastatic prostate cancer, with pixel spacing ranging from [2, 3] mm • Manual segmentation by a board certified radiologist • 100 pairs for training; 40 for validation; 73 for testing • Jaccard index and AUC (AUC is a supplemental metric and evaluated on subset of samples) Fig 4: Segmentation results of 2 cases (a, b) showing rule-based method [2] on the left, and proposed CLF w context on the right. TP, FP, FN. Red arrows indicate regions where false positives were reduced. • CLFw context improves the Jaccard index by 0.09 over the CLFw/o context (Table 1), largely due to the discriminative power of the context features (84% feature importance, Fig. 3) • CLFw context provides incremental improvement (Jaccard index increase of 0.08) over the rule-based method [2] (Table 1). Fig 4 demonstrates 2 cases where the # of false positives were reduced compared to method [2] Feature type Feature importance Symmetry (context) Landmark (context) Offset (context) Intensity + Texture Landmark detection • ASM search matches HOG descriptors using the Mahalanobis distance, searching horizontally, vertically, and diagonally, at 3 resolutions (Fig. 2a) Features • Context features: Offset vector to each landmark (Fig. 2b), intensity difference w.r.t. each landmark, intensity difference w.r.t. symmetric point across the midline • Intensity: 0th, 1st, 2nd order scale-space features • Texture: Grey level cooccurrence matrix properties • Note: all features using intensity are computed at 3 Gaussian scales (1, 2, 4 pixels) Fig. 1: Workflow for proposed tumor segmentation method Sampling for training • Negatives sampled from a probability map of regions prone to false positives (Fig. 2c) Classifier Training • Axis-aligned RF using the Gini impurity measure, with 100 trees and 20 random features at each split node. Bone scan Intensity and texture features [2] CLFw context (a) (b) (c) Landmark detection Context features Classification (CLFw context) Tumor segmentation 0 0.25 0.5 Method Jaccard AUC [2] 0.50 ± 0.31 -- CLF w/o context 0.49 ± 0.26 0.91 ± 0.05 CLF w context 0.58 ± 0.27 0.96 ± 0.03 [2] CLFw context (a) (b) Fig 3: Gini feature importance from RF training Table 1: Jaccard and AUC metrics for rule-based method [2], CLF w/o context features, and CLF w/ context features