Intelligent Medical Imaging for Breast Cancer Detection and Diabetic Retinopathy
adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging for Breast Cancer Detection and Diabetic Retinopathy Hanung Adi Nugroho Inovasi E-Health and Biomedika untuk Indonesia
Research and Development of Intelligent Medical Imaging Profile
Dr. Ir. Hanung Adi Nugroho
Department of Electrical Engineering and Information Technology Faculty of Engineering, UNIVERSITAS GADJAH MADA Jl. Grafika 2, Kampus UGM, Yogyakarta 55281, Indonesia Telp./ fax. +62-274-552305 Email: adinugroho@ugm.ac.id; adinugroho@ieee.org
Research areas: Biomedical signal and image processing and analysis; computer vision; medical instrumentation; pattern recognition; data mining; statistical data analysis.
Bachelor of Engineering (S.T.) – Teknik Elektro, Universitas Gadjah Mada, Yogyakarta, Indonesia (2001) Master of Engineering (M.E.) – School of Information Technology and Electrical Engineering, The University of Queensland, St Lucia, Brisbane, Australia (2005) Doctor of Philosophy (Ph.D.) – Electrical and Electronics Engineering Department, Universiti Teknologi PETRONAS, Seri Iskandar, Malaysia (2012)
Intelligent Medical Imaging Research
Medical imaging - Overview
Currently medical imaging is limited to the acquisition of images of the human organs/ body Medical imaging refers to the techniques and processes used to create images of the human body for clinical purposes (medical procedures seeking to reveal, diagnose or examine disease). Medical imaging can be seen as the solution of mathematical inverse problems. This means that cause (the properties of living tissue) is inferred from
Analysis of the images obtained
effect (the observed signal)
is performed clinically by experts adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research
Medical imaging - Technology
a short-lived isotope, such Gamma ray : positron emission tomography (PET) as 18F, is incorporated into a substance used by the body such as glucose which is absorbed by the tumour of interest
X ray : computed tomography (CT) Expose to x-ray radiation, repeated scans must be limited to avoid health effects
Intelligent Medical Imaging Research
Magnetic resonance imaging (MRI) uses powerful magnets to polarise and excite hydrogen nuclei (single proton) in water molecules in human tissue, producing a detectable signal which is spatially encoded resulting in images of the body excellent soft-tissue contrast no known long term effects of exposure to strong static fields health risks associated with tissue heating from exposure to the RF field and the presence of implanted devices in the body, such as pace makers
adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research Medical imaging - Technology
Ultrasound : ultrasonography H-F sound, 2-10MHz, safe, 2D moving images
Fundus camera Retinal image
Intelligent Medical Imaging Research Issues, challenge and approach
Issues Issues Approach Approach
- Harmful (radiation, contrast agent) • Harmful (radiation, contrast agent)
From From medical medical imaging imaging (image (image
- Specialized device – difficult to use - • Specialized device – difficult to use - acquisition with enhancement) to acquisition with enhancement) to highly trained operator needed highly trained operator needed medical image analysis (feature medical image analysis (feature
- Expensive (Initial cost, Maintenance) • Expensive (Initial cost, Maintenance) extraction, extraction, classification, classification, pattern pattern
- Image Acquisition only, little or no • Image Acquisition only, little or no recognition, measurements) resulting recognition, measurements) resulting analysis for diagnostic purposes, analysis for diagnostic purposes, in in intelligent intelligent imaging imaging (decision (decision subjective subjective support systems) support systems)
Challenge Challenge
To develop intelligent medical imaging system which is objective in analysis that is To develop intelligent medical imaging system which is objective in analysis that is safe to the patients. safe to the patients.
adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research Current research in intelligent medical imaging system at DTETI-UGM
Radiology (Breast cancer) Ophthalmology (Diabetic retinopathy)
Diabetic retinopathy Breast cancer Glaucoma
Intelligent Medical Imaging Research in Breast Cancer Intelligent Medical Imaging Research in Breast Cancer adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Breast Cancer Breast Cancer Breast cancer is a disease in which malignant (cancer) cells form in the tissues of the breast.
- Breast compression
- Low-dose X-ray
- Just for particular patient
CAD CAD : Computer Aided Diagnosis Intelligent Medical Imaging Research in Breast Cancer
Breast Self Exam Mammograms USG MRI
- Limited availability
- Elaborate the radiology knowledge into image processing and analysis technology
- Assist radiologists to diagnose nodule
- Low cost
- Short acquisition time
- No radiations
- High availability
- Convenient more sensitive
- Depend on operator
- No Radiation More Detail
- Radiologist s experience
- Inconsistency of interpretation
- Expensive Limited availability
adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Breast Cancer Breast Cancer Detection
Research Objective To develop a computer aided diagnosis (CAD) system for classifying breast nodule in ultrasound (US) images to distinguish benign and malignant nodules.
Intelligent Medical Imaging Research in Breast Cancer Diagnosis of Breast Cancer using Ultrasound
A breast ultrasound is a scan that uses penetrating sound waves that do not affect or damage the tissue and cannot be heard by humans.
Normal Abnormal adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Breast Cancer
Methodology
Intelligent Medical Imaging Research in Breast Cancer Scheme of CAD System Computer Aided System USG Image Processing Image Image Image Analysis Display Acquisitions
Diagnosis Radiologists adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Breast Cancer System Diagram Block
Preprocessing Noise and GrayScale (4) USG (1) Marker Segmentation RoI (2) Conversion (3) Images Reduction
(5) Feature Extraction
- Moment based
Malignant / Birads based Feature
(8) features
Diagnosis (6) (7) Benign Classification Selection
- Geometry Feature • Texture Feature
Texture Features
adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Breast Cancer
Nodule Background
Segmented Area Posterior Characteristic Margin Characteristic
Echo Pattern Characteristic
Texture Analysis
Intelligent Medical Imaging Research in Breast Cancer Geometry Features
Geometric feature is constructed by a set of geometrical elements such as points, lines, curves or surfaces = =
=
1 ∑
. exp ( 2 ) = =
= = 4 .
Intelligent Medical Imaging Research in Breast Cancer Geometry and Moment Based Features
Nodule Moment Based Analysis
Shape characteristics Geometry Analysis
Margin characteristics Background
adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Breast Cancer Research Roadmap Clinical Integrated Trial Modules
Prototype 2 2018 Margin and Posterior Features
2016 Prototype 1 2015 Echo
Pattern Shape and 2014 Boundary
Intelligent Medical Imaging Research in Breast Cancer Results adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Breast Cancer Diagnosis Rules for i-Brids V.1
Benign
- Circumscr>Round -
- Not Circumscribed Malignant
Unmarked Hypoechoic or Hypoechoic
- Circumscribed Malig>Irreg
- Not Circumscribed Malignant
Intelligent Medical Imaging Research in Breast Cancer
Segmentation and Image ROI and Filtering Diagnosis
Feature Extraction Capturing adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Breast Cancer Performance Analysis of CAD System
100.00% 90.00% 80.00% 70.00% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00%
0.00% Accuracy Sensitivity Specificity PPV NPV
Shape 96.20% 94.70% 97.90% 94.73% 97.91% Margin
80.90% 79.50% 82.50% 78.50% 82.50% Echo
91.23% 95.83% 87.88% 85.19% 96.67%
Intelligent Medical Imaging Research in Breast Cancer Statistical Analysis Diagnosis Statistical Analysis
No
Malignant Benign
1 Number of Features Agreement
19
13
12.5
6.5
2 Number of Features due to Chance
3 Total Number of Subjects
38
4 Total Number of Agreement
32
19
5 Number of Agreement due to chance
6 Kappa
0.68 Kappa statistics are commonly used to indicate the degree of agreement of nominal assessments made by multiple appraisers.
A Kappa 0.68 is in the “substantial” agreement range between radiologists and CAD system.
adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Breast Cancer Diagnosis Rules for i-Brids V.2
Benign/Malignant
- No Posterior Fea>Circumscribed • Enhancement Be
- Shadowing Malignant
Unmarked Hypoechoic or Hypoechoic
Malignant
- No Posterior Fea>Not Circumscribed • Enhancement Malig
- Shadowing Malignant
Intelligent Medical Imaging Research in Breast Cancer
Image Segmentation and ROI and Filtering Diagnosis Feature Extraction
Capturing adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Breast Cancer Accuracy of CAD System
98.00% 96.00% 94.00% 92.00% 90.00% 88.00% 86.00% 84.00% 82.00% 80.00% 78.00% 76.00%
Margin Posterior Diagnosis Radiologist 1
89.47% 84.21% 97% Radiologist 2
89.47% 86.84% 97%
Intelligent Medical Imaging Research in Breast Cancer Performance Analysis of CAD System
100% 99% 98% 97% 96% 95% 94% 93% 92%
Sensitivity Specificity PPV NPV Radiologist 1 100% 94.74% 95% 100% Radiologist 2 100% 94.74% 95% 100%
adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Breast Cancer Statistical Analysis Margin Posterior Diagnosis
No Statistical Analysis
No Circumscribed Indistinct Enhancement Posterior Shadow Malignant Benign
1 Number of Features Agreement
22
16
21
3
9
19
19
2 Number of Features due to Chance
12.74
6.74
11.6
0.79
3.47
9.5
9.5
3 Total Number of Subjects
38
38
38
4 Total Number of Agreement
38
33
38
5 Number of Agreement due to chance
19.47
19
15.87
6 Cohen's Kappa
1 0.774
1 A Kappa 1 is in the “perfect” agreement range between two radiologist
A Kappa 0.74 is in the “substantial” agreement range between two radiologist
Intelligent Medical Imaging Research in Breast Cancer Video of i-Brids
adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Breast Cancer Potential Market Total: 6430
Number of Health Care Fasilities in Indonesia
3451 1599
1380 Health Hospitals/Clinics Puskesmas
Laboratories
Breast Cancer :
adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Breast Cancer Recognition [1] H. A. Nugroho, N. Faisal, I. Soesanti, and L. Choridah, “Analysis of Computer Aided Diagnosis on Digital Mammogram Images,” in 2014 International Conference on Computer, Control, Informatics and Its Applications Analysis, 2014, pp. 25–29.
[2] A. Nugroho, H. A. Nugroho, and L. Choridah, “Active Contour Bilateral Filter for Breast Lesions Segmentation on Ultrasound Images,” in 2015 International Conference on Science in Information Technology (ICSITech) Active, 2015, pp. 36–40. [3] H. A. Nugroho, Y. Triyani, M. Rahmawaty, , I. Ardiyanto ,and L. Choridah, “Performance Analysis of Filtering Techniques for Speckle Reduction on Breast Ultrasound Images,” in 2016 International Electronics Symposium (IES), 2016, pp. 454–458. [4] M. Rahmawaty, H. A. Nugroho, Y. Triyani, I. Ardiyanto, and I. Soesanti, “Classification of Breast Ultrasound Images based on Texture Analysis,” in iBioMed 2016, 2016, pp. 84–89. [5] Y. Triyani, H. A. Nugroho, M. Rahmawaty, I. Ardiyanto, and L. Choridah, “Performance Analysis of Image Segmentation for Breast Ultrasound Images,” in ICITEE 2016, 2016, no. October, pp. 415–420. [6] H. K. N. Yusufiyah, H. A. Nugroho, T. B. Adji, and A. Nugroho, “Feature Extraction for Classifying Lesion ’ s Shape of Breast Ultrasound Images,” 2nd Int. Conf. Inf. Technol. Comput. Electr. Eng., pp. 105–109, 2015. [7] H. A. Nugroho, H. Khuzaimah, N. Yusufiyah, T. B. Adji, and A. Nugroho, “Zernike Moment Feature Extraction for Classifying Lesion ’ s Shape of Breast Ultrasound Images,” in 7th International Conference on Information Technology and Electrical Engineering (ICITEE), 2015, pp. 458–463. [8] H. A. Nugroho, N. Faisal, I. Soesanti, and L. Choridah, “Identification of Malignant Masses on Digital Mammogram Images based on Texture Feature and Correlation based Feature Selection Hanung,” in 6th International Conference on Information Technology and Electrical Engineering (ICITEE),
2014. [9] H.R. Fajrin, H. A. Nugroho, and I. Soesanti“Ekstraksi Ciri Berbasis Wavelet Dan Glcm Untuk Deteksi Dini Kanker Payudara Pada Citra Mammogram,” in SNST, 2015, pp. 47–52.
[10] M. Sahar, H. A. Nugroho, Tanur, I. Ardiyanto, and L. Choridah “Automated Detection of Breast Cancer Lesions Using Adaptive Thresholding and Morphological Operation,” in International Conference on Information Technology Systems and Innovation (ICITSI), 2016. [11] Tianur, H. A. Nugroho, M. Sahar, R. Indrastuti, and L. Choridah, “Classification of Breast Ultrasound Images based on Posterior Feature,” in International Conference on Information Technology Systems and Innovation (ICITSI), 2016.
Intelligent Medical Imaging Research in Breast Cancer Team Members and Collaborators Department Electrical Engineering and Information Technology Faculty of Engineering Universitas Gadjah Mada Department of Radiology Sardjito Hospital, Yogyakarta
- H A Nugroho • I Ardiyanto • M Rahmawaty • Y Triyani • M Sahar • L Choridah • R. Indrastuti • A. Mardhiah • Tianur • A Nugroho • D A Husna • H Khuzaimah • R L Buana
adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Diabetic Retinopathy Intelligent Medical Imaging Research in Diabetic Retinopathy
What is diabetic retinopathy TYPE 1 DIABETES: when the pancreas
doesn’t produce insulin
when the pancreas doesn’t produce enough insulin (or the insulin cannot be processed)
GESTATIONAL DIABETES: when the
insulin is less effective during pregnancy
your body needs insulin to transform glucose into energy
Types of diabetes:
TYPE 1 DIABETES:
Intelligent Medical Imaging Research in Diabetic Retinopathy What is diabetic retinopathy
DR is retinopathy (damage to the retina) caused by
complications of diabetes mellitus, which could eventually lead to blindness.
Fact : Diabetic Retinopathy Diabetic
- Nearly all patients of type-1 diabetes and 60% of
(DR) Cardiomyopathy patients of type-2 diabetes indicate retinopathy.
- DR is the leading cause of the blindness in developing countries among adults aged 20-74 years.
Diabetic Neuropathy Diabetic Nephropathy
Normal vision DR vision adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Diabetic Retinopathy
Diabetes : fact and figures “Worldwide”
2015: 415 million people with diabetes 2040: 642 million people with diabetes
adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Diabetic Retinopathy The pathologies of DR new blood vessels
Haemorrhages
Proliferative DR Severe NPDR Intelligent Medical Imaging Research in Diabetic Retinopathy
exudates Micro aneurysms Moderate NPDR Mild NPDR No DR
Issues, challenges and approaches
Issues Diabetes mellitus affect ~10% population (DR is a real concern - epidemic stage?) Needs access to ophthalmologist with fundus camera equipment Low contrast Fundus images requiring Fluorescein angiography - an invasive procedure
Challenges
1. Can we develop a screening & grading system to be made accessible to all diabetes patients?
2. Can we detect DR early even before patient have visual problems?
3. Can we make non-invasive procedure as effective?
Fundus camera technology
- Image Processing & Computer Vision
Intelligent Medical Imaging Research in Diabetic Retinopathy Haemorrhages detection
Fundus image Enhancement Enhancement Haemorrhages candidates Detected Haemorrhages Haemorrhages candidates Detected Haemorrhages
Green and V band Histogram Retinal vessels Retinal vessels extraction matching detection elimination
Two-dimensional matched filtering Contrast Opening
Double length Masking enhancement operation filtering operation
Haemorrhages Pre-processing Post-processing candidate detection adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Diabetic Retinopathy
Hard exudates detection
Removal OD and
Detected
Fundus image Filtered image Detected OD Candidate Candidate Hard Exudates Exudates
Hard Exudates Hard Exudates Green channel Complement
Removal OD and
extraction operation
Optic disc (OD) Morphological
[1]
detection operation [1] H. A. Nugroho, K. W. Oktoeberza, T. B. Adji, and M. B. Sasongko, "Segmentation of exudates based on high pass filtering in retinal fundus images," in 2015 7th International Conference on Information Technology and Electrical Engineering (ICITEE), 2015, pp. 436-441. Matched filter
Intelligent Medical Imaging Research in Diabetic Retinopathy Research roadmap
General : to develop a system to assist the ophthalmologists in monitoring and diagnosing diabetic retinopathy disease. First year: to develop algorithms in each module to detect structures and pathologies in DR retinal image. Second year: to integrate the modules and develop an algorithm for screening DR system. Third year: to test the system based on clinical study for monitoring and grading system.
DR monitoring and 2018
Clinical study grading system
Analysis of DR System evaluation pathologies 2017
DR pathologies Classification
DR screening detection system
FAZ detection 2016
Haemorrhages and hard Micro aneurysms exudates detection detection
Macula detection 2015
Optic disc detection adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Diabetic Retinopathy
Recognition
International Conferences
[1] H. A. Nugroho, D. A. Dharmawan, I. Hidayah, and L. Listyalina, "Automated microaneurysms (MAs) detection in digital colour fundus images using matched filter," in Computer, Control, Informatics and its Applications (IC3INA), 2015 International Conference on, 2015, pp. 104-108.
[2] H. A. Nugroho, L. Listyalina, N. A. Setiawan, S. Wibirama, and D. A. Dharmawan, "Automated segmentation of optic disc area using mathematical morphology and active contour," in Computer, Control, Informatics and its Applications (IC3INA), 2015 International Conference on, 2015, pp. 18-22.
[3] H. A. Nugroho, D. Purnamasari, I. Soesanti, K. W. Oktoeberza, and D. A. Dharmawan, "Detection of foveal avascular zone in colour retinal fundus images," in 2015 International Conference on Science in Information Technology (ICSITech), 2015, pp. 225-230. [4] H. A. Nugroho, K. W. Oktoeberza, T. B. Adji, and M. B. Sasongko, "Segmentation of exudates based on high pass filtering in retinal fundus images," in 2015 7th International Conference on Information Technology and Electrical Engineering
(ICITEE), 2015, pp. 436-441.
[5] H.A. Nugroho, L. Listyalina, and D. A. Dharmawan, "A New Approach for Detection of Retinal Haemorrhages in Colour Fundus Images," presented at the International Seminar on Sensors, Instrumentation, Measurement and Metrology, 2016. [7] I. Ardiyanto, H.A. Nugroho, and R. L. B. Buana, "Maximum Entropy Principle for Exudates Segmentation in Retinal Fundus Images," presented at the International Seminar on Sensors, Instrumentation, Measurement and Metrology, 2016. [8] H.A. Nugroho, W.KZ. Oktoeberza, I. Ardiyanto, R.L.B. Buana, and M. B. Sasongko, "Automated Segmentation of Hard Exudates Based on Matched Filtering," presented at the International Seminar on Sensors, Instrumentation,
Measurement and Metrology, 2016.
Intelligent Medical Imaging Research in Diabetic Retinopathy Recognition
Journals
[1] H. A. Nugroho, K. W. Oktoeberza, T. B. Adji, and F. Najamuddin, "Detection of Exudates on Color Fundus Images using Texture Based Feature Extraction," International Journal of Technology, vol. 6, p. 04, 2015. [2] H.A. Nugroho, D.A. Dharmawan, and L. Listyalina, "Automated Segmentation of Foveal Avascular Zone (FAZ) in Digital Colour Retinal Fundus Images," International journal of biomedical engineering and technology, 2016.
adinugroho@ugm.ac.id adinugroho@ugm.ac.id Intelligent Medical Imaging Research in Diabetic Retinopathy Team members and Collaborator
Rapid Assessment Diabetic Retinopathy and Intelligent System Research Groups Department of Electrical Engineering and Information Technology, Faculty of Engineering Universitas Gadjah Mada, Indonesia
(Hanung Adi Nugroho, Noor Akhmad Setiawan, Teguh Bharata Adji, Indriana Hidayah,
Igi Ardiyanto, Ratna Lestari Budiani Buana, Dhimas Arief Dharmawan, Latifah Listyalina, Dewi Purnamasari, Widhia Oktoeberza KZ)Department of Ophthalmology, Sardjito Hospital, Yogyakarta, Indonesia (dr. Muhammad Bayu Sasongko, dr. Kartika Dhani)
THANK YOU adinugroho@ugm.ac.id adinugroho@ugm.ac.id