computer vision utk deteksi ost

Upload: ardhiansyah-baskara

Post on 03-Jun-2018

230 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    1/32

    Deteksi Osteoporosis BerbantukanKomputer dengan Dental Panoramic

    Radiographs

    Agus Zainal Arifin

    Fakultas Teknologi Informasi, ITS, Surabaya

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    2/32

    2

    CAD

    Trend CAD (Computer Aided Diagnosis) Tujuan bukan komputer menggantikan manusia (dokter) Tujuan agar komputer (model komputasi yang

    dihasilkan) dapat meningkatkan akurasi dan mengurangierror analisa manusia.

    Perbedaan Performa Ada perbedaan performa antara radiologist senior vs

    dokter junior. Tetapi saat dokter junior memakai CAD sebagai alat

    bantu, tingkat performanya naik signifikan, menyamai

    level dokter senior. T. Kobayashi, et.all Effect of a computer-aided

    diagnosis scheme on radiologists performance,Radiology (1996).

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    3/32

    3

    Osteoporosis

    Loss of integrity and strength of the bone one of the most common disorders

    substantial morbidity rates

    increased medical cost

    high mortality risk in the elderly

    Bone Density & Bone Quality

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    4/324

    Osteoporosis Prevalence

    Di Indonesia, hasil analisa Depkes di 14propinsi, penderita osteoporosis sekitar19,7 persen dari jumlah lansia yang ada

    (IDIOnline.org, 2005). Tanggal 20 September hingga 20 Oktober

    2005 Menteri Kesehatan RI

    mencanangkannya sebagai BulanOsteoporosis Nasional.

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    5/325

    Deteksi Dini

    Penanganan tepat

    Bila berhasil terdeteksi secara dini, makadapat diberi penanganan medis yang sesuai.

    Metode Deteksi

    Dewasa ini

    Pengukuran BMD (Bone Mineral Density) dengan

    DEXA Scanner Metode Usulan

    Identifikasi dengan dental panoramic radiographs

    Dilakukan oleh Dokter Gigi

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    6/326

    BMD pada

    Lumbar Spine

    Dan Femoral Neck

    Pengukuran Bone Mass Density DEXA Scanner

    Dual Energy X-ray Absorptiometry

    Scanning BMD

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    7/327

    7

    The Proposed Early Detection

    Using Dental Panoramic Radiographs

    Frequently visit to dentists for treatment anddental diseases

    Dentists may use it easily for diagnosing otherdeseases Osteoporosis ?

    Normal ?

    Send to

    MedicalProfessional

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    8/32

    8

    Sistem Deteksi Dini

    PengolahanCitra Digital

    EkstraksiFitur

    PemilihanFitur

    DisainKlasifikasi

    EvaluasiSistem

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    9/32

    9

    Existing System

    Saat ini telah kami kembangkan Computer-aided diagnosis dengan mengukur

    Cortical width

    Cortical shape Kepadatan Trabecular bone

    Klasifikasi

    Fuzzy Neural Network, Algoritma genetika, dll

    Input: parameter yang berhasil diukur

    Output: Normal / Osteoporosis

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    10/32

    10

    10

    Cortical Width Measurement

    width ?

    Left and right of lower mandible around mental foramen

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    11/32

    11

    11

    Cortical Shape Analysis

    Cortical

    shape ?

    1. Normal 2. Osteopenia 3. Osteoporosis

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    12/32

    12

    Trabecular Bone Analysis

    Direction (horizontal and vertical), density, branch number, etc.

    Trabecular

    Bone

    Tujuan akhirnya mampu mendeteksi secaratepat inferior contour pada trabeculae

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    13/32

    13

    Original Results

    Line Detection Results

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    14/32

    14

    Line Detection Results

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    15/32

    15

    Pengguna

    Deteksi GarisTransformasi ke

    citra hitam putih

    Pengguna

    4 ROI

    4 Citra deteksi garis

    Jumlah Struktur

    segmen garis

    Rata-rata jumlah

    struktur segmen

    garis semua area

    sampel

    4 citra black and white

    Luas tiap citra

    diagnosa

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    16/32

    16

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    17/32

    17

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    18/32

    18

    Bone Mineral Density

    No. UmurHasil DEXA Scanner

    Lumbar Spine Femoral Neck

    1 84 0.593 (3) 0.496 (3)

    2 57 1.311 (1) 0.999 (1)3 67 0.755 (3) 0.723 (1)

    100 56 0.816 (2) 0.761 (1)

    Arti Index :1. Normal2. Ostepenia3. Osteoporosis

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    19/32

    19

    Rencana Pengembangan

    Identifikasi Parameter Mandible Cortical bone sepanjang mandible

    Analisa kepadatan trabecular bone

    Lokasi mental foramen

    Tinggi crown, akar gigi, dan lain-lain Standarisasi nilai-nilai parameter

    Berdasarkan data medis orang Indonesia asli.

    Optimalisasi Sistem

    Membangun klasifikasi yang lebih robust dan cepat. Integrasi seluruh bagian sistem. Sistem harus simple,

    cepat, dan akurat, agar dokter gigi tidak berkeberatanmenggunakannya diluar tugas utamanya

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    20/32

    20

    Remarks

    Dibandingkan pengukuran secara manual,pengukuran dengan sistem ini jauh lebih mudah,lebih cepat, dan lebih konsisten. Apalagi denganjumlah dental panoramic radiographs masif.

    Cutoff threshold sistem ini mungkin berbedadengan cutoff threshold cara manual, sebab adanyaperbedaan antara persepsi mata manusia dan hasilanalisa numerik.

    Dokter Gigi sangat dimungkinkan menggunakansistem CAD ini, karena mudah dan cepat.

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    21/32

    21

    Needs of Medical Images Processing

    Images were low contrast, unevenlyilluminated, and mainly dark

    Measurement by experts may not be

    consistent in the sense of intra-observerand inter-observer

    Need to an apparatus for Assessing

    massively dental panoramic radiographs General dental practitioners have more

    works

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    22/32

    2222

    Cortical Width Measurement

    l

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    23/32

    2323

    Assessment Results

    Manual vs OtomaticLumbar spine Femoral neck

    Normal Low BMD Normal Low BMD

    Manually measurement

    Cutoff threshold> Cutoff threshold 3045 232 2546 213

    Computer-aided system

    Cutoff threshold> Cutoff threshold

    3144

    223

    3140

    213

    Cutoff of manual 3.94 mm for lumbar spine and 3.91 for pada femoral neck.

    Cutoff of this system 3.08 mm for lumbar spine and 2.69 mm for femoral neck.

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    24/32

    2424

    Another Measurement

    Mandibular Width Measurement

    Binarization using Algorithm of

    Multistage adaptive thresholding

    Morphological Operation for

    Edge Detection

    Applying Genetic Algorithm for

    enhancing edges

    Euclidean distance of mandibular

    upper and lower boundary

    Algo ithms

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    25/32

    2525

    Algorithms

    Input ImagesThresholding and

    morpholgy processing GeneticAlgorothm

    Right

    Left

    C ti l Width M t

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    26/32

    2626

    Cortical Width Measurement

    width ?

    Left and right of lower mandible around mental foramen

    I Th h ldi i h F

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    27/32

    2727

    Optimization Of Ultrafuzziness BasedOn Fuzzy Sets Type II

    Image Thresholding with Fuzzy

    Image Fuzzification

    Determine optimal

    threshold

    Calculate ultrafuzziness

    Of the image

    Thresholding the image

    using threshold value

    Input Image

    Thresholded Image

    I Th h ldi ith F

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    28/32

    28

    Image Thresholding with Fuzzy

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    29/32

    2929

    Segmentationof

    Mandibular Bone

    T t l I f ti

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    30/32

    3030

    Dental Panoramic Radiograph Region of Interest 256 x 256

    Textural Information

    T t l I f ti

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    31/32

    3131

    Entropy

    Uniformity

    Contrast

    Homogenity

    SumMean

    Variance

    Correlation

    MaxProbability

    InverseDifferentMoment

    ClusterTendency

    M

    i

    N

    j

    jiPjiP ],[log],[

    M

    i

    N

    j

    jiP ],[2

    M

    i

    N

    j

    jiPji ],[)( 2

    M

    i

    N

    j ji

    jiP

    ||1

    ],[

    M

    i

    N

    j

    jijPjiiP ]),[],[(

    2

    1

    M

    i

    N

    j

    jiPjjiPi ]),[)(],[)((2

    1 22

    M

    i

    N

    j

    jiPji 2 ],[))((

    ],[,

    ,jiPMax

    NM

    ji

    jiji

    jiPM

    i

    N

    jk

    ||

    ],[

    M

    i

    N

    j

    kjiPji ],[)2(

    Textural Information

    Textural Information

  • 8/12/2019 Computer Vision Utk Deteksi Ost

    32/32

    3232

    Textural Information