Segmentation of avascular necrosis of the femoral head using 3-D MR images
Introduction
Avascular necrosis (AVN) of bone is a process that is characterized pathologically by bone marrow ischemia and eventual death of trabecular bone. Ischemic necrosis, osteonecrosis, and aseptic necrosis are synonyms for the same disease process [1]. Several conditions are clearly related to avascular necrosis of the femoral head (ANFH), a disruption of the blood supply. These include hip fracture, alcohol abuse, marrow storage disease, and corticosteroids, used for immunosuppression in transplant patients as well as for asthma, rheumatoid arthritis, or connective tissue diseases. In nontraumatic ANFH, collapse is a crucial event that leads to functional devastation of the hip joint [2], [3]. MRI [4], [5] is the most sensitive non-invasive examination for detecting AVN [1]. AVN may be detected on MRI examination in every early stages [6], [7] as a low signal region within the femoral head due to loss of water-containing bone marrow. Hence, a prospective study of the prognostic value of MRI in early stages of the disease may allow accurate evaluation of ANFH, avoid surgical operations, and preserve hip function [8]. However, much of the current confusion and contradiction on the treatment of AVN are caused by the lack of an agreed efficient quantitative system for evaluation and staging [9]. Hence, finding criteria to assess the susceptibility for collapse or to predict impending collapse for each case before it occurs [10], [11], is an important aspect of the corresponding research [1], [2], [3], [6], [7], [8], [9], [10], [11].
Our ultimate goal in undertaking this research was to develop a computer assisted diagnostic (CAD) system for dealing with ANFH. In this regard, we aim to provide feasible, fast and robust software tools, for evaluation, quantification, visualization, and prediction of collapse of the femoral head. In this paper, we mainly focus on computerized segmentation and visualization of necrotic and non-necrotic lesions of the femoral head using MRI (T1-weighted) coronal data sets.
Computers have become an integral part in medical image acquisition, enhancement and analysis [12]. Computerized interpretation of medical images is problematic because of statistical, structural, and temporal variations of objects in such images. In particular, the segmentation problem for MRI data depends on a large number of factors including image contrast, resolution, signal-to-noise ratio, slice thickness, complexity of the scene, data set size, RF-coil uniformity, and magnetic susceptibility artifact [4], [5]. For these reasons, identical tissue types will give rise to varying intensities as a function of their spatial locations [13]. This is why developing functional automated systems for MR image analysis and interpretation has been a major challenge in the field of medical image processing.
Computerized techniques for quantification of bone microscopic structure [12], total hip arthroplasty using CT data [14], and bone segmentation in clinical knee MRI [15] are reported in the literature. In the case of ANFH, the previous research [1], [2], [3], [6], [7], [8], [9], [10], [11], [16] was mostly performed by manual segmentation of necrotic lesions of the femoral head. In addition, no user friendly environments were provided for 3-D visualization and assessment of the results. Our motivation for undertaking this research was to develop a new technique and also a free software tool for segmentation and visualization of ANFH. Since the Microsoft Windows operating system (OS) [17] is commonly available in most academic, clinical, and research centers, we implemented that computer software and corresponding graphical user interfaces (GUI) under the Windows environment. The developed algorithms are integrated in the MEDAL free software package and available on the Internet [37].
The remainder of the paper is organized as follows. Section 2 investigates the available data sets. Section 3 reviews the conventional segmentation techniques. 4 Segmentation of non-necrotic lesions of the femoral head, 5 Segmentation of necrotic lesions of the femoral head develop techniques for segmentation of non-necrotic and necrotic parts of the femoral head, respectively. Realization of the algorithms and evaluation of the obtained results are explained in Section 6. Summary of the work and future development are made in Section 7.
Section snippets
Data set
About 1500 coronal images of femoral heads from MRI/T1 scans were acquired from 30 adult patients.1 It was found that the quality and imaging conditions of the acquired images, exact orientation of imaging and
Review of segmentation techniques
Segmentation is the first essential and important step of early vision and image analysis. Computerized segmentation of ANFH is needed for the following quantitative assessments.
- •
Determining the geometrical specifications4 of necrotic lesions.
- •
Visualization of necrotic and non-necrotic lesions of the femoral head.
- •
Finding the relation between collapse and shape of necrotic lesions.
- •
Determining the risk of collapse of necrotic lesions.
- •
Introducing an appropriate
Segmentation of non-necrotic lesions of the femoral head
The proposed segmentation technique consists of the following steps.5
Segmentation of necrotic lesions of the femoral head
As discussed in Section 4, non-necrotic lesions of the femoral head were of high intensity values. In the case of normal bone tissues, the change in intensity is mostly due to the imperfect imaging condition of the MRI modality. However, for necrotic lesions of bone, in addition to non-ideal imaging conditions, the situation and stage of the disorder can dramatically affect the intensity value. That is, gray levels of necrotic tissues are often subject to change, data set by data set, slice by
Realization of the techniques
The techniques of 4 Segmentation of non-necrotic lesions of the femoral head, 5 Segmentation of necrotic lesions of the femoral head were developed in C language. The developed algorithms together with the necessary graphical user interfaces (GUIs) were included in a dynamic link library (DLL). This DLL was based on the Microsoft Foundation Class (MFC) and compatible with the Microsoft Visual C compiler environment [17]. The developed DLL for bone segmentation was added as a plug-in DLL to the
Summary
In this paper, we developed techniques and software for segmentation of avascular necrosis of the femoral head. The software is available on the Internet [37]. We experimentally found that the proposed segmentation techniques were effective for femoral heads with small necrotic regions. As seen in Table 1, these cases had the largest population in the available data sets (18 out of 60). When the necrotic region is large, the segmented non-necrotic lesion is of small size. Hence, after oblique
Acknowledgements
This work was supported by the Japan Society for the Promotion of Science (JSPS Research for the Future Program).
Reza Aghaeizadeh Zoroofi ([email protected]) was born in Rasht, Iran, in 1966. He received the BS degree in electronic engineering from Amirkabir University of Technology, Iran, the MS degree in telecommunication engineering from Khajeh Nassir University of Technology, Iran, and the PhD degree in medical science from Osaka University, Japan, in 1989, 1991, and 1998, respectively. From January to March 1997, he was a NEDO Researcher at Mechanical Engineering Laboratory, Ministry of
References (38)
- et al.
A review of image segmentation techniques
Pattern Recogn
(1993) - et al.
Segmented snake for contour detection
Pattern Recogn
(1998) - et al.
A fast ellipse-circle detector using geometric symmetry
Pattern Recogn
(1995) The robust algorithm for finding the center of an arc
Comput Vision Imag Unders
(1995)- ...
- et al.
Avascular necrosis of the femoral head
Bone Joint Surg (Br)
(1993) - et al.
Prognostication of non-traumatic avascular necrosis of the femoral head
Clin Orthop
(1994) - et al.
MRI: basic principles and applications
(1995) - et al.
Magnetic resonance imaging, theory and practice
(1996) - et al.
Quantitative magnetic resonance imaging predicts clinical outcome of core decompression for osteonecrosis of the femoral head
J Rheumatol
(1995)
Prognosis of early stage avascular necrosis of the femoral head
Clin Orthop
Does lesion size affect the outcome in avascular necrosis?
Clin Orthop
Quantifying the extent of osteonecrosis of the femoral head
J Bone Joint Surg (Br)
Estimation of the extent of osteonecrosis of the femoral head using MRI
J Bone Surg (Br)
Bone image segmentation
IEEE T Bio-med Eng
Automated segmentation of the brain in MRI
Proc Japan Assoc Med Imag Tech (JAMIT) Frontier, Tokyo
Combined acetabular and femoral navigation for resurfacing total hip arthroplasty
CARS
Cited by (23)
Segmentation of histology slides of cortical bone using pulse coupled neural networks optimized by particle-swarm optimization
2013, Computerized Medical Imaging and GraphicsCitation Excerpt :In addition, micro-segments of cortical bone helped the authors [8,9] in building a micro-FEM model which was developed to study the process of orthogonal cutting of bones used widely in orthopedic surgeries. A new computerized approach for segmentation of magnetic resonance images (MRI) of necrotic lesions of the femoral head of humans in [10,11] helped analyzing a vascular necrosis [10]. Automated bone image segmentation proved to be efficient in detecting microscopic bone features (e.g., osteons and Haversian canals) [12].
Automatic segmentation of magnetic resonance images using a decision tree with spatial information
2009, Computerized Medical Imaging and GraphicsImproving segmentation accuracy for magnetic resonance imaging using a boosted decision tree
2008, Journal of Neuroscience MethodsPersonalized MR-based musculoskeletal models compared to rescaled generic models in the presence of increased femoral anteversion: Effect on hip moment arm lengths
2008, Gait and PostureCitation Excerpt :MR based musculoskeletal models were built using a custom-built workflow [8]. In a first step, bone structures were segmented semi-automatically using a hybrid approach [12]. The segmented images were used to create tetrahedral meshes of the bone structures.
Automated segmentation of necrotic femoral head from 3D MR data
2004, Computerized Medical Imaging and GraphicsCitation Excerpt :Experimental results show the superiority of the proposed technique compared to the previous technique [29] in terms of accuracy, user interactions, and clinical usefulness. The developed algorithms are integrated in a software that is developed based on the MEDAL software [30]. The rest of this paper is organized as follows.
Segmentation of Drug-Treated Cell Image and Mitochondrial-Oxidative Stress Using Deep Convolutional Neural Network
2022, Oxidative Medicine and Cellular Longevity
Reza Aghaeizadeh Zoroofi ([email protected]) was born in Rasht, Iran, in 1966. He received the BS degree in electronic engineering from Amirkabir University of Technology, Iran, the MS degree in telecommunication engineering from Khajeh Nassir University of Technology, Iran, and the PhD degree in medical science from Osaka University, Japan, in 1989, 1991, and 1998, respectively. From January to March 1997, he was a NEDO Researcher at Mechanical Engineering Laboratory, Ministry of International Trade and Industry, Tsukuba, Japan. From April 1997 to March 1998, he was a Research Fellow at National Cardiovascular Center Research Institute, Osaka, Japan. From April 1998 to March 2000, he was a Japan Society for Promotion of Science (JSPS) Postdoctoral Fellow at Division of Functional Diagnostic Imaging, Osaka University Medical School. Dr Zoroofi is currently an Assistant Professor at Department of Electrical and Computer Engineering, Faculty of Engineering, University of Tehran, Iran. His research activities include works in the fields of image processing, medical imaging and medical information.
Takashi Nishii graduated and received his MD degree from Osaka University Medical School, Japan, in 1989. He is currently an Associate Professor of the Department of Orthopaedic Surgery of the Osaka University Medical School.
Yoshinobu Sato ([email protected]) was born in Kobe, Japan, in 1960. He received his BS, MS and PhD degrees in information and computer sciences from Osaka University, Japan, in 1982, 1984 and 1988, respectively. From 1988 to 1992, he was a Research Engineer at the NTT Human Interface Laboratories. In 1992, he joined the Division of Functional Diagnostic Imaging of Osaka University Medical School as a Faculty Member. From 1996 to 1997, he was a Research Fellow in the Surgical Planning Laboratory, Harvard Medical School and Brigham and Women's Hospital. He is currently an Associate Professor at Osaka University Medical School, where he leads a group conducting research on 3-D image research analysis and surgical navigation systems in the Division of Functional Diagnostic Imaging.
Nobuhiko Sugano graduated and received his MD degree from Osaka University Medical School, Japan, in 1985. He is currently an Associate Professor of the Department of Orthopaedic Surgery of the Osaka University Medical School.
Hideki Yoshikawa graduated and received his MD degree from Osaka University Medical School, Japan, in 1979. He is currently a Professor of the Department of Orthopaedic Surgery of the Osaka University Medical School.
Shinichi Tamura was born in Nishinomiyam, Japan, in 1944. He received his BS, MS and PhD degrees in electrical engineering from Osaka University, Japan, in 1966, 1968 and 1971, respectively. He is currently Professor of Medical School, and Graduate School of Engineering Science, of Osaka University. He is an author/co-author of over 180 papers in scientific journals and received several paper awards from journals/societies including Pattern Recognition and Investigative Radiology. His current research activities include works in the field of image processing and its medical applications. Dr Tamura is a member of the IEEE, International Society for Computer Aided Surgery, the Institute of Electronics, Information and Communication Engineers of Japan, the Information Processing Society of Japan, the Japanese Society of Medical Imaging Technology, and the Japan Radiological Society, etc. Currently he is an Associate Editor of Pattern Recognition, and a Vice Editor-in-Chief of Medical Imaging Technology.