dc.contributor.author |
Kassab, Hussein Amin. |
dc.date.accessioned |
2013-10-02T09:22:55Z |
dc.date.available |
2013-10-02T09:22:55Z |
dc.date.issued |
2013 |
dc.identifier.uri |
http://hdl.handle.net/10938/9574 |
dc.description |
Thesis (M.E.)--American University of Beirut, Department of Electrical and Computer Engineeering, 2013. |
dc.description |
Advisor : Dr. Mohamad Adnan Al-Alaoui, Professor, Electrical and Computer Engineering Department--Committee Members : Dr. Ali Chehab, Associate Professor, Electrical and Computer Engineering Department ; Dr. Haitham Akkary, Associate Professor, Electrical and Computer Engineering Department ; Dr Aghiad Al-Kutoubi, Professor and Chairperson, Department of Diagnostic Radiology. |
dc.description |
Includes bibliographical references (leaves 155-161) |
dc.description.abstract |
This thesis intended to demonstrate a program that gives an estimate of the human bone age by taking as an input the left hand x-ray image. At the beginning, the program automatically detects the hand profile by undergoing image processing techniques to the input image that include intensity transformation, filtering and morphological operators. After estimating the back ground, the program will automatically detect the hand and correct its orientation based on the middle finger. Many additions to previous literature were contributed, that include enhancement of the hand detection, hand alignment and POI localization. The developed program detects automatically the points of interests (POI) in x-ray images of the left hands by aligning each finger vertically. The POIs are mainly indicated on the Epiphyses-Metaphysis Region of Interest (EMROI). After that, the algorithm extracts the five fingers from the hand and each one of them is then isolated and vertically aligned to perform phalangeal zone extraction according to a specific edge detection algorithm canny. Then, features are extracted and are defined as the ratios of widths and distances. The extracted features are compared with standard figures and values in a database belonging to a normal child having the same status in terms of gender and chronological age. Then, they are classified according to Multi-layered Perceptron (MLP) Neural Network (NN) that was trained by the Levenberg-Marquardt back-propagation algorithm with Al-Alaoui Mean Square Error (MSE) classification method with cloning. Hence, bone age for each finger is determined. Then, each finger is given a weight and based on a “maximum occurrence” voting process the final bone age is found. The program was tested on different image groups that include different ages and both genders and gave excellent results. |
dc.format.extent |
xv, 161 leaves : ill. ; 30 cm. |
dc.language.iso |
eng |
dc.relation.ispartof |
Theses, Dissertations, and Projects |
dc.subject.classification |
ET:005812 AUBNO |
dc.subject.lcsh |
Bones -- Growth. |
dc.subject.lcsh |
Hand -- Radiography. |
dc.subject.lcsh |
Skeletal maturity. |
dc.subject.lcsh |
Children -- Age determination. |
dc.subject.lcsh |
Human beings -- Age determination. |
dc.subject.lcsh |
Bones -- Radiography. |
dc.subject.lcsh |
Image processing. |
dc.subject.lcsh |
X-ray optics. |
dc.subject.lcsh |
Neural networks (Computer science). |
dc.title |
Automatic bone age assessment using hand x-ray images |
dc.type |
Thesis |
dc.contributor.department |
American University of Beirut. Faculty of Engineering and Architecture. Department of Electrical and Computer Engineering. |