Pada mulanya, imej perubatan dianalisa secara analog iaitu imej dicetak di atas filem. Kebanyakan imej ini adalah di dalam bentuk hitam dan putih, serta amat sukar untuk dianalisa kerana mata manusia tidak dapat membezakan imej kelabu sebaik gambar berwarna. Apabila pemprosesan imej digital pertama kali digunakan untuk menganalisa imej-imej perubatan, sistem yang digunakan memerlukan ruang storan yang luas. Oleh kerana itu, projek ini memfokuskan pemprosesan imej digital untuk imej- imej perubatan. Algoritma pewarnaan palsu telah diaplikasikan ke atas imej perubatan untuk memudahkan
penganalisaan oleh doktor. Dua teknik pewarnaan palsu telah dibangunkan iaitu teknik pewarnaan palsu manual menggunakan kaedah ambang konvensional dan teknik pewarnaan palsu automatik menggunakan kaedah penentuan ambang automatik. Algoritma pengelompokan Purata-K dan Purata K Boleh Gerak digunakan untuk menentukan nilai ambang secara automatik dan seterusnya mengelompokan imej kepada beberapa kelompok sebelum algoritma pewarnaan palsu diaplikasikan. Selepas algoritma pewarnaan palsu dibangunkan, algoritma tersebut diimplementasikan kepada Pemproses Isyarat Digital TMS320C6416T supaya ia besifat mudahalih tanpa memerlukan komputer peribadi. Proses pengoptimuman TMS320C6416T dilakukan bertujuan mengoptimumkan prestasinya. Pengoptimuman ini melibatkan proses
pemilihan pengoptimuman atucara pemprograman dan juga modifikasi pada struktur algoritma yang dibangunkan. Pengujian ke atas beberapa jenis imej perubatan menunjukan bahawa algoritma Purata-K Boleh Gerak adalah lebih sesuai diapikasikan untuk pewarnaan palsu automatik. Proses pengoptimuman pula berjaya mengurangkan masa pelaksanaan sebanyak separuh. Daripada hasil keputusan yang diperolehi, jelas terbukti bahawa penyelidikan ini telah berjaya dilakukan dan sistem yang dibangunkan dapat menukarkan imej perubatan kepada gambar pewarnaan palsu yang setara.
_______________________________________________________________________________________________________
Before the advancement of digital image processing, medical images were analyze in analogue where the images were printed on films. The images were mostly in black and white and were hard to be analyzed because human eyes
cannot differentiate greyscale image as good as the colour images. When digital image processing was first used to analyze medical images, the system was bulky and often consumed large space. Thus, this project was undertaken to focus on the digital image processing for the medical images. Pseudo colouring algorithm was developed and applied to the medical images for easier analysis by medical doctors. Two pseudo-colouring algorithms were developed which were the conventional method (i.e. the user needs to input the threshold value manually) and the second algorithm was the automatic threshold
adjustment (i.e. the threshold value was automatically calculated for the images). The automatic pseudo-colouring algorithm used two clustering algorithms which were the K-Mean and the Moving K-Mean algorithms. The clustering algorithm clustered the images into several clusters before the pseudo-colouring algorithm was applied to the images. After the pseudo
colouring algorithm was created, the algorithm was implemented onto the TMS320C6416T Digital Signal Processor so that it can be mobile and portable without the needs of a personal computer. When the implementation was completed, the TMS320C6416T target board was then optimized such that it can operate or process the image at its optimum performance. The performance of the TMS320C6416T was based on the execution time taken for it to process the medical images. Higher performances mean that the time required to process the images was shorter. The optimization included compiler's optimization options and also the structure of the algorithm being created. After the algorithm was implemented and optimized on the TMS320C6416T, several medical images were tested using the developed system. Results from the developed system showed that Moving K-Mean algorithm was more suitable for automatic pseudo-colouring. The clustering
process was more efficient where it can cluster better than the K-Mean. Optimization was proven that it can increase the performance of the TMS320C6416T processing. The execution time after the optimization process was reduced by half. Other than that, three medical images which were the mammogram, pap-smear and rat sperm images were successfully converted to its equivalent pseudo-colour images using three level of pseudo-colour algorithm designed specifically for each images. From the result, it was proven that the research was successfully carried-out and the developed system can convert medical images to its equivalent pseudo-colour images.
Dsp-based pseudo colouring system for medical images / Mohd Naim Mohd Jain Noordin