Digitaalinen kontrastin parantaminen pehmytkudosten kuvantamisessa

Innomentarium on ollut toimeksiantajana mukana tutkimuksessa, jossa on selvitetty digitaalisen kontrastin parantamisen vaikutusta kuvanlaatuun pehmytkudosten kuvantamisessa. Lopputuloksena on valmistunut pro gradu -tutkielma aiheesta.

Onnittelut Sherille valmistumisen johdosta!

Pro gradu -tutkielman tiivistelmä luettavissa alla. Koko tutkielma saatavilla Oulun yliopiston kirjastossa.

Author: Shahrzad Behrouzi
Digital Contrast Enhancement for Soft Tissue Imaging
Degree Programme in Biomedical Engineering


Breast cancer remains a significant cause of mortality among women, necessitating effective strategies for early diagnosis and successful treatment. Mammogram imaging, through the use of low-energy X-rays, has proven to be the most reliable and established method for the early detection of breast cancer. Lesion detection, including characteristic masses and microcalcifications, plays a crucial role in the diagnosis process. Furthermore, interoperative specimen imaging is essential to determine whether patients may require a second surgery due to cancer proximity to resection margins. This research aims to investigate methods to increase the quality of soft tissue imaging because detail is crucial in soft tissue imaging. The study utilizes phantoms and open-source data, employing varying image capturing setups for phantoms.The primary objective is to explore and evaluate techniques for enhancing image contrast, which is critical for accurate diagnosis. This thesis tests and evaluates methods from the following prevalent categories: Histogram Equalization (HE), Brightness Preserving Bi-histogram Equalization (BBHE), Contrast-Limited Adaptive Histogram Equalization (CLAHE), dynamic histogram equalization (DHS), and recursive mean separate histogram equalization (RMSHE). Python is used for image processing throughout research. The thesis further aims to optimize contrast, as perceived by the human visual system and in extent make assessments based on the just-noticeable-difference (JND) transform. The performance of different models using metrics which are entropy ratio, peak signal-to-noise ratio (PSNR), and the structural similarity index measure (SSIM). The results show better performance of the CLAHE-static JND and DHE methods compared with other methods in terms of high PSNR and SSIM. The overall objective is to employ sophisticated methodologies for processing and analyzing raw image data to enhance the contrast of soft tissue imaging. For future work, the technique will be evaluated with deep learning methods.