Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is a valuable modality for evaluating breast abnormalities found in mammography and performing early disease detection in high-risk patients. However, images produced by various MRI scanners (e.g., GE Healthcare & Siemens) differ in terms of intensity and other image characteristics such as noise distribution. This is a challenge both for the evaluation of images by radiologists and for the computational analysis of images using radiomics or deep learning. For example, an algorithm trained on a set of images acquired by one MRI scanner may perform poorly on a dataset produced by a different scanner. Therefore, there is an urgent need for image harmonization. Traditional image to image translation algorithms can be used to solve this problem, but they require paired data (i.e. the same object imaged using different scanners). In this study, we utilize a deep learning algorithm that uses unpaired data to solve this problem through a bi-directional translation between MRI images. The proposed method is based on a cycle-consistent adversarial network (CycleGAN) that uses two generator-discriminator pairs. The original CycleGAN struggles in preserving the structure (i.e. breast tissue characteristics and shape) during the translation. To overcome this, we modified the discriminator architecture and forced the penalization based on the structure at the scale of smaller patches. This allows the network to focus more on features pertaining to breast tissue. The results demonstrate that the transformed images are visually realistic, preserve the structure and harmonize intensity across images from different scanners.
1 September 2021 in Computer Methods and Programs in Biomedicine
Production of gestures does not require vision. Everyone—even those who are blind since birth—produces hand gestures while interacting with others. In our previous study, a gesture-based technique called Dactylology has been developed for the visually impaired to help them interact with computers. Cognitive load is an important indicator and a critical research issue while designing and adopting such new techniques. Hence, a study was conducted to compare the performance on a task between two computer input techniques (i.e., Dactylology and Braille) under varying levels of cognitive load (low, medium, and high), introduced by manipulating the task complexity. For the purpose of the study, 14 visually impaired participants were trained on Dactylology and Braille techniques to interact with the computer. The task performance was measured through the response time (RT) and false responses (FR). The results confirm that the participants had significantly lower RT and committed fewer FR in the Dactylology technique than the Braille under all cognitive load conditions. Altogether, these results render sufficient support to consider gesture-based Dactylology as a potential technique for the visually impaired to interact with the computer.
25 August 2017 in IEEE Transactions on Human-Machine Systems
Although a lot of work has been performed in the gesture-based human–computer interface, blind users still feel it is difficult to interact with computers. One of the major stumbling blocks is the lack of knowledge about their preferences toward hand gestures. A user evaluation study is conducted with 25 blind users to understand this fact and an optimal gesture set is devised. These gestures are selected based on the performance and preference measure analysis. Performance measure includes rating of gestures on four subjective criteria: easiness, naturalness, learning, and reproducibility. In preference measure, a new parameter called as preference index is proposed in this paper. The optimal gestures are further categorized into two groups: tier-1 and tier-2. On the basis of these gestures, a dactylology is proposed for blind users so that they can interact with a computer easily. A prototype model of the proposed interactive system has been developed and encouraging experimental results are obtained.