The Medical Imaging & Image Analysis Laboratory
Department of Biomedical Engineering, The George Washington University
On February 19, Nada Kamona, a recent BME M.S. graduate from our lab, received an Honorable Mention for her poster “Automated detection of simulated motion blur in mammograms” at SPIE Medical Imaging, Houston. The poster was based on her M.S. thesis, supervised by Dr. Loew. Congratulations!
Our students, Shuyue (Frank) Guan and Ange Lou, have attended the 48th Annual IEEE AIPR 2019: Cognition, Collaboration, and Cloud in Washington DC, October 15th–17th. The Applied Imagery Pattern Recognition (AIPR) workshop sponsored by IEEE is to bring together researchers from government, industry, and academia across a broad range of disciplines. The 2019 IEEE AIPR Workshop explored cognitive applications of vision, dynamic scene understanding, machine learning, the associated supporting applications, and the system engineering to support the dynamic workflows. Shuyue gave a presentation about the evaluation of Generative Adversarial Network (GAN) Performance and Ange presented his work about segmentation of infrared breast images using Deep Neural Networks. A brief summary of Shuyue’s project and presentation: Recently, a number of papers address the theory and applications of the Generative Adversarial Network (GAN) in various fields of image processing. Fewer studies, however, have directly evaluated GAN outputs. Those that have been conducted focused on using classification performance and statistical metrics. In this paper, we consider a fundamental way to evaluate GANs by directly analyzing the images they generate, instead of using them as inputs to other classifiers. We consider an ideal GAN according to three aspects: 1) Creativity: non-duplication of the real images. 2) Inheritance: generated images should have the same style, which retains key features of the real images. 3) Diversity: generated images are different from each other. Based on the three aspects, we have designed the Creativity-Inheritance-Diversity (CID) index to evaluate GAN performance. We compared our proposed measures with three commonly used GAN evaluation methods: Inception Score (IS), Fréchet Inception Distance (FID) and 1-Nearest Neighbor classifier (1NNC). In addition, we discussed how the evaluation could help us deepen our understanding of GANs and improve their performance. Check the poster for more information of this project. A brief summary of Ange’s project and presentation: Convolutional and deconvolutional neural networks (C-DCNN) model had been applied to automatically segment breast areas in breast IR images in our previous studies. In this study, we applied a state-of-the-art deep learning segmentation model: MultiResUnet, which consists of an encoder part to capture features and a decoder part for precise localization. We also expanded our database to include 450 images, acquired from 14 patients and 16 volunteers. We used a thresholding method to remove interference in the raw images and remapped them from the original 16-bit to 8- bit, and then cropped and segmented the 8-bit images manually. Experiments using leave-one-out cross-validation (LOOCV) and comparison with the ground-truth images by using Tanimoto similarity show that the average accuracy of MultiResUnet is 91.47%, which is about 2% higher than that of the C-DCNN model. The better accuracy shows that MultiResUnet offers a better approach to perform the infrared breast segmentation task than our previous model.
The American Society for Radiation Oncology (ASTRO) Annual Conference was held from September 15th to 18th in Chicago. Our alum, Apurva Singh, presented her project in an oral presentation, ”A Novel imaging- genomic approach to predict the outcomes of radiation therapy”. It describes a portion of her thesis work related to developing a prediction system based on the image-genome analysis of pre-treatment PET images. The motivation of her work is to determine, whether information present in the heterogeneity of tumor regions in the pre-treatment PET scans of patients and in their gene mutation status can predict the efficacy of radiation therapy in their treatment. Apurva’s thesis project was done under Dr. Loew’s supervision.
Nada Kamona, our recent graduate student, and Dr. Loew attended the Medical Imaging Perception Society (MIPS) XVIII conference on July 14th-17th at Salt Lake City, Utah. Nada presented her thesis project in a 20-minutes oral talk on automatic detection approaches for simulated motion blur in mammograms. Scientists, physicians, radiologists, and students from around the world had attended the conference, and presented research in human and computer perception of medical image information. Nada was also awarded the MIPS XVIII Student Award Scholarship, which helped cover travel and registration expenses.
The Medical Image Perception Conference is a biennial conference dedicated to bringing together people interested in human and computer perception of medical image information and related subjects such as, detection and discrimination of abnormalities, cognitive and psychophysical processes, perception errors, and search patterns. This year it was hosted by the University of Utah.
For more information about MIPS, click here.
Three of our masters students have successfully defended their masters’ theses this semester, in partial fulfillment of the requirements for the degree of Master of Science. We congratulate them all on their accomplishments and hard work!
Kristina Landino, a student in BME, presented her work on salience in mammograms where she compared various salience detection algorithms in mammograms. Her committee members were Dr. Emilia Entcheva and Dr. Vesna Zderic.
Nada Kamona, a student in the 5-year BS/MS BME program, presented her project on automatic detection of simulated motion blur in mammograms, where she simulated blur mathematically to mimic real blur and developed methods to automatically quantify the blur in the images. Her committee members were Dr. Jason Zara and Dr. Emilia Entcheva.
Lastly, Apurva Singh in ECE, presented her work on tumor heterogeneity and gene mutation. This is a combined study for analysis of radiation therapy efficacy in head-and-neck carcinoma patients. The analysis was also extended to lung and cervical cancer patients. Her committee members were Dr. Milos Doroslovacki and Dr. Kie-Bum Eom.
Apurva Singh
Nada Kamona
Kristina Landino
