logo

:::Location:Home > Research

 Research

Artificial, augmented, and human intelligence for medical image analysis for our health

Artificial, augmented, and human intelligence for medical image analysis for our health
We all aspire to be healthy. Can artificial intelligence (AI) help us in our quest?
 
   Thanks to ubiquitously available datasets, fast-growing computing power, and the latest innovative algorithms, AI continues to surprise us by changing the world at a rapid pace. AI unmanned vehicles have been around for some time. The FDA approved the first AI-powered diabetic retinopathy diagnostic device that does not require a doctor's supervision in April 2018. Google recently claimed that its AI voice technology passed the Turing test in making salon reservations in May. Therefore, can AI do just about anything? Maybe not, but we think AI can lend us a big hand in staying healthy by enhancing medical image analysis.
 
   Nearly all stages of medical care take advantage of medical images. Various medical imaging modalities, such as magnetic resonance imaging (MRI), computed tomography (CT), pathological specimens, and neuroimaging markers, are standard and essential to clinicians. However, these methods can be time consuming and sometimes their results are not interpreted effectively. We investigate how AI can tackle the challenges associated with medical image analysis.
 
   To push the frontier of medical care, such as the search for a cure for cancer and psychiatric disorder phenotyping, we try to answer the following two questions. (i) How can the accuracy, efficiency, and robustness of medical image analysis be improved by developing novel AI software tools? (ii) How can these developed techniques be applied to deploy innovative and effective medical workflows to assist clinical practitioners in medical diagnosis, treatment, and planning?
 
   Regarding the first question, we are developing a modularized AI Engine for performing medical image analysis tasks by AI methods. This AI Engine consists of theoretically and numerically proven software toolboxes. The toolboxes include image processing, quantitative analysis, deep learning, machine learning, and high dimensional data analysis. This AI Engine is the result of interdisciplinary innovations involving medical domain knowledge, interpreters’ experience, deep learning and machine learning techniques, mathematical and geometrical insights, statistical analysis, and high-performance computational methods. This novel AI Engine will act as an enabling technology for medical workflows, as described next.
 
   Regarding the second question, we are applying the toolboxes developed in the AI Engine to deploy standardized Augmented Intelligence Workflows (AI Workflows) in clinical settings. For example, four AI Workflows are dedicated to precision cancer treatments for lung, hypopharyngeal, hepatocellular carcinoma, and prostate cancers. AI Workflows are also dedicated to pancreatic mass classification and detection. Another AI Workflow addresses radiotherapy treatment planning in lung cancer. By standardizing data collection, processing, and interpretation, the AI Workflows can improve physicians’ therapeutic decision process with niches in large-scale quantitatively based computer-aided diagnosis and decision making.
 
   Combining the AI Engine and AI Workflows, we are building an Artificial Intelligence for Medical Image Analysis Platform (AIMIA Platform). The platform is a complete solution for medical image analysis and connects international and interdisciplinary researchers and professionals to foster academic excellence and commercial clinical applications. We anticipate that the AIMIA Platform will boost human-machine intelligence in medical image analysis and medical care. That is, we study how to use machine intelligence to advance human intelligence in medical care. The platform will assist physicians in making creative discoveries with insightful biomedical innovations.
 
   We want to emphasize that the development of the AIMIA Platform is rooted in an experienced and closely connected international and interdisciplinary team. The interdisciplinary background of the team members provides multiple methodologies that are new in medical image analysis. The international connections further allow us to meet people and exchange ideas to be well informed and inspired by broad viewpoints. International collaborations are critical for medical solution verification and promotion globally.
 
   Last but not least, one essential goal of our R&D efforts is education. We want to invite students to work together as a team and connect students with internationally renowned researchers and professionals to foster academic excellence and commercial clinical applications.
 
   In short, the AI Engine can act as a prospective tool for conducting highly efficient, accurate, and robust medical image analysis. The AI Workflows can enhance cognitive processes by helping clinical practitioners make better decisions and perform more effective actions. Combining AI Engine and AI Workflows can lead to AI-boosted human intelligence via creative discoveries made with insightful biomedical innovations.
 
 
References
1. Lin, C., Tsai, Y. M., Wang, W., Chen, L. (2018). GPU-accelerated high-resolution image stitching with better initial guess. ICCE 2018, 1-3.
2. Tsai, Y., Huang, H., Chou, C., Wang, W., and Hsiao, I.* (2015). Effective Anatomical Priors for Emission Tomographic Reconstruction. Journal of Medical and Biological Engineering. 35(1), 52-61.
3. Chou, C., Dong, Y., Hung, Y., Kao, Y., and Wang, W.*, Chien-Min Kao, and Chin-Tu Chen (2012). Accelerating Image Reconstruction in Dual-Head PET System by GPU and Symmetry Properties. PLOS ONE, 7(12): e50540. DOI:10.1371/journal.pone.0050540
4. Huang, C., Wang, W., Tzen, K., Lin, W., and Chou, C. (2012). FDOPA kinetics analysis in PET images for Parkinson's disease diagnosis by use of particle swarm optimization. IEEE International Symposium on Biomedical Imaging (ISBI) 2012.
5. Chou, C., Chuo, Y., Hung, Y., and Wang, W.* (2011). A Fast Forward Projection Using Multithreads for Multirays on GPUs in Medical Image Reconstruction. Medical Physics, 38(7), 4052-4065.
 
Weichung Wang
Professor, Institute of Applied Mathematical Sciences