Undergraduate / The essays for a MIT fellowship in the field medical image processing [4]

The essays for a MIT fellowship in the field medical image processing

Dear All:

May I ask if you can help me on these essays. I'm not a native English speaker. Hopefully I didn't do ant thing extremely stupid. Thanks.

Please respond to the following 3 essay questions. Please put your answers in one file, with the response to each starting on a separate page.

Question 1: Please describe a time in the recent past when you went beyond what was defined, expected, established, or popular (max 500 words).
Histo-pathological examination is a powerful method for the prognosis of critical diseases. But, despite significant advances in high-speed and high-resolution scanning devices or in virtual exploration capabilities, the clinical analysis of whole slide images (WSI) largely remains the works of medical doctors. Thus their workload is usually heavy. For example, for a pathologist, one slide costs about 10 seconds to 10 minutes (depends on his/her experiences). However, in a large-scale hospital, he/she usually needs to screen several hundred slides per working day.

In 2008, I joined a project, in which the goal was to go beyond the state-of-the-art in multimedia information retrieval in a medical information system. The proposed system can provide more efficient and precise methods for accessing medical information. As a result, the work load of the medical doctors can be reduced.

In the results of the project, I proposed the world's first region of interest (ROI) detection algorithm for carcinoma specimen using microscope. We demonstrated the capabilities of the algorithm on the diagnosis and prognosis of breast cancer. In theoretical speaking, the algorithm can be applied on various diseases as long as the training patterns are well defined. Eventually, we proposed an innovative platform, in which multi-scale computer vision algorithms perform fast analysis of a histo-pathological WSI. It can support and improve the performance of pathologists. In the project, we demonstrated the capabilities of the system on the support of breast cancer grading and its prognosis.

During the development, the very first problem that we met was that since the amount of the data and image was extremely huge, thus the processing time was incredibly long. For instance, our first version needed 72-96 hours for a slide by using a cutting-edge computer.

Thus, I decided to look into the problem on a higher level. I started from understanding how does the human visual system play its role while a pathologist is diagnosing a case. I introduced some concepts from an emerging research field of artificial intelligences and neural networks, called bio-inspired computer vision (BioCV). Based on the BioCV, I was able to mimic some behaviors of the human visual system. Meanwhile, the proposed system has been developed by using graphic processing units (GPUs), as a result, the proposed system is able to identify the ROI of a slide less than 100 seconds per slide (by using a cutting-edge computer, it can achieve 30 seconds per slide), while the accurate rate remains good (91.83% in average).

This platform allows medical professionals to efficiently locate the information that they seek for diagnosis, medical research, and teaching. In the information age, where information is the most valuable resource, the system can help to improve the quality and reduce the cost of health care and medical research and education.

This system is promising, and it is potentially high impact. The result has been published on the journal of computerized medical images and graphics (CMIG) in 2011.

 
Question 2: Please describe a time in the recent past when you took responsibility for achieving an objective (max 500 words).
Neural Stem Cells (NSCs) have received much attention in cell-transplantation therapy for central nervous disorders such as Parkinson's and Alzheimer's diseases. The key concept is using the NSCs to regenerate the neurons. In this field, the study on the epidermal growth factors (EGFs) has been highlighted. Some studies suggested that EGF is one of the major mitogens for immortalized neural progenitors.

Since 2010, I have joined a project which is focusing on searching the EGFs of the NSCs for various purposes, such as the treatment of these diseases. In the beginning of the project, a major problem was the dynamic appearance of the suspension neurospheres. That is, in order to observe the status of a neurosphere, culturing the cells in suspension status (in contrast to adherent status) is necessary. However, the nature of suspension neurospheres result difficulty on monitoring, tracking, and acquiring the information from them. More precisely, a suspension neurosphere is not attached on the plate; they can move, rotate and spin rapidly. Meanwhile, most of the bio-markers are prevented in our experiments due to the phototoxicity. As a result, to the best of our knowledge, most of the automatic microscopes and microscopic technologies are not available for this situation.

I was taking responsibility of solving this problem. First, I analyzed the requirement: in our experimental environment, the cycle of the observing cells is roughly about 22-26 hours, which means that in order to observe the doubling procedure of a batch of cells (usually a batch is about 100 cells), the human operators need to relocate those cells in every 15-30 minutes over at least 2 days. Sometimes it is even longer than weeks. As a result, these kinds of operations are usually labor intensive. In the most of the time, the performance is usually not well.

In order to tackle this problem, I dedicated myself to the collaborative laboratory and worked on the biological experiments by myself, such that I can study and practice of using the microscope, which include the devices and skills that the biologists needed for the experiments in this project. The purpose is to obtain the first-hand experience. This experience is unique in my career since the biology practical skills are never being the keys that I need in the research.

In weeks, I had picked enough experience and fully understood how it is working in a lab. Based on my knowledge in computer vision and machine learning, I developed the world's first 3-d suspension cell tracking system for automatic monitoring and acquiring information through a long-term time-lapse image acquisition. Using the system, it is possible to obtain the image and information during the whole progress of forming neurospheres. Based on this system, our collaborative biologists can discover some things meaningful. This system also attracted the attentions from the industries and academia. At this moment we are working on the commercialization with an international company, and there is another commercializing plan in initializing state.

 
Question 3: In looking around at the lives of your family, friends, neighborhood, country, and/or world, please give one or two examples of what you would consider issues that need to be addressed via biomedical imaging research. Please comment on any imaging-based approaches that you think might help address those issues. Please refrain from listing huge global issues (e.g. ridding the world of infectious disease; eliminating obesity; curing cancer) and instead be more specific. (500 word limit)

As the medical and healthcare technologies are improving, the longevity of human is increasing. The incidence of some disorders increases with age in both humans and laboratory animals. Many diseases are highly related to the aging: for example, the neurodegenerative diseases and/or the cancers. A clear understanding of the causes of the age-related increase in these disorders incidence is needed to develop a strategy for primary disorder prevention.

The integration system of quantitative microscopy, machine learning and computer vision system could be an important tool for tackling these issues. Quantitative microscopy has been extensively used in biomedical research and has provided significant insights into structure and dynamics at the cell and tissue level. Meanwhile, machine learning and computer vision have been widely studied in the last decades. They are capable to mimic some capabilities of a human expert.

In the following, how can the suggested system help on the research of these disorders are discussed:

- Neurodegenerative diseases:
Recently, neural stem cells (NSCs) have received much attention in cell-transplantation therapy for or neurodegenerative diseases such as Parkinson's and Alzheimer's diseases. A major problem of using the NSCs as the treatment is that the fate of a NSC needs to be predicted. A NSC can produce progenitor cells, eventually they generate various types of cells, including neurons, astrocytes, and oligodendrocytes. In order to use NSCs as the treatment, we want to enlarge the population of the generated neurons. Our research suggested that the fate prediction of the NSCs is possible. It needed the supports of the suggested system. Although it is still far away from a clinic therapy, however our research indicated a promising future.

- Cancers:
Recently the research on telomeres has been highlighted for the investigation of the relationship between aging and cancers. Recent research suggested that via investigating the status of the telomeres, we might be able to predict the cancers. This kind of research requires the suggested system in some aspects: a practical example is evaluating the morphology of the telomeres. This is not an easy task since the scale of the telomeres is rather small. Laser confocal microscopy has been frequently used in order to observe the telomeres. But despite the advantages, the method is limited in the problem of resolution. For the sake of solving this problem, the suggested system might be helpful.

On the other hand, the relationship between aging and cancers is still unclear. In order to discover it, more investigations and more data are necessary. Hence, the suggested system might be able to play an important role in this research.

Finally, we conclude that the suggested integration system of quantitative microscopy, machine learning and computer vision system is not only for the mentioned disorders, but also potentially capable to support the research on various diseases from the aspect of medical/biomedical imaging, bioinformatics and image processing. It may not be able to show us the answers directly, but can be an important tool for revealing the secrets.