Image Analysis And What I Learned
Welcome back to my blog! This time, I will begin by talking about the next phase of my project: image analysis. After obtaining patient MRI scans through the Notion software last week, my goal this week was to begin examining them and identifying regions of interest.
It took a while to familiarize myself with ImageJ, and I encountered many problems along the way, but eventually I learned to use it. The scans are taken from the transverse plane, and scrolling through them reveals additional cross-sections in the z-axis, thus providing a 3-Dimensional understanding. Legions in the liver are sometimes difficult to locate, but they appear as a discolored and misshapen mass that appears and then disappears as you scroll through. Each patient contained also several different types of MRI Scans, with names like Apparent Diffusion Coefficient mm2 s Apparent Diffusion Coefficient mm2 s AX DualEcho FSPGR_op phase, AX DualEcho FSPGR_op phase, AX DualEcho FSPGR_in phase, AX DualEcho FSPGR_in phase, AX LAVA PRE, MULTI-B DWI RTr_b1000, MULTI-B DWI RTr_b500, Ph1 AX LAVA MPh, Ph2 AX LAVA MPh, Ph3 AX LAVA MPh. I'm just beginning to understand what they all mean.
Unfortunately, my mentor was busy for much of the week, so I could not progress further. However, I was able to learn much more about my project. In response to some of the comments I received from previous posts, I would also like to elaborate more on the newly identified type of HCC I am working with. It is a distinct histological subtype called SteatoHepatitic Hepatocellular Carcinoma (SH-HCC). The term steatohepatitic refers to the accumulation of liver fat in people who drink little or no alcohol (alcohol is a major cause of "normal" HCC). It has pathological features resembling large droplet steatosis, ballooning of malignant hepatocytes, Mallory-Denk bodies, inflammaton, and pericellular fibrosis.
However, although these features can be identified through surgical pathology on dead patients, it is much harder to do so through MRI examination, which is why all past MRI exams have completely missed the SH-HCC subtype. In an MRI scan, tissues are shaded dark and light depending on their density. Radiologists analyze the size and distribution of these bright and dark areas to non-invasively diagnose diseases real-time in living patients. Regions of interests containing legions associated with SH-HCC will have a unique shading, different from regular HCC, but the distinction is still unkown. This is why texture analysis and machine learning on the pixel image data will be used to establish distinguishing image metrics and discover correlations. If successful, we can apply this to future MRI scans so medical professionals will have the ability to diagnose and treat the specific SH-HCC subtype in patients.
It took a while to familiarize myself with ImageJ, and I encountered many problems along the way, but eventually I learned to use it. The scans are taken from the transverse plane, and scrolling through them reveals additional cross-sections in the z-axis, thus providing a 3-Dimensional understanding. Legions in the liver are sometimes difficult to locate, but they appear as a discolored and misshapen mass that appears and then disappears as you scroll through. Each patient contained also several different types of MRI Scans, with names like Apparent Diffusion Coefficient mm2 s Apparent Diffusion Coefficient mm2 s AX DualEcho FSPGR_op phase, AX DualEcho FSPGR_op phase, AX DualEcho FSPGR_in phase, AX DualEcho FSPGR_in phase, AX LAVA PRE, MULTI-B DWI RTr_b1000, MULTI-B DWI RTr_b500, Ph1 AX LAVA MPh, Ph2 AX LAVA MPh, Ph3 AX LAVA MPh. I'm just beginning to understand what they all mean.
Unfortunately, my mentor was busy for much of the week, so I could not progress further. However, I was able to learn much more about my project. In response to some of the comments I received from previous posts, I would also like to elaborate more on the newly identified type of HCC I am working with. It is a distinct histological subtype called SteatoHepatitic Hepatocellular Carcinoma (SH-HCC). The term steatohepatitic refers to the accumulation of liver fat in people who drink little or no alcohol (alcohol is a major cause of "normal" HCC). It has pathological features resembling large droplet steatosis, ballooning of malignant hepatocytes, Mallory-Denk bodies, inflammaton, and pericellular fibrosis.
However, although these features can be identified through surgical pathology on dead patients, it is much harder to do so through MRI examination, which is why all past MRI exams have completely missed the SH-HCC subtype. In an MRI scan, tissues are shaded dark and light depending on their density. Radiologists analyze the size and distribution of these bright and dark areas to non-invasively diagnose diseases real-time in living patients. Regions of interests containing legions associated with SH-HCC will have a unique shading, different from regular HCC, but the distinction is still unkown. This is why texture analysis and machine learning on the pixel image data will be used to establish distinguishing image metrics and discover correlations. If successful, we can apply this to future MRI scans so medical professionals will have the ability to diagnose and treat the specific SH-HCC subtype in patients.
Hi, Richard. Cool post. Do you think that eventually MRI machines can be equipped with the ability to recognize HCC and use that to adjust the scan?
ReplyDeleteThanks for posting! It seems difficult to analyze different types of MRI Scans. Do you know why it would be like this? To me, it seems to create additional work for the doctor if they had to analyze different MRI scans? How are they different from one another?
ReplyDeleteHi Richard! It's very impressive to see you beginning to understand all those different types of MRI scans! I hope to see some more about them in the weeks to come!
ReplyDeleteHi Richard! I am really enjoying your blog posts. I would appreciate you going into more detail so I can understand what you are doing. What are you looking forward to at this point? :)
ReplyDeleteHi Richard. Your blog posts are very compelling! when you say "discolored and misshapen mass", could you provide images? Thank you!
ReplyDeleteHi Richard! Your blog so far is really interesting; I really appreciate how you use fairly simple and straightforward wording to make it really clear how you accomplished various tasks, especially since it makes it much easier for people like me to understand what you're discussing! I'm excited to hear what happens next week!
ReplyDeleteHey, Richard. It's unfortunate you couldn't make much progress this week but it is much appreciated that you took the time to explain things for us. I am better able to wrap my heard around your project. Can't wait to see you back working on the project!
ReplyDeleteHi Richard! Thanks for elaborating on SH-HCC. I hope you make lots of progress in the weeks to come!
ReplyDeleteHi Richard, really interesting blog post, it's a shame that your mentor wasn't with you for much of this week, I look forward to more posts in the future.
ReplyDeleteHi Richard, sorry to hear that you didn't make as much progress as you wanted this week, but i am happy to hear about your collaboration with the SH-HCC.
ReplyDeleteHi Richard! Its great that you were able review some of the comments. However, it sucks that you didn't complete what you wanted to complete this weekend! What is the difference between the different MRIs.
ReplyDeleteI didn't know that there were this many nuances in the field of image analysis. Hopefully you can get what you wanted to get done soon.
ReplyDeleteThat's pretty cool. I've heard of ImageJ, but how does that tool help you in particular? Are you marking the regions that look like they have cancer for the computer to use?
ReplyDeleteThank you for elaborating on SH-HCC. I am sorry that you were not able to make much progress this week but thank you for further explaining your project. I can't wait to see your next post!
ReplyDelete