Overall it is still a good course and worth doing but I won't expect to be able to clear a job interview in medical machine learning based on this course. It touches many nice topics such as what to do if data is unbalanced, different metrics about evaluating the models. However the part about MRI segmentation seems very rushed. I would consider this as a very basic course and the student would have to spend significant personal time exploring on his/her own to really understand the concepts presented in the class. It wasn't easy for me to get help on some programming assignments when I got stuck a. Moreover, when I didn't get a perfect score on the programming assignments, I don't know where I made the mistakes, which makes it impossible to correct them.

The course has been designed well, learnt new terminology which I was not aware of previously when working on 2D datasets. Good introduction to 3D images. The course could be a bit more detailed, for example, since data preprocessing is very crucial, it would have been great to have had an assignment on cleaning 3D data using image registration, alignment, etc. Additional references for reading mainly books would have been nice. Finally, brief details on the type of computing power and memory is required especially for 3D images would have been very helpful. If I run the code on my laptop, I am sure it will crash, would be nice to have an idea of the requirements. Anyways, thank you for the course, very nice introduction to AI in medical field.

As a Radiologist from the Philippines who has been desperately trying to find some kind of "grounded center" for all the AI/ML topics I've been studying online, this is a really great way to consolidate what I've learned so far especially for AI applied to Radiology. I've been training models for computer vision (based on free tutorials on-line) but this has definitely given me better insight as to how those models actually work and how they come together from simple numpy arrays, to tensors, layers, and finally into compiled models.... giving me a better appreciation for how activation functions and convolutions actually fit into the development of convolutional neural networks. More power to the team.

The course was interesting but did not make me feel ready to apply a DL model on such data. It'a like being in a sandbox all the time: you play, you see things, then you are required to build your own, little, insignificant castle with your little basket, but no more than that. I think that real problems in AI application in this field are not about calculating sensitivity, specificicity or standardazing data, things for whom there are already functions built in libraries. I feel I know more this job, but i wouldn't be ready if i didn't know it yet before.

I really enjoyed and learned a lot from the material in this course. The lectures were clear and concise. Short lectures made it easy to retain the material. Also helpful were non-graded exercises embedded with the lectures. The graded labs were correct and had helpful hints.

The only improvement that I would want is to have the discussion forums back on Coursera and not on Slack. I found it difficult to search for similar questions on Slack and frequently ran into a limit on the number of messages I could search through.

Overall an excellent course!

Overall, it is a good decision to take the course. Although it focuses on practical aspects of the AI in medicine, it falls short explaining the basic CNN architecture for image segmentation or classification. That said if you wish to fully take advantage of the course, spend some time understanding some of the key functions available in the util.py scripts which can be accessed through the notebooks. There, you could benefit from the course and learn interesting implementation stuff if you feel like the assignments are too practical.

Good Course

I find that it is always tougher to teach when the audience is heavily segmented.

I see this course audience as:

- Medical practitioner who want to learn about ML

- ML practitioner who want to apply ML in a specific context.

I am of the second group.

The course is at its best when the topic are the most general like:

- The importance of correctly preparing the test, validation and training sets.

- Understanding the meaning of model accuracy in the real world.

But the implementation specifics are a bit dated now (April 2021).

Thanks

The course is ok - after this course you will be ready for real tasks. but the course is not ideal: 1) you can not solve some tasks with different possible ways. As example in week 3 programming, you can not use np.empty, but you need to use np.zeros, cause another vay is incorrect. And the sub volume task - random crop loop with tries is no optimal way to solve it, but another way is incorrect. 2) I wanted to hear more about U-Net. 3) i think you need to report copies of your course on github

Overall good. The lectures are easy to follow, but the programming assignments (especially week 3) need clearer instructions. The automatic grader also needs improvement, as the grader not only false alarms in a correct function and fails to detect errors in another function, but also requires very specific implementation (you can't implement in a different way, and you can't miss any argument) , even though the function works well and correct.

This is an amazing course for people who know AI and want to know about it's applications in the healthcare industry. I had fun learning from the instructor Pranav who is concise and delivers lessons comprehensively. Overall an amazing course. Could have asked for more assignments and hands-on stuff, hence I'm being conservative on granting 4-stars only...

Very good course on applying AI for image-based medical diagnosis. Some things that could be improved are : 1. adding content relevant to using AI in non-image based diagnosis 2. could be made more comprehensive with more applications, exercises and theoretical content by extending course duration to a longer time

The video segments could be made longer to incorporate more information on how the modeling is done. A lot of new information was thrust into the weekly exercises. It would be better if the weekly exercises were a test of what we had learnt. A great course on the whole, anyway. The instructor was very clear.

This is an outstanding course. I am a physician and this has been very helpful in bridging the knowledge gap between what I learned in other deep learning courses and the unique situation of working with medical data. I would however appreciate a deeper dive into how to work with the DICOM format.

Introduces the fundamentals of using AI for medical diagnoses. Concepts are clearly explained and the assignments are well framed. More lectures regarding subtle concepts like MRI Image registration and calculation of confidence interval would have made the course more interesting and comprehensive

The course serves as an introduction to AI applications on medical diagnosis. The assignments are easy. However, video lectures are missing some minor concepts that suddenly appear in the programming assignment. It is recommended to take this course after you took Deep Learning Specialization.

Good course but, as often is the case, too much time was spent on fixing small errors in notebooks, such as using the "wrong" function (i.e. np.multiply doesn't work when * does due to the very specific setup of the exercise, even though they are both element-wise multiplication).

A very well structured course that covers most of the practical design challenges of deep learning applications in healthcare sector. A good foundation for people who want to pursue a career as a Machine Learning Engineer for medical diagnosis and/or computer vision.

Great course! Although the coding exercises focus more on lower level details of matrix manipulation, and not on the parts for selecting a model, building and training it. Most of the model related code is provided if form of utility code or as pretrained weights.

Hello,

I enjoyed taking this course. It would be great if assignments could be debuged, I tried downloading the assignments to debug using vscode but some parts of the assignments(datasets or some functions) were not there to be downloaded.

Thank you

This was a great course for getting a high-level understanding of AI's applications in medical diagnosis.

The only issue is that the assignments are auto-graded which, coupled with bugs, can make submitting assignments very frustrating.

Great but 1) all notebooks must be moved to Tensorflow 2 and Pytorch 2) it's not a Deep Learning course but a data course (for people who want to really understand the classification/Unet models, they need to study another DL course)

Course data and related util files with reasonable explanations will make this course magnificent. I spent a lot of time figuring out differences while I try it in my local engine due to version differences.

Pleasant pacing, very clear and concise lecture material. I was really frustrated with the final assignment though. Would be nice if the grader gives something more instructive than correct/incorrect.

A good course to understand the use of Deep Learning and AI in Medical Diagnosis. In this course, you can understand different ways to segment and analyze the images of brain tumors and X-Rays.

Use cases selected were really nice, Videos should carry more detail technical aspects and could be bit more lengthy and Assignments should consider multiple options to solve given problem