• AI in Diagnostic Imaging: How computers are learning to read X-rays, CT scans, and MRIs to help doctors catch diseases

• AI-Powered Drug Discovery: How AI is speeding up the process of finding and developing new medicines

• Personalized Medicine Systems: How AI analyzes a person's genes and health history to create treatment plans designed specifically for them.

  
  

For each topic, I will explain what technology is being used, what benefits it brings, what problems it still has, and what I personally find interesting about it. Where AI in healthcare is heading in the future, what ethical issues we need to think about, and how all of this affects society - things like jobs, privacy, and whether everyone will have equal access to these new tools.

Technologies used - AI diagnostic doesn't only use deep learning to function, but also few others:

• Deep Learning & Convolutional Neural Networks (CNNs): This is the main technology behind medical image analysis. Deep learning is the main way to train any AI model, so it is no surprise it is also being used here.

• Computer Vision: Pretty self explanatory name: allows AI to analyze the imagine, just like a radiologist would.

• Machine Learning: The AI gets better at finding diseases the more images it sees

• Image Segmentation: AI outlines specific parts of an image, like drawing a border around a tumor, which highlights the suspected area of treatment.

• Transfer Learning: Instead of starting from scratch, AI can use knowledge from one type of scan to help read another type

Benefits and improvements:

Like we discussed previously, AI diagnostics are able to catch details that human eyes might miss and, as such, reduce possible misdiagnoses. It's capable of delivering faster results, allowing radiologists to review more scans in less time so patients receive answers sooner. It's worth mentioning the consistency, since AI provides the same quality of reading whether it is 3 p.m. or 3 a.m., without any breaks. It also provides precise measurements, such as calculating tumor size exactly rather than relying on estimates. With AI automatically flagging urgent cases, the sickest patients are always prioritized, which significantly improves workflow and ensures that all patients who need treatment as soon as possible will receive it. And finally, AI expands access to quality diagnostics in rural or underserved areas where specialist doctors may not be available.

Challenges and Limitations:

But, with all those benefits, there are guaranteed challenges. One of which is the data bias. Not every human being is the same, so treating unique cases may be not impossible, but highly difficult. The “black box” problem also creates unease, since doctors may not fully understand how AI comes up with its decisions. And of course what about the legal responsibility? If AI contributes to a wrong diagnosis, it is not very clear whether the doctor, hospital, or the company should be held accountable. We can continue on by mentioning things like overreliance, or maybe the issues with regulatory approval from FDA, but I think that those challenges, are a little less significant than the ones I mentioned. (Najjar 6, 2023)

Video Resources:

Using artificial intelligence in radiology clinical practice - https://youtu.be/dCDuMyzWS8Q?si=35qXdWTaenLtGcmu

Personal Insight:

I think that as long as there is a use for AI and it actually helps to save lives, it is undoubtedly amazing news. Of course, there are still things that need to be resolved and improved, but I think this is a very expected outcome from a new technology. I am sure that if we give it a few more years of development, AI will be highly unlikely to make a mistake.

Technologies used :

• Natural Language Processing (NLP): Allows AI to read scientific papers, patents, and clinical trial reports to find connections and ideas for new drugs

• Machine Learning & Deep Learning: This is what we will see across all AI models. Since it was already explained in the AI in Diagnostic post, I won't use time to repeat the same information.

• Generative AI: Creates brand new molecules that have never existed before

• Molecular Modeling: Simulates how drug molecules attach to and affect proteins in the body

• Reinforcement Learning: Keeps tweaking drug designs to make them better and better

• Knowledge Graphs: Connects different pieces of medical information to find unexpected connections

Benefits and Improvements:

I think the fact that decades of research for the new drugs can be condensed down into 2-3 years is one of the most of the most overlooked benefit of AI technology. With lowering need research time it would also make development far cheaper, lowering costs that typically average $2.6 billion per drug. How about the ability to “fail faster,” where AI can identify weak potential drugs early before companies waste millions on their testing. In a nutshell, AI is a huge resource saver for pharmaceutical industry.

Challenges and Limitations

Well, as fantastic as it sounds, AI is obviously far from perfect. Biology is extremely complex subject, and AI cannot capture every single detail of how the human body works and reacts to certain drugs, which means that it may overlook some life-risking problems. And this will sound controversial to myself, but even when AI predicts how a drug will work, it still requires years of laboratory and human testing, since we cannot blindly trust it as of now. And just like with AI Diagnostic, things like regulations, patent ownership, lack of data, and lack of faith in AI are still the issues that are waiting to be resolved.

Video Resources:

How AI Is Accelerating Drug Discovery - https://youtu.be/mqBvitxD05M?si=lAX9uRGDZ4GAhcrD

Personal Insight:

Drug creation by AI honestly seems like such an insane idea, but it is a very important one. What excites me most is the potential for rare diseases. Right now, drug companies mainly focus on diseases affecting millions of people because developing drugs is so expensive. But if AI makes it ten times cheaper and faster, suddenly it becomes realistic to develop treatments for rare diseases that only affect a few thousand people. For families dealing with rare genetic conditions that have no treatment, this could be life-changing. And of course, after the events of the COVID-19 pandemic, it is pretty obvious that we should be able to create medicine in a matter of years rather than decades when we need new treatments. However, with that said, it seems like there is this one singular trend continuing after AI in diagnostics: AI is still a little too undeveloped to be used in healthcare. The promises are very ambitious and full of potential. However, with that said, AI as a technology is still way too new to be used in something like drugs for patients. I am not denying that it will improve in the next few years and become very reliable, but as of right now, this is way too far from perfect. Personally, if I were told that I was prescribed a medication that was made by AI without human supervision, there is a 0% chance I would take it. I just cannot trust AI with my life or health. But I certainly hope that this opinion will change in the future - which is not a matter of if, but when.

Technologies Used:

• Machine Learning & Predictive Analytics: Analyzes patient's health data to predict how diseases will progress and how you'll respond to treatments

• Natural Language Processing: Pulls important information from patient's medical records

• Genomic Analysis: Interprets patient's genetic code and figures out what different genetic variations mean for their health

• Deep Learning Neural Networks: Finds complicated patterns across all different types of patient data

• Real-Time Monitoring: Constantly analyzes data from fitness trackers and medical devices

• Decision Support Systems: Gives doctors evidence-based treatment recommendations

• Data Integration: Combines genetic info, medical records, environment, and lifestyle into one complete picture

Benefits and Improvements:

There are many benefits to AI-driven personalized medicine. For example, it helps to match patients with treatments that are most likely to work for them, while also predicting and preventing harmful drug reactions. It can also calculate the exact right dosage of medicine based on how an individual’s body processes it, ensuring safer and more effective care. It could function as an early warning system, spotting disease risks years before symptoms appear. It doesn;t waste time on unnecessary procedures by avoiding treatments unlikely to help, saving lots of valuable time for the patient, which is crucial when dealing with diseases.

Challenges and Limitations:

I think one of the major concerns with this AI usage is the data protection. Genetic and health data are highly sensitive and, if they are leaked, it could have lifelong consequences. Your own data could be maliciously used against you, and you wouldn't even find that out until it's too late. Another one in my opinion is the accessibility as these quite advanced treatments may initially be available only to wealthy patients, which could worsen the healthcare inequality. And of course data bias: as I already explained, every human has a unique approach, and generalizing the data for every case could be consequential.

Video Resources:

AI develops personalized treatment for ultra-rare genetic condition: HealthLink - https://youtu.be/BTX7UP0sbmk?si=6XILrBt1vlAZP0U9

Personal Insight:

I think the most exciting thing about this AI is its approach to treatment. As I already established here and in previous posts, right now, healthcare focuses on treating the symptoms of the sickness. But with this AI model, we can spot risks decades early and help to prevent disease before it even starts. Imagine knowing at 25 that you have a high risk for heart disease and getting a specific plan, which is tailored exactly to your genes and lifestyle that prevents a heart attack at 50. That's just insane to even think about. But as exciting as this can be, I am also worried about the potential data breaches. In previous works for this class, I mentioned a data leak that happened in of the hospitals in the U.S. Back then the biggest damage was the stolen insurance transactions. Now let's consider what would happen if now, thieves would have access to everything possibly known about the patient, starting from their daily routine, and ending with their literal genetic code. And if not thieves, what is the company itself would want to go against the patients? Those are the questions I can't give an answer to at the moment, but they are certainly still very important ones. I don't think I will entrust so many details about my own body and habits to AI unless I would know for sure that the odds of this data being leaked or weaponized are near zero.

But with all of those advancements, I brought up earlier the point of privacy and data protection. Privacy is huge - AI needs tons of sensitive patient data, and if there is valuable data, there are people who would want to access it in any way possible. It is really hard to trust companies the safety of your data, especially when we are talking about healthcare. And as such, unless there is a system to protect that data, I don't think that those AI advancements will go far. Another consideration is the data bias. Across all three examples, one thing was consistent: deep learning. Health is not something one can generalize all the time. There are bound to be cases that have never occurred or have unique challenges to them. And because of that, I fear that not all the patients will be able to access the proper care they need. And this brings up another point: liability. God forbid, in case of the patients being wrongfully treated, which led to a bad outcome, who is to take the accountability? Is it the AI model that misdiagnosed the patient? Is it the company that created that AI model? I don't think there is a universal answer. Just like health, I think those issues should be looked at case-by-case.

To summarize my thoughts, I think that with the proper approach and with determined mindset, we can achieve the perfect results with implementing AI into healthcare. As long as we are careful, as long as everything is taken into account, this idea does not seem to be impossible. One thing I would like to add to the point of mistreatment in Ethical Consideration is that fails are inevitable. No matter how hard we try, no matter how careful we are, if something can go wrong, it will go wrong. Data breaches will happen; some people will unfortunately suffer from it. And yet, I don't think that should stop us. I think that if anything, this should encourage us to do better, and show the areas where the improvement should be made. If we cannot accept fails, we cannot make any progress.

With having said all of that, as scary as those ideas may appear, I am very hopeful about our future. There is no guarantee that the points I've made won't ever happen - they very much might - but I think that there are more people on this planet that would do anything to prevent that from happening, than there are the people who are all for that type of control. I think that the worse case scenario is when all of the bad things mentioned will happen, we just wouldn't know it. But, there is no point in guessing, we just have to wait and see where this AI future brings us. At the moment, I am very scared of the AI replacing jobs. It is scary to even consider the 2nd great depression being possible.

This source gives current real-world examples of AI being used across different healthcare areas, which I used for understanding what's actually happening in the industry right now.

- National Library of Medicine. (2025). 2025 Watch List: Artificial Intelligence in Health Care. https://www.ncbi.nlm.nih.gov/books/NBK613808/
A medical source that provided evidence-based analysis of AI's clinical applications and regulatory issues, which highlighted benefits and challenges

- Bajwa, J., et al. (2021). Artificial intelligence in healthcare: transforming the practice of medicine. PMC. https://pmc.ncbi.nlm.nih.gov/articles/PMC8285156/
A peer-reviewed article that gives technical details about AI methods used in medical practice, which was crucial for understanding the technology section.

- Cellina, M., et al. (2023). Redefining Radiology: A Review of Artificial Intelligence Integration in Medical Imaging. PMC. https://pmc.ncbi.nlm.nih.gov/articles/PMC10487271/
Focuses specifically on diagnostic imaging AI, which I used for understanding how computer vision works in healthcare for the first example.

- Singh, N., et al. (2024). AI in diagnostic imaging: Revolutionising accuracy and efficiency. ScienceDirect. https://www.sciencedirect.com/science/article/pii/S2666990024000132
Recent research showing measurable improvements in diagnostic accuracy from AI, providing concrete evidence rather than just hype.

- Sharma, A., et al. (2024). Artificial Intelligence Applications in Drug Discovery and Drug Delivery. PMC. https://pmc.ncbi.nlm.nih.gov/articles/PMC11510778/
Detailed look at AI's role in pharmaceutical development