Machine Learning based Models for Early Screening and Detection of Hematologic Cancer using Complete Blood Count(CBC)
The modern AI systems and algorithms have disrupted every walk of our life, but industries such as hospitals and medical healthcare labs are still quite reluctant when it comes to incorporating or even taking help from these AI or CS based systems, they think and feel as if they?re too unreliable or
2025-06-28 16:28:31 - Adil Khan
Machine Learning based Models for Early Screening and Detection of Hematologic Cancer using Complete Blood Count(CBC)
Project Area of Specialization Artificial IntelligenceProject SummaryThe modern AI systems and algorithms have disrupted every walk of our life, but industries such as hospitals and medical healthcare labs are still quite reluctant when it comes to incorporating or even taking help from these AI or CS based systems, they think and feel as if they’re too unreliable or worse they don’t qualify to be taken help from. So in order to integrate AI with the healthcare system we are building ML/DL based models which would be able to not just detect but, this model would be able to achieve early screening of blood cancer or as the term goes hematologic malignancies.
Project ObjectivesThe detection and screening of cancer are some of the most tiresome and depressive things that a person could go through, and on top of that the our healthcare system and the overall world’s healthcare system requires an awful amount of tests that have to be done before the doctors could come to a possible conclusion as to what kind of cancer does the patient have and where is the cancer growing. There are a bunch of different tests that have to happen once a person starts showing the possible early signs of blood cancer, some common symptoms might be fatigue, infection, bone pain, fever, bruising/bleeding, and petechiae. The first thing that a doctor might recommend is the complete blood count or the CBC test. CBC is one of the basic and fundamental tests to evaluate a variety of health disorders including hematological malignancies. We according to our research findings and with the help of a researcher from UoK’s laboratory of genomic sequencing have found that with the help of CBC report we can detect 10 different kinds of blood malignancies and then we can also find 34 different sub-groups of those malignancies, but what does that contribute towards the already well defined frameworks that the industry uses in order to detect the malignancies? Well, we are not here to completely revolutionize the healthcare industry or to make the doctors go obsolete through some science fiction AI, we are here to give a helping support system or you can imagine these ML based models to be more like a doctor who always has the most accurate and most educated guess, but only for hematologic cancer. A practical example would be, that usually when screening the blood cancer, the patients have to go through at least 10-15 tests, these different tests consists of all the possibilities that might have affected the patient’s body, but what if we could pinpoint exactly at the place where the cancer was developing, that would very legitimately save the time and money of patient, our ML/DL based model can predict where exactly has this blood cancer ventured by seeing the CBC report of a patient. Not just that, but finding the cancer through a test report such as CBC would be incredible why? Because CBC report is something that almost every person has to go through at some point in life, I mean to say that CBC test is a common is easy test compared to many other cancer detection tests. By having a CBC test as early as possible we and then the ML/DL model detecting the malignancy at such an early stage could significantly help in reducing the fatality rate.
Project Implementation MethodThe given methodology which we have choosen is going to follow an AI approach of dealing with the problem at hand. The project is being implemented using the framework called tensorflow which is a deep learning based framework. We are also making the use of different Machine Learning based libraries such as keras and sci kit learn(sklearn). We were able to implement this project or at least start implementing it because we first aquired the data from NIBD through our external supervisor.
Benefits of the ProjectThe early detection and screening of any cancer is a crucial and critical thing in itself. Making this thing easier to access and easier to use would result in better and advanced technology, new methodology and possibly resulting in less fatalities due to blood cancer, some of the salient features of our research are given below,
-
Detection of Hematological Cancer, along with 10 different types of blood cancer using only one Model.
-
Detection of Hematological Cancer using a simple CBC Report. CBC test is an easy test which a common person is reachable towards.
-
Detecting something is not enough, our model would be able to detect these hematologic malignancies earlier in life of tumor growth.
-
With a little further research we would be able to classify 10 types of blood cancer along with 34 subtypes of those cancers making this an extremely reliable method of detecting and screening cancer.
-
The patient won't have to take about 10-15 extra confirmatory tests for screening and making sure which blood cancer he/she has, rather using only one test(our test) and a few other confirmatory tests he/she would be able to pin point what cancer they have.
-
Our method could reduce the amount of work and energy that is consumed in detecting the cancer, and those energies and the expenses of those energies could be consumed in treating the patient.
-
The method we propose is extremely cost efficient compared to almost any other medical method, on top of that our method is a generalized way to solve this issue, rather than many different methods we solve(detect cancer) using only 1!
This would be deployed as an API, along with it’s CLI based app and if our internal advises then we can deploy this entire system as a desktop app and a mobile app, but if we think about it, you would see that Desktop app and a mobile app is something a common person would use, something with flashy GUI and easy to use interface, rather than that, a CLI based app or a desktop app would do it for the workplace that this system would be deployed for. Further discussion on this with our advisor. This research would result into a research publication in a journal with good impact factor.
Final Deliverable of the Project Software SystemCore Industry MedicalOther IndustriesCore Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Good Health and Well-Being for PeopleRequired Resources| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 70000 | |||
| GPU | Equipment | 1 | 50000 | 50000 |
| API hosting on AWS(E2 instance) | Equipment | 1 | 20000 | 20000 |