The WBC (White Blood Cell) Differential Count is a vital clinical test involving the microscopic counting and identification of five types of white blood cells, along with detecting abnormalities. I am contemplating the feasibility of developing software tailored for this purpose. Specifically, this software would utilize machine learning algorithms to analyze digital images of blood samples captured through a microscope. The aim is to automate the counting process and enable recognition of any abnormal cell morphology.
For instance, consider a scenario where a prepared blood slide is observed under a microscope, as depicted below. This image would be transmitted to the specialized software, which would leverage image processing techniques and machine learning models to accurately count and identify different types of white blood cells, as well as flag any abnormal cellular characteristics.
Dear Colleagues,
The company: Sysmex has developped such a system named DM 5000, with analyzed of about a hundred fields for each colored slides (by an automatic way, with direct coloration).
See the characteristic on the Roche diagnostic society website, saling this apparatus.
Best regards
Didier J
Thanks, I also used these kind of machines that provides Complete Blood Count but my purpose was more directed to using software to analyze the prepared slide. seems these techniques only embedded in these machines with technology and sold together.
Dear Hezha,
Couple of years back I attended a presentation by one of my colleague, a biomedical engineer doing similar to what you intent to achieve but with cytology slides for breast cancer diagnostics. He was using image procesing, fuzzy logic and AI. You can reach his profile here, http://ee.eng.usm.my/v4/index.php/our-people/75-prof-madya-dr-nor-ashidi-bin-mat-isa. His email is in his website. Good Luck
Dear Hezha,
The reocognition of differences between cells by imaging , completed by cytochemistry and fluorescence staining , is a complex problem studied since about 30 years, and imaging (and machines) have difficulties to be specific (the reason for combination of different technics for recognition)). So we (specialist) have due let ingeeners find solutions (not our field) and they now sale integrated solutions. So, we could used part of them (imaging) for other purposes, but have we sufficiently time ?
Best regards
Didier
There's a Software, Cellavision provided by Sysmex Corporation. It's based in comparing the images with previously established morphological patterns(based on a great bank of blood cells images)
Dear Mr. Didier,
As I understood you, someone there working on such problem and they've come with a solution right?
I was just thinking if someone has such software. it is about Machine learning and pattern recognition so I don't think it would be easy to be used for that purpose especially if we make a bad staining.
Regards
Hezha
Dear Hezha,
You are right. Machines have been develop and the principal difficuly is to have a good sofware , able to differentiate cells according to different criteria: size, morphology, staining ... and in generlal, different combined technologyiesare used (morphology and staining essentially).
This recognition has been the difficulty for 30 years, and in general the problem of image analyzing.
So, industrial would be happy for new software development, more efficient , and able to differentiate cells by less steps. And also the mean that can be used by this sofware.
Initially, the first difficulty was simply for a count , to differentiate a cell from another (when cells are too near ...).
You can obtain informations and knowledges from international societies of image analysis.
Best regards
Didier
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis