Novel Methods Discovered through Which Bone Marrow Cells Can Be Better Visualized

Table of Contents

1 CPPs shed light on the cytology of the bone marrow
2 Clinical significance
3 Conclusion
- - 3.0.1 References

The spongy part of bones contains a semi-solid tissue called bone marrow. It’s made up of hematopoietic cells that produce blood cells and stroma that are not directly involved in blood cell production but have a supportive role. A way in which these cells can be visualized is known as bone marrow cytology. This is the identification and counting of bone marrow cells, it can be very useful in making a hematological diagnosis. Digital pathology has seen promise from artificial intelligence (AI), particularly deep networks. In bone marrow cytology, deep models for cell detection have proven useful, yielding remarkable accuracy and computational efficiency outcomes. These models enable and automate numerous pathology diagnostic workflow components, including information extraction from digital whole slide images (WSIs) of pathology specimens. Unfortunately, AI models are still not frequently used in diagnostic hematopathology processes.

Bone Marrow

CPPs shed light on the cytology of the bone marrow

In cytopathology, quite a number of AI models have been developed for cell detection and classification using bone marrow aspirate (BMA) images. The output of these models frequently takes the form of a nucleated differential count (NDC), where a variety of distinct bone marrow cell types are numbered and grouped according to minute physical characteristics. Due to its disregard for the morphological complexity of BMA, the NDC is only marginally useful in several hematological diagnoses, such as myelodysplastic neoplasm (MDS). A further drawback of such models is that they do not spare pathologists the tiresome work of looking at a huge WSI to analyze thousands of cells. A more effective method called cell projection plot(CPP) has been introduced. Here, BMA is visualized via deep neural networks. The reduction in dimension is well utilized. As a result, combining deep feature extraction and dimensionality reduction is a natural way to communicate BMAs to pathologists in a way that is easy for them to use and understand.

A study was carried out by a group of scientists to determine the effectiveness of this method. A total of 60 BMA samples are evaluated by three hematopathologists using CPP samples from 20 individuals each. To view and assess WSIs and CPPs, the pathologists used specialized software. In the investigation, bone marrow cell detection was performed using a pre-trained YOLO object detection algorithm. Bone marrow cells are identified and categorized by YOLO using digital WSI aspiration data. Marrow cells were identified and evaluated. The experiment was created to gauge how well CPPs work on their own. The results suggested that pathologists might swiftly analyze these CPPs as an interpretable implementation of AI in BMA cytology as well as a condensed representation of BMA cytology.

Clinical significance

CPP can be utilized to improve the comprehension of cytology results by clinical stakeholders other than pathologists like the patients and patients’ relatives. Pathologists could make notes on CPPs and discuss their findings with other medical experts because CPPs are more transferrable than WSIs. It could also be utilized as a diagnostic tool in hematological diseases like multiple myeloma, myelodysplastic syndromes, acute myeloid leukemia, etc.

Conclusion

To effectively and compactly portray the data obtained by an objection detection model from BMA WSI, a new and efficient visualization method (CPP) has been provided. This tool is very useful in providing pathologists with clearer and more accurate results and analyses.