Hematology is the study of the normal and pathologic aspects of blood and blood elements. Blood is a very unique fluid composed of many cellular elements as well as a liquid portion consisting of proteins, amino acids, carbohydrates, lipids, and other macromolecules and low-molecular-weight precursors.
Blood is a circulating tissue composed of fluid plasma and cells. It is composed of different kinds of cells. These formed elements of the blood constitute about 45% of whole blood. The other 55% is blood plasma.
Blood plasma when the formed elements are removed from blood, a straw-colored liquid called plasma is left. Plasma is about 91.5% water and 8.5% solutes, most of which by weight (7%) are proteins. Some of the proteins in plasma are also found elsewhere in the body, but those confined to blood are called plasma proteins.
These proteins play a role in maintaining proper blood osmotic pressure, which is important in total body fluid balance. Most plasma proteins are synthesized by the liver. The formed elements of the blood are classified as red blood cells (erythrocytes), white blood cells (leucocytes) and platelets (thrombocytes) and their numbers remain remarkably constant for each individual in health.
Red Blood Cells they are the most numerous cells in the blood. In adults, they are formed in the marrow of the bones that form the axial skeleton. Mature red cells are non-nucleated and are shaped like flattened. In stained smears, only the flattened surfaces are observed hence the appearance is circular with an area of central pallor the indented regions. They are primarily involved in tissue respiration.
White Blood Cells they are a heterogeneous group of nucleated cells that are responsible for the body’s defenses and are transported by the blood to the various tissues where they exert their physiologic role.
WBCs are present in normal blood in smaller number than the red blood cells in adults. Their production is in the bone marrow and lymphoid tissues, lymph nodes, lymph nodules and spleen.
There are five distinct cell types each with a characteristic morphologic appearance and specific physiologic role. These are:
- Polymorphonuclear leucocytes (granulocytes).
- Neutrophils seg
- Eosinophils
- Basophiles
- Mononuclear leucocytes
- Lymphocytes
- Monocytes
Polymorphonuclear Leucocytes have a single nucleus with a number of lobes. They Contain small granules in their cytoplasm, and hence the name granulocytes. There are three types according to their staining reactions.
Neutrophils are a type of granulocyte and are the most abundant type of white blood cell in the body their size ranges from 10-12µm in diameter. They are capable of amoeboid movement. There are 2-5 lobes to their nucleus that stain purple violet. The cytoplasm stains light pink with pinkish dust like granules. They play a critical role in the immune response, particularly in combating bacterial infections.
Neutrophils are named for their multi-lobed or segmented nuclei. A typical neutrophil nucleus is divided into several distinct lobes connected by thin strands of chromatin. The segmented appearance of the nucleus gives rise to the term “segmented neutrophils” or “segmented granulocytes.”
Segment in the context of white blood cells (WBCs) refers to the appearance of the nucleus in a neutrophils.
The presence of segmented neutrophils in the blood is a common indicator used in medical laboratories to assess the overall health of the immune system. An increase in the percentage of segmented neutrophils might indicate a bacterial infection or another condition that triggers an inflammatory response. Conversely, a decrease in segmented neutrophils could be associated with conditions like bone marrow disorders or certain infections.
Stab cells refer to a specific type of immature neutrophil, which is a type of white blood cell (WBC) involved in the body’s immune response against infections. These immature neutrophils are characterized by their nucleus, which is elongated and shaped like a band rather than being fully segmented.
Basophils their size ranges from 10-12µm in diameter. Basophiles have a kidney shaped nucleus frequently obscured by a mass of large deep purple or blue staining granules. Their cytoplasmic granules contain heparin and histamine that are released at the site of inflammation.
Mononuclear Leucocytes Lymphocytes
Small Lymphocytes size ranges from 7-10µm in diameter. Small lymphocytes have round, deep-purple staining nucleus which occupies most of the cell. There is only a rim of pale blue staining cytoplasm. They are the predominant forms found in the blood.
Large Lymphocytes their size ranges from 12-14µm in diameter. Large lymphocytes have a little paler nucleus than small lymphocytes that is usually eccentrically placed in the cell. They have more plentiful cytoplasm that stains pale blue and may contain a few reddish granules.
Lymphocytes are responsible for recognizing and responding to specific antigens, which are foreign substances such as bacteria, viruses, and other pathogens.
Monocytes are the largest white cells measuring 14-18µm in diameter. They have a centrally placed, large and horseshoe shaped nucleus that stains pale violet. Their cytoplasm stains pale grayish blue and contains reddish blue dust-like granules and a few clear vacuoles.
They are capable of ingesting bacteria and particulate matter and act as scavenger cells at the site of infection.
Platelets these are small, non-nucleated, round/oval cells/cell fragments that stain pale blue and contain many pink granules. Their size ranges 1-4µm in diameter. Platelets are produced in the bone marrow by fragmentation of cells called megakaryocytes which are large and multinucleated cells. Their primary function is preventing blood loss from hemorrhage.
When blood vessels are injured, platelets rapidly adhere to the damaged vessel and with one another to form a platelet plug.
Deep Learning (DL)
Deep learning is a subset of machine learning, which in turn is a subset of AI. AI can just be a programmed rule that tells the machine how to behave in certain should behave in situations.
Machine learning requires extensive intervention of the user to minimize errors of the model and predictions of the model versus the Meet ground truth data.
Modern deep learning provides a very powerful framework. By adding more layers within a layer, a network with are presented with increasing complexity.
Most of the tasks that consist of a Mapping an input vector to an output vector and doing it easily and quickly for one person are done can be solved with deep learning.
Whole slide images (WSI)
Whole slide images (WSI) also known as digital pathology images or virtual slides, are high-resolution digital representations of entire microscope slides used in the field of pathology. A glass microscope slide is scanned using a specialized high-resolution scanner to produce a WSI, which records the complete slide at various magnifications. With the use of appropriate software, the resulting digital image can often be studied and browsed as a sizable file.
I am text block. Click edit button to change this text. Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.
By using a WSI pyramid structure, digital pathology systems can achieve a balance between fast navigation and detailed visualization, even for extremely large images.
In the context of Whole Slide Images (WSI), magnification refers to the ability to zoom in and out on the digital image to view different levels of detail. The concept of magnification in WSI is closely tied to the pyramid structure.
A WSI pyramid consists of multiple levels or layers, each containing an image of different resolution. These levels represent different magnification levels, where higher levels correspond to lower magnifications (more overview) and lower levels correspond to higher magnifications (more detail).
A low magnification is when viewing the WSI at a low magnification, the system loads the highest level of the pyramid (with the highest numbers). This level provides an overview of the entire slide, allowing you to quickly navigate and locate regions of interest.
High Magnification is when zooming in further can involve transitioning through multiple levels of the pyramid to reach the desired level of detail. The highest pyramid level (with the lowest number) provides the highest magnification and the most detailed view of the tissue sample.
Digital Representation of Image
A digital image is composed of tiny picture elements called pixels. Each pixel represents a single point in the image and contains information about its color, brightness, and other attributes pixels are arranged in a grid to form the overall image.
The resolution of a digital image refers to the number of pixels per unit of measurement it means higher resolution images contain more pixels, resulting in greater detail and clarity.
Color depth is a fundamental concept in digital representation of images that determines the number of colors that can be displayed or represented in an image. It quantifies the range and precision of colors that can be captured, stored, and displayed in a digital image. Color depth is measured in bits per pixel, and it affects the level of detail and realism an image can convey.
Digitalization of slides with HSA KIT
With the help of (HSA SCAN M) software you can easily convert pathology slides into WSI by scanning manually the slide with microscope. It is easy and inexpensive method with high quality. The software’s user-friendly interface allows for easy navigation and customization of scanning settings to meet specific needs.
Additionally, the integration with HSA KIT software allows for advanced image analysis and interpretation, providing valuable insights for research and diagnosis. Overall, HSA SCAN M software offers a comprehensive solution for labs seeking to streamline their slide digitization process while maintaining quality control.
Ground Truth Data
After the scanning process is finished and we have WSI files in HAS KIT that are ready to start annotating them manually and in order to train a model in deep learning Ground Truth Data is needed which is accomplished by creating Base ROIs in files and annotating all of the existing cells within the Base ROIs.
GTD numbers for SKK project that had 4 files with +19,000 GTD is shown in (table 1).The annotations are set within the ROI with regard to a function of the HSA KIT.
Then the blood cells structure are annotated and this structure has 3 sub-structures which are thrombocytes, leucocytes and erythrocytes, the leukocytes is further divided into 4 different classes including (Lymphocytes, Monocytes, stab and Segment) and the numbers of the classes are (1338, 258, 267 and 2175) respectively. The quantity and quality of these annotations depends on the location, clarity and size of the base ROI.
Metrics
The effectiveness of algorithms for comprehending medical images is evaluated using a variety of measures. The table used to visualize algorithm performance and determine multiple assessment metrics is called the confusion matrix. Confusion matrices are used to evaluate deep learning models and give a more realistic view of their performance. The output could have two or more classes. In the table, there are four possible combinations of anticipated and actual values.
Mean Average Precision
Mean Average Precision (mAP) is a metric used to evaluate object detection models. Confusion Matrix, Intersection over Union (IoU), Recall, Precision are the sub-metrics that form the backbone of the formula for the mAP accuracy. The mAP is calculated by finding Average Precision(AP) for each class and then average over a number of classes.
Explainable Deep Learning (XDL)
Explainable Deep Learning (XDL) refers to the practice of making deep learning models, which are often complex and difficult to interpret, more understandable and transparent. Deep learning models, particularly deep neural networks, have demonstrated impressive performance across a wide range of tasks, but their inherent complexity can make it challenging to comprehend how they arrive at their predictions.
XDL techniques aim to address this issue by providing insights into the inner workings of deep learning models, making their decisions more interpretable for humans.
Stochastic Gradient Descent
Stochastic Gradient Descent (SGD) is a foundational optimization method used to minimize the loss function of models, especially in deep learning. Given the complexity of neural networks. The random fluctuations in SGD can help the model escape from shallow local minima in non-convex loss landscapes, which are typical in deep learning.
Heat Maps
Heat Maps are graphical representations of data that utilize color-coded systems. The primary purpose of Heat Maps is to better visualize the volume of locations/events within a dataset and assist in directing viewers towards areas on data visualizations that matter most.
Vision Transformer
ViT model represents images as sequences and predicts a class label for the image. This makes it possible for the models to learn the image structure independently. The input images are treated as a sequence of image patches, where each image patch is flattened into a single vector by concatenating the channels of all pixels and linearly projected onto the desired input dimension.
Mask R-CNN Architecture
The Mask R-CNN Architecture is based on the R-CNN structure. R-CNN has two outputs for each object candidate, a class label and a bounding box offset. There is also a third branch that outputs the object mask. So the Mask-R-CNN is a natural and intuitive idea. But the additional mask output is different from the class and bounding box outputs and requires the extraction of a much finer spatial arrangement of an object.
IHC vs flow cytometry
Both IHC and flow cytometry use antibodies to detect proteins, their results are different. IHC provides spatial information you can see exactly where in a tissue sample a protein is expressed. While Flow cytometry provides more quantitative information about the number or percentage of cells in a sample that express specific proteins and to what degree they express them.
HSA KIT
HS Analysis GmbH provides software for the automatic analysis of tissue samples and cell cultures so that researchers and pathologists in hospitals and the pharmaceutical industry can quickly find answers to crucial questions about disease processes. Neural networks and deep learning are used for this. They help to automatically identify and evaluate your region of interest (ROI). In this way you avoid time-consuming manual annotation of ROIs and achieve analyzes with high throughput and short calculation times.
