Gaussian Blurring Technique for Detecting and Classifying Acute Lymphoblastic Leukemia Cancer Cells from Microscopic...

Area of Science:

• Medical image analysis
• Computational pathology
• Artificial intelligence in diagnostics

Background:

• Visual inspection of peripheral blood smears is crucial for leukemia diagnosis.
• Current methods can be time-consuming and lack uniformity, especially in telemedicine.
• Automated solutions offer potential for improved accuracy and efficiency.

Purpose of the Study:

• To propose a novel method, GBHSV-Leuk, for automated segmentation and classification of Acute Lymphoblastic Leukemia (ALL) cancer cells.
• To enhance the accuracy and uniformity of leukemia cell detection using artificial vision.
• To facilitate earlier and more reliable diagnosis of ALL.

Main Methods:

• A two-stage approach: Gaussian Blurring (GB) for noise reduction and Hue Saturation Value (HSV) color space segmentation with morphological operations.
• Image pre-processing to reduce noise and reflections.
• Segmentation techniques to differentiate cancer cells from background for improved prediction accuracy.

Main Results:

• The GBHSV-Leuk method achieved 96.30% accuracy on a private dataset.
• The method demonstrated 95.41% accuracy on the public ALL-IDB1 dataset.
• The results indicate high performance in identifying and classifying ALL cancer cells.

Conclusions:

• The proposed GBHSV-Leuk method shows significant potential for accurate and efficient automated detection of ALL cancer cells.
• This technique can improve the diagnostic process, particularly in telemedicine applications.
• Early and reliable detection of ALL cancer is facilitated by this artificial vision approach.