Automatic recognition of myeloma cells in microscopic images using bottleneck algorithm, modified watershed and SVM...

Area of Science:

• Hematology
• Computational Pathology
• Medical Diagnostics

Background:

• Plasma cells, derived from B lymphocytes, produce antibodies crucial for fighting infections.
• Multiple myeloma is a bone marrow cancer characterized by abnormal plasma cells (myeloma cells).
• Current diagnosis relies on manual microscopic examination of bone marrow smears, which is subjective, time-consuming, and prone to errors.

Purpose of the Study:

• To develop and evaluate a computer-aided diagnostic method for accurate and efficient identification of myeloma cells in bone marrow smears.
• To reduce diagnostic time and subjectivity associated with manual pathological review.
• To provide a potential second opinion tool for pathologists in multiple myeloma diagnosis.

Main Methods:

• A computer-aided diagnostic approach was developed for analyzing digital bone marrow smear images.
• The method involved separating white blood cells from red blood cells and background.
• Plasma cells were detected using feature extraction and decision rules, followed by classification of normal plasma cells versus myeloma cells.

Main Results:

• The algorithm was tested on 50 digital images containing 678 cells (132 normal plasma cells, 256 myeloma cells, 290 other marrow cells).
• The computer-aided diagnostic method demonstrated high performance metrics.
• Achieved sensitivity of 96.52%, specificity of 93.04%, and precision of 95.28% in identifying myeloma cells.

Conclusions:

• The developed computer-aided diagnostic method shows significant potential for accurate and efficient myeloma cell detection.
• This automated approach can assist pathologists, reduce diagnostic errors, and expedite the diagnosis of multiple myeloma.
• Further validation on larger datasets is recommended to solidify its clinical utility.