Friday 14 March 2025
Scientists have been working on a new way to diagnose leukemia, a type of cancer that affects the blood and bone marrow. Leukemia is typically diagnosed through microscopic examination of blood samples, but this process can be time-consuming and requires highly trained professionals.
A team of researchers has developed a new approach using deep learning techniques to identify leukemia cells in blood smear images. The method uses convolutional neural networks (CNNs), which are commonly used for image recognition tasks such as facial recognition and object detection.
The researchers created two models: one custom-built model and another based on the popular MobileNetV2 architecture. Both models were trained on a dataset of 3,256 blood smear images from patients with acute lymphoblastic leukemia (ALL).
The custom-built model consisted of three convolutional layers followed by a flattening layer and then fully connected layers. The output layer used the softmax activation function to predict one of four classes: benign, early, pre, or pro ALL.
The MobileNetV2-based model used a similar architecture but with some modifications to the classification head. The output layer also used the softmax activation function, but with a different set of weights and biases.
Both models were evaluated using metrics such as precision, recall, F1 score, and accuracy. The results showed that the custom-built model achieved an accuracy of 98.6%, while the MobileNetV2-based model achieved an accuracy of 99.69%. These results are promising for the development of a reliable and efficient diagnostic tool.
The researchers also tested their models on a separate dataset and found that they were able to accurately identify leukemia cells even when the images were degraded or had low contrast. This suggests that the models could be used in real-world settings where image quality may vary.
The use of deep learning techniques for diagnosing leukemia has several advantages over traditional methods. For example, it can reduce the time and cost associated with manual examination of blood samples by trained professionals. Additionally, the models can process a large number of images quickly and accurately, making them well-suited for high-throughput applications.
However, there are also some limitations to this approach. For example, the accuracy of the models may depend on the quality of the training data and the complexity of the image features used. Additionally, the models may not perform as well on images that are outside the range of what was seen during training.
Cite this article: “Deep Learning Techniques for Accurate Leukemia Diagnosis”, The Science Archive, 2025.
Leukemia, Cancer, Diagnosis, Deep Learning, Convolutional Neural Networks, Cnns, Image Recognition, Blood Smear Images, Acute Lymphoblastic Leukemia, All
