Synthetic Blood Smears Generation Using Locality Sensitive Hashing and Deep Neural Networks

Abstract

Peripheral Blood Smear (PBS) analysis is a vital routine test carried out by hematologists to assess some aspects of humans' health status. PBS analysis is prone to human errors and utilizing computerbased analysis can greatly enhance this process in terms of accuracy and cost. Recent approaches in learning algorithms, such as deep learning, are data hungry, but due to the scarcity of labeled medical images, researchers had to find viable alternative solutions to increase the size of available datasets. Synthetic datasets provide a promising solution to data scarcity, however, the complexity of blood smears' natural structure adds an extra layer of challenge to its synthesizing process. In this work, we propose a methodology that utilizes Locality Sensitive Hashing (LSH) to create a novel balanced dataset of 2500 synthetic blood smears. This dataset, which was automatically annotated during the generation phase, will be made public for research purposes and covers 17 essential categories of blood cells. We proved the effectiveness of the proposed dataset by utilizing it for training a deep neural network, this model got a very high accuracy score of 98.72% when tested with the well known ALL-IDB dataset. The dataset also got the approval of 5 experienced hematologists to meet the general standards of making thin blood smears .

Deep Learning for the Detection of Acute Lymphoblastic Leukemia Subtypes on Microscopic Images: A Systematic Literature Review

“…The method employs image processing techniques and deep learning to normalize the radius of the cell, estimate focus quality, adaptively improve image sharpness, and then perform classification. Al-Qudah et al [82] Classification + Localization…”

“…The new YOLOv2 architecture makes it simpler to choose a model based on its speed and precision and has fewer parameters. In a previous study [82], the YOLOv2 model was used to identify blood cells in microscopic images of multicellular blood. The YOLOv2 model is modified through random resizing of the network input.…”

A Systematic Review on Recent Advancements in Deep and Machine Learning Based Detection and Classification of Acute Lymphoblastic Leukemia

“…YOLOv3 is an improved version of YOLO that applies a logistic regression-based prediction approach to perfectly detect bounding boxes [138]. Ai-Qudah and Suen [140] have employed YOLOv2 [137] to classify healthily and ALL efficiently. YOLOv2 [137] is an improvised version of YOLO [136] that boosts both precision and speed.…”

“…Thus, transfer learning schemes are emerging as popular deep learning schemes for ALL detection because of their promising performance even in small datasets. Transfer learning-based method suggested in [35] achieves outstanding performance with 100% accuracy, whereas YOLOV2 based ALL classification method presented in [140] Most of the researchers use two quite popular publicly available ALL datasets: ALLIDB1 and ALLIDB2 to detect and classify ALL. ALLIDB1 dataset is the most popular dataset for ALL detection, which is used in around 43.33% cases, whereas the ALLIDB2 dataset is used in around 40.00% cases.…”