A Novel Shallow ConvNet-18 for Malaria Parasite Detection in Thin Blood Smear Images

Abstract

Accurate diagnosis of plasmodium parasite from blood cell images is essential to prevent the further spreading of the deadliest disease, malaria. It is an infectious disease, mainly transmitted by female Anopheles bite. Conventionally, microscopists can diagnose this disease by examining the thick and thin blood smears. Due to inter/intraobserver errors, the classification accuracy may get affected. To overcome this, a robust and shallow convolutional neural network is developed for the automatic detection of malaria parasite using thin blood smear images. The network is trained with 80% of images (11,023 parasitized and 11,023 uninfected) and tested with 20% (2756 parasitized and 2756 uninfected) of images. Several standard pre-trained models like Alexnet, VGG-16, VGG-19, Resnet-18, Resnet-50, Resnet-101, Squeezenet, Mobilenet-v2, Inception-v3, Googlenet, Xception, and Densenet-201 are implemented and the results obtained are compared with the proposed method. Classification accuracy, sensitivity, specificity, positive predictive value, and F1-score are the metrics used to evaluate the performance of the networks. Compared to existing pre-trained models, the proposed CNN has achieved better results with classification accuracy of 97.8%, sensitivity of 97.9%, specificity of 97.8%, positive predictive value of 97.8%, and F1-score of 97.84%. The proposed method performs the training much faster compared to the pre-trained network due to less number of parameters.

Microscopic parasite malaria classification using best feature selection based on generalized normal distribution optimization

“…The features are derived from classical handcrafted techniques and achieved a 91% accuracy by using a naïve Bayes classifier ( Maqsood et al, 2021 ). In this work, the features were extracted using different handcrafted methods and achieved an accuracy of 98.35% ( Elangovan & Nath, 2021 ). The fusion of CNN and handcrafted features gives an accuracy of 88% ( Sadafi et al, 2021 ).…”

Tile-based microscopic image processing for malaria screening using a deep learning approach

“…However, thick blood film is the most commonly used slide preparation technique for malaria diagnosis, and the development of robust automated tools is critical in reducing the difficulties associated with manual microscopy based malaria diagnosis. The majority of existing studies on malaria parasite detection or classification use thin smear blood films [ 42 – 46 ]. This could be due to the ease with which infected and uninfected RBCs can be distinguished due to their larger size in thin film blood smear microscopic images.…”

Diagnosis of malaria disease by integrating chi-square feature selection algorithm with convolutional neural networks and autoencoder network

“…InceptionV3 is a CNN model trained on more than one million images from the ImageNet database (Elangovan and Nath, 2021). This model has a 48-layer structure.…”