Table 1 Summary of recent deep learning approaches in automated malaria diagnostic systems
Author | Database | Plasmodium species | Classification | Technique | Results |
|---|---|---|---|---|---|
Sriporn et al. [18] | NLM, 7000 cell images | P. falciparum | Binary | Xception, Inception-V3,ResNet-50,NasNetMobile,VGG-16, AlexNet | Best performing model: Xception Accuracy: 99.28% Precision: 99.29% Recall: 99.29% F1-Score: 99.28% |
Umer et al. [19] | NLM, 27558 cell images (150 infected, 50 healthy patients) | P. falciparum | Binary | Customised CNN | Accuracy: 99.96% Precision: 100% Recall: 99.93% |
Zhao et al. [20] | NLM, 27558 cell images (150 infected, 50 healthy patients) Broad Institute dataset contains 1364 blood smear images with 80,000 infected cells | P. falciparum P.vivax | Binary | ResNet50V2, VGG16, VGG19, InceptionV3, DenseNet121, MobileNetV2 | Best performing model: VGG-16 Accuracy: 96.53% Sensitivity: 95.0% Specificity: 98.07% AUC: 99.40% F1-Score: 96.48% MCC: 93.30% Cross-dataset: AUC:94.5% |
Ragb et al. [21] | NLM, 27,558 cell images (150 infected, 50 healthy patients) | P. falciparum | Binary | SqueezeNet, MobileNetV2, GoogleNet, ResNet18, DarkNet19, InceptionV3, AlexNet, Xception, AlexNet, DenseNet201, ResNet101, VGG19, Ensembled model | Best performing model: Ensembled model Sensitivity: 97.94% Specificity: 97.78% Precision: 97.8% |
Cinar et al. [22] | NLM, 27558 cell images (150 infected, 50 healthy patients) | P. falciparum | Binary | AlexNet, ResNet50, DenseNet201, VGG19, GoogleNet, InceptionV3 | Best performing model: DenseNet201 Accuracy: 97.83% |
Maqsood et al. [23] | NLM, 27558 cell images (150 infected, 50 healthy patients) | P. falciparum | Binary | VGG16, VGG19, Xception, Densenet121, Densenet169, Densenet201, Inceptionv3, InceptionResnet_v2, Resnet50, Resnet101, Resnet152, SqueezeNet, Customised CNN | Specificity: 97.78% sensitivity: 96.33% Precision: 96.82% Accuracy: 96.82% F1-Score: 96.82% MCC: 93.64% |
Diyasa et al. [24] | NLM, 27,558 cell images (150 infected, 50 healthy patients) | P. falciparum | Binary | GoogleNet | Accuracy:93.89% |
Loddo et al. [25] | NLM, 27558 cell images (150 infected, 50 healthy patients) MP-IDB, 229 images | P. falciparum P. vivax | Binary and Multi-class | AlexNet, DenseNet-201, ResNet-18, ResNet-50, ResNet101, GoogleNet, ShuffleNet, SqueezeNet,MobileNetV2, Inceptionv3, VGG-16 | Best performing models Binary: ResNet-18 Accuray: 97.68% Multi-class: DenseNet-201: Accuracy: 99.40% Cross-dataset validation: Accuracy: 97.45% |
Shambhu et al. [26] | NLM, 27558 cell images (150 infected, 50 healthy patients) | P. falciparum | Binary | Customised CNN | 96.02% |
Vijayalaskhmi et al. [27] | 1030 infected images and 1520 non-infected images | P. falciparum | Binary | LeNet-5, AlexNet, GoogleLeNet, VGG16, VGG19 | Best performing model: VGG19 Sensitivity: 93.44% Specificity: 92.92% Precision: 92.92% Accuracy: 93.13% F1-Score: 93.13% |
Arshad et al. [28] | IML-malaria | P. vivax | Multi-class | VGG16, VGG19, ResNet50V2, DenseNet169, DenseNet201 | Best performing model: ResNet-50v2: 79.61% |
Rahman et al. [29] | BBBC041V1:1364 images MP-IDB:229 images | P. falciparum P. vivax | Binary | VGG-16,VGG-19, Xception, ResNet-50, customised CNN | Best performing model: VGG19 Accuracy: 99.35% F1-Score: 96.85% Sensitivity: 92.31% Specificity: 99.76% AUC: 96.03% Cross Dataset validation: Accuracy: 85.18% Sensitivity: 70.19% Specificity: 100% F1-Score: 84.82% AUC: 85.09% |
Yang et al. [30] | 2567 thin blood smear images | P. vivax | Binary | YOLOv2 | 79.22% |
Krishnadas et al. [31] | MP-IDB | P. falciparum P. vivax P. malariae P. ovale | Multi-class | Yolov5 and Scaled Yolov4 | Best performing model: Scaled Yolov4 Parasite classification: Accuracy: 83% |
Sukumarran et al. [32] | MP-IDB and Malaria Research Centre, Unimas Sarawak | P. falciparum P. vivax P. malariae P. ovale P. knowlesi | Binary | YOLOv4, Faster R-CNN, SSD-300 | Best performing model: YOLOv4 mAP: 93.87% Cross dataset: mAP: 84.04% |