TY - GEN
T1 - Automated Detection of Malaria implemented by Deep Learning in Pytorch
AU - Krishnadas, Padmini
AU - Sampathila, Niranjana
N1 - Publisher Copyright:
© 2021 IEEE.
PY - 2021
Y1 - 2021
N2 - The diagnoses of diseases as widespread as malaria has proven to be tough in rural areas. This is because of the lack of resources and professionals working in these places. In such cases, and for urban areas as well, the automation of diagnoses can play a vital role in detecting, diagnosing, treating, and preventing malaria in such areas of the world. The automation procedure uses a computerised approach for the diagnosis of malaria from the acquired microscopic images of PBC (Peripheral Blood Smear) images. In this paper, we focus on deep learning algorithms implemented in Pytorch through transfer learning to detect malaria in segmented red blood cell images. The process involved using pre-trained ImageNet models (Namely ResNet and DenseNet) and fine-tuning them to the dataset at hand to classify cell images as either parasitized or uninfected. Reported here the results obtained. The ResNet50 model achieved an accuracy of 91.72%. The DenseNet121 model achieved the highest accuracy of 94.43%. The ResNet50 model performance parameters measured, and the specificity and sensitivity are respectively 89.03% and 88.91%. This branch of digital telepathology can enable the healthcare industry to distribute quality services even to unreachable and rural areas of the world.
AB - The diagnoses of diseases as widespread as malaria has proven to be tough in rural areas. This is because of the lack of resources and professionals working in these places. In such cases, and for urban areas as well, the automation of diagnoses can play a vital role in detecting, diagnosing, treating, and preventing malaria in such areas of the world. The automation procedure uses a computerised approach for the diagnosis of malaria from the acquired microscopic images of PBC (Peripheral Blood Smear) images. In this paper, we focus on deep learning algorithms implemented in Pytorch through transfer learning to detect malaria in segmented red blood cell images. The process involved using pre-trained ImageNet models (Namely ResNet and DenseNet) and fine-tuning them to the dataset at hand to classify cell images as either parasitized or uninfected. Reported here the results obtained. The ResNet50 model achieved an accuracy of 91.72%. The DenseNet121 model achieved the highest accuracy of 94.43%. The ResNet50 model performance parameters measured, and the specificity and sensitivity are respectively 89.03% and 88.91%. This branch of digital telepathology can enable the healthcare industry to distribute quality services even to unreachable and rural areas of the world.
UR - http://www.scopus.com/inward/record.url?scp=85123345485&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85123345485&partnerID=8YFLogxK
U2 - 10.1109/CONECCT52877.2021.9622608
DO - 10.1109/CONECCT52877.2021.9622608
M3 - Conference contribution
AN - SCOPUS:85123345485
T3 - Proceedings of CONECCT 2021: 7th IEEE International Conference on Electronics, Computing and Communication Technologies
BT - Proceedings of CONECCT 2021
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 7th IEEE International Conference on Electronics, Computing and Communication Technologies, CONECCT 2021
Y2 - 9 July 2021 through 11 July 2021
ER -