UltraLightSqueezeNet: A Deep Learning Architecture for Malaria Classification With Up to 54× Fewer Trainable Parameters for Resource Constrained Devices

Lightweight deep learning approaches for malaria detection have gained attention for their potential to enhance diagnostics in resource constrained environments. For our study, we selected SqueezeNet1.1 as it is one of the most popular lightweight architectures. SqueezeNet1.1 is a later version of S...

Full description

Saved in:
Bibliographic Details
Main Authors: Suresh Babu Nettur, Shanthi Karpurapu, Unnati Nettur, Likhit Sagar Gajja, Sravanthy Myneni, Akhil Dusi, Lalithya Posham
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/11007680/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849325378729934848
author Suresh Babu Nettur
Shanthi Karpurapu
Unnati Nettur
Likhit Sagar Gajja
Sravanthy Myneni
Akhil Dusi
Lalithya Posham
author_facet Suresh Babu Nettur
Shanthi Karpurapu
Unnati Nettur
Likhit Sagar Gajja
Sravanthy Myneni
Akhil Dusi
Lalithya Posham
author_sort Suresh Babu Nettur
collection DOAJ
description Lightweight deep learning approaches for malaria detection have gained attention for their potential to enhance diagnostics in resource constrained environments. For our study, we selected SqueezeNet1.1 as it is one of the most popular lightweight architectures. SqueezeNet1.1 is a later version of SqueezeNet1.0 and is 2.4 times more computationally efficient than the original model. We proposed and implemented three ultra-lightweight architecture variants to SqueezeNet1.1 architecture, namely Variant 1 (one fire module), Variant 2 (two fire modules), and Variant 3 (four fire modules), which are even more compact than SqueezeNetV1.1 (eight fire modules). These models were implemented to evaluate the best performing variant that achieves superior computational efficiency without sacrificing accuracy in malaria blood cell classification. The models were trained and evaluated using the NIH Malaria dataset. We assessed each model&#x2019;s performance based on metrics including accuracy, recall, precision, F1-score, and Area Under the Curve (AUC). The results show that the SqueezeNet1.1 model achieves the highest performance across all metrics, with a classification accuracy of 97.12%. Variant 3 (four fire modules) offers a competitive alternative, delivering almost identical results (accuracy 96.55%) with a <inline-formula> <tex-math notation="LaTeX">$6\times $ </tex-math></inline-formula> reduction in computational overhead compared to SqueezeNet1.1. Variant 2 and Variant 1 perform slightly lower than Variant 3, with Variant 2 (two fire modules) reducing computational overhead by <inline-formula> <tex-math notation="LaTeX">$28\times $ </tex-math></inline-formula> and Variant 1 (one fire module) achieving a <inline-formula> <tex-math notation="LaTeX">$54\times $ </tex-math></inline-formula> reduction in trainable parameters compared to SqueezeNet1.1. These findings demonstrate that our SqueezeNet1.1 architecture variants provide a flexible approach to malaria detection, enabling the selection of a variant that balances resource constraints and performance.
format Article
id doaj-art-edd3ea0058d44f2c98adb56e56a6817b
institution Kabale University
issn 2169-3536
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-edd3ea0058d44f2c98adb56e56a6817b2025-08-20T03:48:26ZengIEEEIEEE Access2169-35362025-01-0113894288944010.1109/ACCESS.2025.357169611007680UltraLightSqueezeNet: A Deep Learning Architecture for Malaria Classification With Up to 54&#x00D7; Fewer Trainable Parameters for Resource Constrained DevicesSuresh Babu Nettur0https://orcid.org/0009-0007-7453-7870Shanthi Karpurapu1https://orcid.org/0009-0004-0957-2010Unnati Nettur2Likhit Sagar Gajja3Sravanthy Myneni4Akhil Dusi5Lalithya Posham6Independent Researcher, Virginia Beach, VA, USAIndependent Researcher, Virginia Beach, VA, USADepartment of Computer Science, Virginia Tech, Blacksburg, VA, USADepartment of Computer Science, BML Munjal University, Kapriwas, Haryana, IndiaIndependent Researcher, Virginia Beach, VA, USADepartment of Information Systems, Indiana Tech, Fort Wayne, IN, USASchool of International Education, Nanjing Medical University, Nanjing, Jiangsu, ChinaLightweight deep learning approaches for malaria detection have gained attention for their potential to enhance diagnostics in resource constrained environments. For our study, we selected SqueezeNet1.1 as it is one of the most popular lightweight architectures. SqueezeNet1.1 is a later version of SqueezeNet1.0 and is 2.4 times more computationally efficient than the original model. We proposed and implemented three ultra-lightweight architecture variants to SqueezeNet1.1 architecture, namely Variant 1 (one fire module), Variant 2 (two fire modules), and Variant 3 (four fire modules), which are even more compact than SqueezeNetV1.1 (eight fire modules). These models were implemented to evaluate the best performing variant that achieves superior computational efficiency without sacrificing accuracy in malaria blood cell classification. The models were trained and evaluated using the NIH Malaria dataset. We assessed each model&#x2019;s performance based on metrics including accuracy, recall, precision, F1-score, and Area Under the Curve (AUC). The results show that the SqueezeNet1.1 model achieves the highest performance across all metrics, with a classification accuracy of 97.12%. Variant 3 (four fire modules) offers a competitive alternative, delivering almost identical results (accuracy 96.55%) with a <inline-formula> <tex-math notation="LaTeX">$6\times $ </tex-math></inline-formula> reduction in computational overhead compared to SqueezeNet1.1. Variant 2 and Variant 1 perform slightly lower than Variant 3, with Variant 2 (two fire modules) reducing computational overhead by <inline-formula> <tex-math notation="LaTeX">$28\times $ </tex-math></inline-formula> and Variant 1 (one fire module) achieving a <inline-formula> <tex-math notation="LaTeX">$54\times $ </tex-math></inline-formula> reduction in trainable parameters compared to SqueezeNet1.1. These findings demonstrate that our SqueezeNet1.1 architecture variants provide a flexible approach to malaria detection, enabling the selection of a variant that balances resource constraints and performance.https://ieeexplore.ieee.org/document/11007680/Malaria NIH datasetlightweightcompactSqueezeNetfire moduleAlexNet
spellingShingle Suresh Babu Nettur
Shanthi Karpurapu
Unnati Nettur
Likhit Sagar Gajja
Sravanthy Myneni
Akhil Dusi
Lalithya Posham
UltraLightSqueezeNet: A Deep Learning Architecture for Malaria Classification With Up to 54&#x00D7; Fewer Trainable Parameters for Resource Constrained Devices
IEEE Access
Malaria NIH dataset
lightweight
compact
SqueezeNet
fire module
AlexNet
title UltraLightSqueezeNet: A Deep Learning Architecture for Malaria Classification With Up to 54&#x00D7; Fewer Trainable Parameters for Resource Constrained Devices
title_full UltraLightSqueezeNet: A Deep Learning Architecture for Malaria Classification With Up to 54&#x00D7; Fewer Trainable Parameters for Resource Constrained Devices
title_fullStr UltraLightSqueezeNet: A Deep Learning Architecture for Malaria Classification With Up to 54&#x00D7; Fewer Trainable Parameters for Resource Constrained Devices
title_full_unstemmed UltraLightSqueezeNet: A Deep Learning Architecture for Malaria Classification With Up to 54&#x00D7; Fewer Trainable Parameters for Resource Constrained Devices
title_short UltraLightSqueezeNet: A Deep Learning Architecture for Malaria Classification With Up to 54&#x00D7; Fewer Trainable Parameters for Resource Constrained Devices
title_sort ultralightsqueezenet a deep learning architecture for malaria classification with up to 54 x00d7 fewer trainable parameters for resource constrained devices
topic Malaria NIH dataset
lightweight
compact
SqueezeNet
fire module
AlexNet
url https://ieeexplore.ieee.org/document/11007680/
work_keys_str_mv AT sureshbabunettur ultralightsqueezenetadeeplearningarchitectureformalariaclassificationwithupto54x00d7fewertrainableparametersforresourceconstraineddevices
AT shanthikarpurapu ultralightsqueezenetadeeplearningarchitectureformalariaclassificationwithupto54x00d7fewertrainableparametersforresourceconstraineddevices
AT unnatinettur ultralightsqueezenetadeeplearningarchitectureformalariaclassificationwithupto54x00d7fewertrainableparametersforresourceconstraineddevices
AT likhitsagargajja ultralightsqueezenetadeeplearningarchitectureformalariaclassificationwithupto54x00d7fewertrainableparametersforresourceconstraineddevices
AT sravanthymyneni ultralightsqueezenetadeeplearningarchitectureformalariaclassificationwithupto54x00d7fewertrainableparametersforresourceconstraineddevices
AT akhildusi ultralightsqueezenetadeeplearningarchitectureformalariaclassificationwithupto54x00d7fewertrainableparametersforresourceconstraineddevices
AT lalithyaposham ultralightsqueezenetadeeplearningarchitectureformalariaclassificationwithupto54x00d7fewertrainableparametersforresourceconstraineddevices