SSANet: Side-by-Side Additive Network for Knee Osteoarthritis Severity Detection from X-Ray Images

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F62690094%3A18450%2F24%3A50021554" target="_blank" >RIV/62690094:18450/24:50021554 - isvavai.cz</a>

Výsledek na webu

<a href="http://dx.doi.org/10.1007/978-981-97-2611-0_24" target="_blank" >http://dx.doi.org/10.1007/978-981-97-2611-0_24</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1007/978-981-97-2611-0_24" target="_blank" >10.1007/978-981-97-2611-0_24</a>

Jazyk výsledku

angličtina

Název v původním jazyce

SSANet: Side-by-Side Additive Network for Knee Osteoarthritis Severity Detection from X-Ray Images

Popis výsledku v původním jazyce

Knee Osteoarthritis (KOA) is becoming one of the most frequent degenerative and irreversible diseases in elderly people. Its early detection and diagnosis is also a difficult and time-intensive task. The recent advances in Machine Learning (ML) and Computer Vision (CV) created new paths to automatically detect KOA in the early stages. Detecting and grading the severity of KOA requires a dedicated ML model. This paper proposes a Side-by-Side Additive Network (SSANet)-based model for detecting and grading the severity of KOA. Since KOA affects the knee joint space by narrowing the gap between the joints, we preprocess the X-ray images to detect the Region of Interest (ROI) before the training of the model. From experimental results, it is evidenced that the ROI selection is really improving the detection accuracy of the model. Unlike sequentially connected convolution layer-based deep learning models, the proposed SSANet is based on a parallel convolution layer to reduce the degradation of feature quality. Again, to enhance the edge feature maps, we apply an additive layer that couples the previous layers’ convolutional output. SSANet captures both high-level and low-level features to predict the KOA severity effectively. The presence of a smaller number of layers and fewer numbers of learnable parameters than the existing popular deep learning models, make SSANet a time and space-efficient network. Moreover, the performance of the proposed SSANet in detecting KOA severity is significantly superior to that of other popular state-of-the-art networks.

Název v anglickém jazyce

SSANet: Side-by-Side Additive Network for Knee Osteoarthritis Severity Detection from X-Ray Images

Popis výsledku anglicky

Knee Osteoarthritis (KOA) is becoming one of the most frequent degenerative and irreversible diseases in elderly people. Its early detection and diagnosis is also a difficult and time-intensive task. The recent advances in Machine Learning (ML) and Computer Vision (CV) created new paths to automatically detect KOA in the early stages. Detecting and grading the severity of KOA requires a dedicated ML model. This paper proposes a Side-by-Side Additive Network (SSANet)-based model for detecting and grading the severity of KOA. Since KOA affects the knee joint space by narrowing the gap between the joints, we preprocess the X-ray images to detect the Region of Interest (ROI) before the training of the model. From experimental results, it is evidenced that the ROI selection is really improving the detection accuracy of the model. Unlike sequentially connected convolution layer-based deep learning models, the proposed SSANet is based on a parallel convolution layer to reduce the degradation of feature quality. Again, to enhance the edge feature maps, we apply an additive layer that couples the previous layers’ convolutional output. SSANet captures both high-level and low-level features to predict the KOA severity effectively. The presence of a smaller number of layers and fewer numbers of learnable parameters than the existing popular deep learning models, make SSANet a time and space-efficient network. Moreover, the performance of the proposed SSANet in detecting KOA severity is significantly superior to that of other popular state-of-the-art networks.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

—

Návaznosti

S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2024

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Proceedings of 4th International Conference on Frontiers in Computing and Systems

ISBN

978-981-9726-10-3

ISSN

2367-3370

e-ISSN

2367-3389

Počet stran výsledku

12

Strana od-do

349-360

Název nakladatele

Springer Singapore

Místo vydání

Singapore

Místo konání akce

Mandi, India

Datum konání akce

16. 10. 2023

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—