Glomeruli Detection Using Faster R-CNN and CenterNet

Histopathology is a complex discipline that requires the expertise of skilled pathologists to analyze tissue specimens accurately and provide guidance to clinicians in the management of a patient's healthcare. Digital histopathology has revolutionized the field of pathology and has indeed offer...

Full description

Saved in:
Bibliographic Details
Published in2023 3rd Asian Conference on Innovation in Technology (ASIANCON) pp. 1 - 6
Main Authors P, Jasitha, N, Pournami P
Format Conference Proceeding
LanguageEnglish
Published IEEE 25.08.2023
Subjects
Analytical models
Biopsy
digital histopathology
Faster R-CNN
glomeruli detection
Histopathology
Location awareness
Medical diagnosis
Object detection
Technological innovation
Online AccessGet full text

Cover

More Information
Summary:Histopathology is a complex discipline that requires the expertise of skilled pathologists to analyze tissue specimens accurately and provide guidance to clinicians in the management of a patient's healthcare. Digital histopathology has revolutionized the field of pathology and has indeed offered a helping hand to pathologists. The integration of Artificial Intelligence(AI) based digital pathology not only streamlines the pathology workflow but also offers a more comprehensive and individualized perspective, empowering pathologists to better manage the progression of intricate diseases. The renal biopsy is an essential diagnostic tool for identifying glomerular diseases. The detection and localization of glomeruli are crucial in the initial stages of diagnosing kidney diseases. Accurate counting of glomeruli in WSIs can aid in research studies aimed at under-standing the pathophysiology of kidney diseases and developing new treatments. This work aims at finding the best possible object detection mechanism to accomplish the automated detection of glomeruli in renal disease diagnosis. The experiment is carried out with the variations of developing Faster R-CNN which takes the input of single glomeruli and, the Faster R-CNN model and CenterNet model which takes WSI patches with multiple glomeruli separately. This study is a comparative analysis among the developed models and the results obtained show that Faster R-CNN is achieving promising results with an average IoU of 64.2% and an mAP of 65.7% with WSI image patches.
DOI:10.1109/ASIANCON58793.2023.10270511