Comparison of YOLO and transformer based tumor detection in cystoscopy

Background: Bladder cancer (BCa) is the second most common type of cancer in the genitourinary system and causes approximately 165,000 deaths each year. The diagnosis of BCa is primarily done through cystoscopy, which involves visually examining the bladder using an endoscope. Currently, white light...

Full description

Saved in:
Bibliographic Details
Main Authors: Eixelberger Thomas, Maisch Philipp, Belle Sebastian, Kriegmaier Maximilian, Bolenz Christian, Wittenberg Thomas
Format: Article
Language:English
Published: De Gruyter 2024-12-01
Series:Current Directions in Biomedical Engineering
Subjects:
Online Access:https://doi.org/10.1515/cdbme-2024-2055
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850032100012457984
author Eixelberger Thomas
Maisch Philipp
Belle Sebastian
Kriegmaier Maximilian
Bolenz Christian
Wittenberg Thomas
author_facet Eixelberger Thomas
Maisch Philipp
Belle Sebastian
Kriegmaier Maximilian
Bolenz Christian
Wittenberg Thomas
author_sort Eixelberger Thomas
collection DOAJ
description Background: Bladder cancer (BCa) is the second most common type of cancer in the genitourinary system and causes approximately 165,000 deaths each year. The diagnosis of BCa is primarily done through cystoscopy, which involves visually examining the bladder using an endoscope. Currently, white light cystoscopy is considered the most reliable method for diagnosis. However, it can be challenging to detect and diagnose flat, small, or poorly textured lesions. The study explores the performance of deep learning systems (YOLOv7- tiny, RT-DETR18), originally designed for detecting adenomas in colonoscopy images, when retrained and tested with cystoscopy images. The deep neural network used in the study was pre-trained on 35,699 colonoscopy images (some from Mannheim) and both architectures achieved a F1 score of 0.91 on publicly available colonoscopy datasets. Results: When the adenoma-detection network was tested with cystoscopy images from two sources (Ulm and Erlangen), F1 scores ranging from 0.58 to 0.81 were achieved. Subsequently, the networks were retrained using 12,066 cystoscopy images from Mannheim, resulting in improved F1 scores ranging from 0.77 to 0.85. Conclusion: It could be shown that transformer based networks perform slightly better than YOLOv7-tiny networks, but both network types are feasable for lesion detection in the human bladder. The retraining of the network with additional cystoscopy data led to an improvement in the performance of urinary lesion detection. This suggests that it is possible to achieve a domain-shift with the inclusion of appropriate additional data.
format Article
id doaj-art-c8f8cbd98a744eb1a238e72b76fbb0a7
institution DOAJ
issn 2364-5504
language English
publishDate 2024-12-01
publisher De Gruyter
record_format Article
series Current Directions in Biomedical Engineering
spelling doaj-art-c8f8cbd98a744eb1a238e72b76fbb0a72025-08-20T02:58:46ZengDe GruyterCurrent Directions in Biomedical Engineering2364-55042024-12-0110422823110.1515/cdbme-2024-2055Comparison of YOLO and transformer based tumor detection in cystoscopyEixelberger Thomas0Maisch Philipp1Belle Sebastian2Kriegmaier Maximilian3Bolenz Christian4Wittenberg Thomas5Friedrich-Alexander-Universitat Erlangen-Nurnberg & Fraunhofer IIS,Erlangen, GermanyDepartment of Urology, University Hospital Ulm,Ulm, GermanyUniversitats-Medizin,Mannheim, GermanyUrology Clinics Munchen-Planegg,Munchen, GermanyDepartment of Urology, University Hospital Ulm,Ulm, GermanyFriedrich-Alexander-Universitat Erlangen-Nurnberg & Fraunhofer IIS,Erlangen, GermanyBackground: Bladder cancer (BCa) is the second most common type of cancer in the genitourinary system and causes approximately 165,000 deaths each year. The diagnosis of BCa is primarily done through cystoscopy, which involves visually examining the bladder using an endoscope. Currently, white light cystoscopy is considered the most reliable method for diagnosis. However, it can be challenging to detect and diagnose flat, small, or poorly textured lesions. The study explores the performance of deep learning systems (YOLOv7- tiny, RT-DETR18), originally designed for detecting adenomas in colonoscopy images, when retrained and tested with cystoscopy images. The deep neural network used in the study was pre-trained on 35,699 colonoscopy images (some from Mannheim) and both architectures achieved a F1 score of 0.91 on publicly available colonoscopy datasets. Results: When the adenoma-detection network was tested with cystoscopy images from two sources (Ulm and Erlangen), F1 scores ranging from 0.58 to 0.81 were achieved. Subsequently, the networks were retrained using 12,066 cystoscopy images from Mannheim, resulting in improved F1 scores ranging from 0.77 to 0.85. Conclusion: It could be shown that transformer based networks perform slightly better than YOLOv7-tiny networks, but both network types are feasable for lesion detection in the human bladder. The retraining of the network with additional cystoscopy data led to an improvement in the performance of urinary lesion detection. This suggests that it is possible to achieve a domain-shift with the inclusion of appropriate additional data.https://doi.org/10.1515/cdbme-2024-2055deep learningyolotransformerscystoscopylesion detection
spellingShingle Eixelberger Thomas
Maisch Philipp
Belle Sebastian
Kriegmaier Maximilian
Bolenz Christian
Wittenberg Thomas
Comparison of YOLO and transformer based tumor detection in cystoscopy
Current Directions in Biomedical Engineering
deep learning
yolo
transformers
cystoscopy
lesion detection
title Comparison of YOLO and transformer based tumor detection in cystoscopy
title_full Comparison of YOLO and transformer based tumor detection in cystoscopy
title_fullStr Comparison of YOLO and transformer based tumor detection in cystoscopy
title_full_unstemmed Comparison of YOLO and transformer based tumor detection in cystoscopy
title_short Comparison of YOLO and transformer based tumor detection in cystoscopy
title_sort comparison of yolo and transformer based tumor detection in cystoscopy
topic deep learning
yolo
transformers
cystoscopy
lesion detection
url https://doi.org/10.1515/cdbme-2024-2055
work_keys_str_mv AT eixelbergerthomas comparisonofyoloandtransformerbasedtumordetectionincystoscopy
AT maischphilipp comparisonofyoloandtransformerbasedtumordetectionincystoscopy
AT bellesebastian comparisonofyoloandtransformerbasedtumordetectionincystoscopy
AT kriegmaiermaximilian comparisonofyoloandtransformerbasedtumordetectionincystoscopy
AT bolenzchristian comparisonofyoloandtransformerbasedtumordetectionincystoscopy
AT wittenbergthomas comparisonofyoloandtransformerbasedtumordetectionincystoscopy