A clinical microscopy dataset to develop a deep learning diagnostic test for urinary tract infection
Natasha Liou1,2, Trina De3,4, Adrian Urbanski3,4
1Bladder Infection and Immunity Group (BIIG), UCL Centre for Kidney & Bladder Health, Division of Medicine, University College London, Royal Free Hospital Campus, London, UK.
Abstract:
Urinary tract infection (UTI) is a common disorder. Its diagnosis can be made by microscopic examination of voided urine for markers of infection. This manual technique is technically difficult, time-consuming and prone to inter-observer errors. The application of computer vision to this domain has been slow due to the lack of a clinical image dataset from UTI patients. We present an open dataset containing 300 images and 3,562 manually annotated urinary cells labelled into seven classes of clinically significant cell types. It is an enriched dataset acquired from the unstained and untreated urine of patients with symptomatic UTI using a simple imaging system. We demonstrate that this dataset can be used to train a Patch U-Net, a novel deep learning architecture with a random patch generator to recognise urinary cells. Our hope is, with this dataset, UTI diagnosis will be made possible in nearly all clinical settings by using a simple imaging system which leverages advanced machine learning techniques.
Insights
This study introduces a new dataset of annotated urinary cells to aid in diagnosing urinary tract infections (UTIs). This resource enables machine learning for faster, more accessible UTI diagnosis using simple imaging systems.
Area of Science:
- Medical Imaging
- Computer Vision
- Machine Learning
Background:
- Urinary tract infection (UTI) diagnosis relies on manual urine microscopy, which is labor-intensive and error-prone.
- A lack of clinical image datasets has hindered the adoption of computer vision for UTI diagnosis.
Purpose of the Study:
- To present an open-access dataset of annotated urinary cells from UTI patients.
- To demonstrate the utility of this dataset in training a deep learning model for automated cell recognition.
Main Methods:
- An open dataset of 300 images with 3,562 manually annotated urinary cells (seven classes) was created from unstained urine of symptomatic UTI patients.
- A novel deep learning architecture, Patch U-Net with a random patch generator, was trained using this dataset.
Main Results:
- The Patch U-Net model was successfully trained to recognize urinary cells.
- The dataset facilitates the development of automated UTI diagnostic tools.
Conclusions:
- This annotated dataset is a valuable resource for advancing computer-aided UTI diagnosis.
- The developed deep learning approach shows promise for near-universal UTI diagnosis in clinical settings using simple imaging.
More Related Videos
Related Concept Videos
01:26Urinary Tract Infection I: Introduction
01:25Urinary Tract Infection II: Pathophysiology
01:30Urinary Tract Infection III: Diagnostic Studies and Interprofessional Care
01:26Urine Studies II: Urine Culture and Sensitivity Test
01:28Microbiota of the Urogenital Tract
01:24Automated Microbial Diagnostics