Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Nasal Cytology Dataset for Object Detection and Deep Learning

Mauro Camporeale, Giovanni Dimauro, Matteo Gelardi, Giorgia Iacobellis, Mattia Sebastiano Ladisa, Sergio Latrofa, Nunzia Lomonte

TL;DR

This study addresses the slow, manual counting in nasal cytology by introducing the Nasal Cytology Dataset (NCD), an open benchmark containing over 10,000 bounding boxes across 500 field images to train object-detection models for nasal mucosa cells. The authors establish baselines using DETR and YOLOv8 for two tasks: cell recognition (multiclass cytotype labeling) and cell detection (general cell localization). Results show robust detection performance, with YOLOv8 outperforming DETR in recognition due to class imbalance, while both models achieve strong AP50 scores for detection. The work enables AI-aided rhinology and highlights the need for balanced data and potential two-stage detection-classification pipelines to improve minority cytotype recognition, laying groundwork for general-purpose liquid cell detectors in nasal cytology.

Abstract

Nasal Cytology is a new and efficient clinical technique to diagnose rhinitis and allergies that is not much widespread due to the time-consuming nature of cell counting; that is why AI-aided counting could be a turning point for the diffusion of this technique. In this article we present the first dataset of rhino-cytological field images: the NCD (Nasal Cytology Dataset), aimed to train and deploy Object Detection models to support physicians and biologists during clinical practice. The real distribution of the cytotypes, populating the nasal mucosa has been replicated, sampling images from slides of clinical patients, and manually annotating each cell found on them. The correspondent object detection task presents non'trivial issues associated with the strong class imbalancement, involving the rarest cell types. This work contributes to some of open challenges by presenting a novel machine learning-based approach to aid the automated detection and classification of nasal mucosa cells: the DETR and YOLO models shown good performance in detecting cells and classifying them correctly, revealing great potential to accelerate the work of rhinology experts.

A Nasal Cytology Dataset for Object Detection and Deep Learning

TL;DR

This study addresses the slow, manual counting in nasal cytology by introducing the Nasal Cytology Dataset (NCD), an open benchmark containing over 10,000 bounding boxes across 500 field images to train object-detection models for nasal mucosa cells. The authors establish baselines using DETR and YOLOv8 for two tasks: cell recognition (multiclass cytotype labeling) and cell detection (general cell localization). Results show robust detection performance, with YOLOv8 outperforming DETR in recognition due to class imbalance, while both models achieve strong AP50 scores for detection. The work enables AI-aided rhinology and highlights the need for balanced data and potential two-stage detection-classification pipelines to improve minority cytotype recognition, laying groundwork for general-purpose liquid cell detectors in nasal cytology.

Abstract

Nasal Cytology is a new and efficient clinical technique to diagnose rhinitis and allergies that is not much widespread due to the time-consuming nature of cell counting; that is why AI-aided counting could be a turning point for the diffusion of this technique. In this article we present the first dataset of rhino-cytological field images: the NCD (Nasal Cytology Dataset), aimed to train and deploy Object Detection models to support physicians and biologists during clinical practice. The real distribution of the cytotypes, populating the nasal mucosa has been replicated, sampling images from slides of clinical patients, and manually annotating each cell found on them. The correspondent object detection task presents non'trivial issues associated with the strong class imbalancement, involving the rarest cell types. This work contributes to some of open challenges by presenting a novel machine learning-based approach to aid the automated detection and classification of nasal mucosa cells: the DETR and YOLO models shown good performance in detecting cells and classifying them correctly, revealing great potential to accelerate the work of rhinology experts.
Paper Structure (12 sections, 3 equations, 7 figures, 4 tables)

This paper contains 12 sections, 3 equations, 7 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Nasal Cytology
  3. Dataset Construction
  4. Methods
  5. YOLO v8
  6. The DETR model
  7. Metrics
  8. Experiments and Results
  9. Cell Recognition
  10. Cell Detection
  11. Discussion
  12. Conclusions

Figures (7)

  • Figure 1: Examples of observable items with MGG staining.
  • Figure 2: Examples of particularly OD difficult cases
  • Figure 3: Example of sorting for IoU threshold = 0.5
  • Figure 4: Comparison between GTs (a,c,e) and DETR predictions (b,d,f) for the task of Cell Recognition.
  • Figure 5: Comparison between GTs (a,c,e) and YOLOv8 predictions (b,d,f) for the task of Cell Recognition.
  • ...and 2 more figures