Advancing mold identification in the routine laboratory: Performance of smartphone-based imaging and a newly developed Convolutional Neural Network

Background: Mold identification in clinical diagnostics is traditionally labor intensive and is dependent on expert interpretation. MoldVision is a deep learning approach that uses smartphone images ofmold cultures to automate identification.
Methods: We analyzed 161 clinical isolates across four common mold genera. Penicillium spp., Aspergillus spp. (with A. flavus and A. fumigatus), Fusarium spp., and Cladosporium spp. Daily images were captured from the top and bottom of culture plates over five days using a standardized smartphone setup, generating over 4,000 images. We trained three variations of VGG16 convolutional neural networks (CNN) and benchmarked the best-performing model (VGG16 with dual classification heads) against LightGBM models trained on pre-extracted features and human expert assessments at various time points.
Results: The best performing VGG16 model achieved a mean (SD) ROC-AUC of 92.7% ± 1.8% and sensitivity of 68.7% ±2.6% across all species. Here, the performance..., , # Data from: Advancing mold identification in the routine laboratory: Performance of smartphone-based imaging and a newly developed Convolutional Neural Network

Dataset DOI: [10.5061/dryad.cjsxksnj4](10.5061/dryad.cjsxksnj4)

## Description of the data and file structure

This dataset contains image files collected from *Penicillium spp.*, *Aspergillus spp.* (with *Aspergillus flavus* and *Aspergillus fumigatus*), *Fusarium spp.*, and *Cladosporium spp.* isolates over a 5 day period from the top and bottom of a Sabouraud Agar Plate. Each file is named according to a standardized convention to facilitate identification and reuse.

### Files and variables

#### File: moldvision_database.csv

**Description:** A CSV file for ease of access for Machine Learning applications, as well as providing a database for the image files

##### Variables

* filename
* class: species of fungal mold
* top/bottom: orientation of the agar plate in the picture (\"top\" or \"bottom\")
* ID: three‐digit subject i...,

背景：临床诊断中的霉菌鉴定传统上工作强度较高，且依赖专家的主观解读。MoldVision是一种深度学习方法，利用霉菌培养物的智能手机图像实现自动化鉴定。

方法：本研究分析了覆盖4个常见霉菌属的161株临床分离株，包括青霉属（*Penicillium spp.*）、曲霉属（*Aspergillus spp.*，包含黄曲霉（*A. flavus*）与烟曲霉（*A. fumigatus*））、镰刀菌属（*Fusarium spp.*）以及枝孢菌属（*Cladosporium spp.*）。采用标准化智能手机成像装置，连续5天每日从培养皿的顶面和底面采集图像，共生成超过4000张图像。我们训练了3种变体的VGG16卷积神经网络（CNN），并将性能最优的模型（带双分类头的VGG16）与基于预提取特征训练的LightGBM模型，以及不同时间点的人类专家评估结果进行了基准对比。

结果：性能最优的VGG16模型在所有菌种上的平均（标准差）受试者工作特征曲线下面积（ROC-AUC）为92.7%±1.8%，灵敏度为68.7%±2.6%。此处性能数据……
# 数据来源：《推进常规实验室霉菌鉴定：基于智能手机成像与新型卷积神经网络的性能表现》

数据集DOI：[10.5061/dryad.cjsxksnj4](10.5061/dryad.cjsxksnj4)

数据集描述与文件结构
本数据集包含5天内从沙氏琼脂平板（Sabouraud Agar Plate）顶面和底面采集的青霉属、曲霉属（包含黄曲霉与烟曲霉）、镰刀菌属及枝孢菌属分离株的图像文件。所有文件均遵循标准化命名规范，便于识别与复用。

### 文件与变量
#### 文件：moldvision_database.csv
**描述**：该CSV文件便于机器学习应用调用，同时为图像文件提供数据库支撑。
##### 变量
* filename：文件名
* class：真菌霉菌的菌种类别
* top/bottom：照片中培养皿的朝向（"top"即顶面或"bottom"即底面）
* ID：三位数字的受试者编号……