`ChemoPADNNtraining2024` Dataset

The ChemoPADNNtraining2024 Dataset is a curated collection of Paper Analytical Device (PAD) images used for chemotherapy drug identification and analysis.

Description

This dataset consists of 3,609 images from 624 unique samples, divided into development and test sets. The dataset is designed for training machine learning models to identify and analyze various chemotherapy drugs and related substances using PAD technology. The Active Pharmaceutical Ingredients (APIs) are available in one of four concentrations (0%, 33%, 66%, 100%), providing a comprehensive range for concentration-dependent analysis.

Data Distribution

The dataset is split into:

Development/Validation Set: 2,847 images (492 unique samples)
Test Set: 762 images (132 unique samples)

Class Distribution

	class	#dev	#test	#total
0	hydroxyurea	561	141	702
1	mesna	509	128	637
2	doxorubicin	474	118	592
3	cisplatin	421	105	526
4	mtx-oral	282	71	353
5	oxaliplatin-2-mg-ml	248	62	310
6	oxaliplatin-5-mg-ml	210	52	262
7	mtx-injectable	141	35	176
8	lactose	41	10	51
-	#total	2891	723	3614

Dataset Visualizations

All Data Distribution

Development Set Distribution

Test Set Distribution

Camera and Drug Distribution Camera and Drug Heatmap

Directory Structure

The dataset is organized within the datasets/Leiberman-Lab_ChemoPADNNtraining2024_Partial-Drug-Set_v1.0 directory. Below is an overview of its structure and contents:

datasets/Leiberman-Lab_ChemoPADNNtraining2024_Partial-Drug-Set_v1.0/
├── metadata_dev.csv
├── metadata_test.csv
├── labels.csv
├── projects.csv
├── class_distribution.csv
├── dataset_sizes.md
├── croissant.jsonld
└── figs/
    ├── camera_drug_heatmap.png
    ├── dev__tsize_20.png
    ├── test__tsize_20.png
    └── all_data.png

PAD Dataset Registry