sc890 commited on
Commit
2e68f30
1 Parent(s): ef7d2ce

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +81 -83
README.md CHANGED
@@ -42,88 +42,86 @@ configs:
42
  path: data/Labels_*
43
  ---
44
 
45
- # Multimodal Dataset of Tuberculosis Patients including CT and Clinical Case Reports
46
-
47
- Zhankai Ye <br>
48
- NetID: zy172
49
-
50
- ## Dataset Summary
51
- This dataset is curated from the original “The MultiCaRe Dataset” to focus on the chest tuberculosis patients. This is a multimodal dataset consisting of lung computed tomography (CT) imaging data and the clinical case records of tuberculosis patients, along with their case keywords, the captions of their CT images, patient_id, gender, and age information.
52
-
53
- ## Dataset Sources
54
- - Homepage: https://zenodo.org/records/10079370
55
- - DOI: 10.5281/zenodo.10079370
56
- - Data article: https://www.sciencedirect.com/science/article/pii/S2352340923010351
57
-
58
- ## Supported Tasks:
59
- This dataset can be utilized for:
60
- - Developing algorithms of the segmentation of chest CT images and the classification of tuberculosis positive or control.
61
- - Developing novel natural language processing (NLP) methods and unsupervised machine learning methods to extract medical terms from clinical notes.
62
-
63
- ## Languages:
64
- English
65
-
66
- ## Data Structure and Instance:
67
- The data will follow the structure below:
68
- {
69
- - `"case_id"`: "PMC10129030_01",
70
- - `"gender"`: "male",
71
- - `"age"`: 62,
72
- - `"case_text"`: "A 62-year-old man presented with acute dyspnea at rest, requiring high-flow…",
73
- - `"keywords"`: "["dendriform pulmonary ossification", "lung transplant", "pulmonary fibrosis"]",
74
- - `"pics_array"`: image
75
- - `"Caption"`: "coronal. chest CT shows ground-glass and reticular opacities in the dependent…"
76
- }
77
-
78
- ## Data Fields:
79
- - **case_id (string)**: ID of the patient, created combining the PMC of the article plus a sequential number.
80
- - **gender (string)**: Gender of the patient. It can be either Female, Male, Transgender or Unknown.
81
- - **age (int)**: Age of the patient. Ages lower than 1 y.o. are assigned 0 as age.
82
- - **case_text (string)**: Self-explanatory.
83
- - **keywords (string)**: Keywords are taken from the keywords section that is sometimes available in the content of the article.
84
- - **pics_array (int)**: image
85
- - **Caption (string)**: Image caption.
86
-
87
- ## Initial Data Collect and Preprocessing
88
- 1. The original MultiCaRe Dataset, approximately 9GB in size, encompasses a diverse range of medical specialties including oncology, cardiology, surgery, and pathology. To create your tuberculosis-focused subset, the dataset undergoes a filtering process based on specific criteria:
89
- - Case Report Selection: The selection criterion for case reports is the presence of keywords such as 'tuberculosis' or 'tb'. This ensures that only reports relevant to tuberculosis are included.
90
- - Caption Filtering: The dataset is further refined by filtering captions that contain keywords like 'ct', 'lung', or 'chest'.
91
- - Image Labeling: Finally, the images are chosen based on the presence of labels 'ct' and 'lung'. This dual-label requirement ensures that the selected images are relevant to CT scans of the lungs, which are instrumental in detecting and assessing tuberculosis.
92
- - Through this meticulous filtering process, an initial tuberculosis dataset is compiled from the broader MultiCaRe Dataset. This dataset is messy, contains many diffferent files.
93
- 2. To enhance the quality and relevance of the tuberculosis dataset, additional processing steps are implemented after in the Hugging Face python script after the initial filtration from the MultiCaRe Dataset:
94
- - Exclusion of Records with Missing Age Information.
95
- - Merge of data from difference files, including .csv, .JSON, and .jpg.
96
-
97
- ## Social Impact
98
- The multimodal dataset of tuberculosis patients, meticulously curated from the larger MultiCaRe Dataset, stands to have a significant social impact, particularly in the field of public health and medical research. Tuberculosis (TB) remains a major global health issue, especially in low- and middle-income countries, and the integration of CT imaging with clinical case reports in this dataset provides a rich resource for advanced diagnostic and treatment research. By facilitating the development of more precise algorithms for CT image segmentation and classification, as well as enhancing natural language processing (NLP) techniques for extracting medical terms from clinical notes, this dataset has the potential to improve the accuracy and efficiency of TB diagnosis.
99
-
100
- ## Personal and Sensitive Information
101
- Case reports are designed with the intention of being publicly accessible, and as a result, they deliberately omit any personal identifying details of the patients to ensure their privacy and confidentiality.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
102
 
103
  ## Bias, Risks, and Limitations
104
- ### Bias
105
- 1. Selection Bias: The original MultiCaRe Dataset was generated from 75,382 open access PubMed Central articles spanning the period from 1990 to 2023. Therefore, the random sampling of the cases from difference demographic groups cannot be guaranteed. The data may have bias as the collection process was not representative of the broader population. For example, the dataset may predominantly includes cases from a specific geographic location, age group, or socioeconomic status, and the findings may not apply to other groups.
106
- 2. Technology Bias: Advanced imaging technologies might not be equally available in all settings, leading to a dataset that disproportionately represents patients from better-equipped facilities. This can skew the dataset towards conditions that are more likely to be diagnosed in such settings.
107
- 3. Interpreter Bias: For the `"case_text"` and the `"caption"`, variability in the expertise and experience of radiologists or clinicians interpreting the images can lead to differences in diagnosis or findings reported in the dataset.
108
-
109
- ### Risks
110
- 1. Privacy and Confidentiality Risks: Patient data, including case records and images, are highly sensitive. There's a risk of identifying individuals even if the data is properly anonymized.
111
- 2. Data Integrity and Quality Risks: Inaccuracies, missing data, and inconsistencies within the dataset can compromise the validity of research findings or clinical decisions based on the data. This could lead to ineffective or harmful interventions.
112
-
113
- ### Limitations
114
- - Data Quality:
115
- 1. For textual data, certain patient records are missing key descriptive terms. Meanwhile, cases where imaging studies were not conducted lack both the images and their respective descriptive captions.
116
- 2. Regarding images, a primary concern is also the incomplete nature of the dataset, as images do not accompany all patient records. Additionally, the image resolution varies, which can impede detailed examination. The inconsistency in image sizes and variations in the positioning of patient photographs may also pose challenges for consistent image analysis.
117
-
118
- ## Citation
119
- ```bibtex
120
- @dataset{NievasOffidani2023MultiCaRe,
121
- author = {Nievas Offidani, M. and Delrieux, C.},
122
- title = {The MultiCaRe Dataset: A Multimodal Case Report Dataset with Clinical Cases, Labeled Images and Captions from Open Access PMC Articles},
123
- year = {2023},
124
- version = {1.0},
125
- publisher = {Zenodo},
126
- doi = {10.5281/zenodo.10079370},
127
- url = {https://doi.org/10.5281/zenodo.10079370},
128
- }
129
- ```
 
42
  path: data/Labels_*
43
  ---
44
 
45
+ # DeepFruit Dataset
46
+
47
+ <!--The dataset is from Mendeley, comprises 21,122 images of 20 diverse fruit types across 8 different combinations and 2 csv files. -->
48
+
49
+
50
+ ## Dataset Details
51
+
52
+ This dataset contains total of 21,122 fully labeled images, featuring 20 different kinds of fruits. It is structured into an 80% training set (16,899 images) and a 20% testing set (4,223 images), facilitating a ready-to-use framework for model training and evaluation.
53
+
54
+ Additionally, there are two CSV files that label the types of fruits depicted in each image.
55
+
56
+ ### Dataset Description
57
+
58
+ The "DeepFruit" dataset is a comprehensive collection designed for the advancement of research in fruit detection, recognition, and classification. It encompasses a wide array of applications, including but not limited to, fruit recognition systems and calorie estimation. A total of 21,122 fully labeled images, featuring 20 different kinds of fruits. It is structured into an 80% training set (16,899 images) and a 20% testing set (4,223 images), facilitating a ready-to-use framework for model training and evaluation. This dataset provides a valuable resource for researchers aiming to develop automated systems leveraging deep learning, computer vision, and machine learning techniques for fruit image analysis.
59
+
60
+
61
+ - **Language(s):** en
62
+ - **License:** Mendeley License: CC BY 4.0
63
+
64
+ ### Dataset Sources
65
+
66
+ Data: https://data.mendeley.com/datasets/5prc54r4rt/1
67
+
68
+ Paper: https://www.sciencedirect.com/science/article/pii/S2352340923006248#sec0003
69
+
70
+ ## Uses
71
+
72
+ Convert Fruit Dataset From Image to PIL.
73
+
74
+ ### Direct Use
75
+
76
+ This section describes suitable use cases for the dataset.
77
+
78
+ ## Dataset Structure
79
+
80
+ "Train" & "Test": Datasets
81
+
82
+ "image_id": datasets.Value("string")
83
+
84
+ "number":datasets.Value("int32")
85
+
86
+ "image": datasets.Image()
87
+
88
+ "image_path": datasets.Value("string")
89
+
90
+ "label": datasets.Value("string")
91
+
92
+
93
+
94
+ ### Curation Rationale
95
+
96
+ It lies in its foundational role for enabling advanced machine learning applications in dietary and health management. By converting fruit images to the PIL format, it prepares data for analysis that could lead to innovations in recognizing and understanding fruit characteristics. This groundwork is crucial for developing technologies that assist in dietary planning, nutritional education, and managing health conditions through better food choices, thereby having a broad positive effect on public health and awareness.
97
+
98
+ #### Data Collection and Processing
99
+
100
+ Image Format: All images are expected to be in JPEG format. Non-JPEG files are excluded during the data processing phase, ensuring consistency in file format.
101
+
102
+ Label Extraction: Labels are extracted from separate CSV files (Labels_Train.csv and Labels_Test.csv), which map image names to their corresponding fruit labels. This method ensures that labels are organized and accessible.
103
+
104
+ Data Splitting: The dataset is split into training and testing sets, as indicated by the separate ZIP files for train and test data. This standard practice facilitates the evaluation of model performance on unseen data.
105
+
106
+ Python Imaging Library (PIL): Used for opening and manipulating images in the Python Imaging Library format. This choice is made for its wide adoption and ease of integration with other Python libraries for data science and machine learning tasks.
107
+
108
+ Datasets Library from Hugging Face: Facilitates the creation, distribution, and loading of the dataset. This library provides a standardized way to work with datasets, including features for splitting, processing, and accessing dataset information.
109
+
110
+ #### Supported Tasks
111
+
112
+ The fruit images were captured under various conditions, including different plate sizes, shapes, and situations, as well as varying angles, brightness levels, and distances.
113
+
114
+ 1. Foundation for Advanced ML Models/ Algorithms Training: By converting the fruit dataset into PIL format, we ensure that the data is in a uniform, accessible format that is compatible with various machine learning and deep learning libraries. This standardization is vital for the efficient training, validation, and testing of different classification models.
115
+
116
+ 2. Enables Comprehensive Analysis: The dataset, featuring a wide variety of fruit images, is essential for developing a deep understanding of fruit characteristics. This includes not only basic identification but also detailed analyses such as sugar content, calorie count, and vitamin composition, which are crucial for dietary planning and health management.
117
+
118
+ 3. Basis for Practical Applications: The dataset's conversion and subsequent use in machine learning model training are not academic exercises but are intended for real-world applications. The insights gained from this project could significantly impact dietary planning, particularly for individuals with specific health considerations like diabetes, by providing accurate, detailed information about fruit characteristics.
119
+
120
 
121
  ## Bias, Risks, and Limitations
122
+
123
+ Representation Bias: Given the dataset comprises 20 diverse fruit types across 8 combinations, there might be an underrepresentation of certain fruits, particularly those that are less common or indigenous to specific regions. This could lead to a model trained on this dataset performing less accurately on fruit types or varieties not included or underrepresented.
124
+
125
+ Misclassification Risk: In critical applications where accurate fruit identification is crucial (e.g., dietary management apps, agricultural sorting mechanisms), misclassification could lead to adverse outcomes. This risk is heightened if the dataset contains mislabeled examples or if the model struggles with fruits that have similar appearances.
126
+
127
+ Scope of Application: The dataset's utility is primarily confined to the domain of fruit recognition and classification. It may not be suitable for more nuanced tasks within agricultural technology, such as detecting fruit diseases or assessing ripeness, unless supplemented with additional, specialized data.