Thermal AI Open Source datasets: Part I

on

Are you intrigued by the potential of thermal imaging for clinical applications and eager to delve into developing AI algorithms using open source datasets? Look no further! In this Part I of the blog, we'll introduce you to a diverse thermal AI datasets tailored for different clinical use cases. These datasets can serve as foundational resources for developing and training initial versions of AI algorithms aimed at revolutionizing healthcare by combining AI with this affordable, non-invasive, non-contact and radiation free thermal imaging.

.

1. Thermal AI Datasets for Breast Cancer

Database For Mastology Research (DMR) for Breast Cancer Detection

DMR database is an open source dataset from Visual Lab, Universidade Federal Fluminense. It consists of breast thermal imaging data of 200+ participants along with their medical records and diagnosis from standard imaging tests such as mammography.  Each participant data include breast thermal images at five standard views along with series of thermal images capture over different time stamps during the imaging.


                             

DBT-TU-JU Breast Thermogram Dataset

 DBT-TU-JU Breast Thermogram dataset is from Regional Cancer Centre (RCC), Agartala Government Medical College (AGMC) of Govind Ballav Pant (GBP) Hospital, Agartala, Tripura, India. The includes 1100 thermograms from 100 subjects captured at six different views involving the five standard views and supine view. Clinical findings, X-ray mammography and Fine Needle Aspiration Cytology (FNAC) reports are available to obtain the final clinical diagnosis.



Using the above thermal AI datasets, researchers can tackle various AI problem statements as listed below related to breast cancer research. if you want more information on these problem statements, read the Thermal Imaging for Low-cost, Portable, Non-invasive and Privacy Aware Breast Cancer Detection
  1. Classification of Breast Cancer from Thermal Images
  2. Segmentation of Breast Region of Interest
  3. Localization of Abnormal Regions in Thermal Images
  4. 3D Surface Generation
  5. 3D Internal Modeling

2. Thermal AI Datasets for COVID-19

NTIC COVID-19 Screening Dataset


NTIC COVID-19 Screening dataset is an open source dataset from Niramai Health Analytix. This dataset was collected on Indian population during COVID-19.  Broadly, This comprehensive dataset is a consortium of multiple datasets for different COVID-19 screening purposes.

(a) Thermal Surveillance Dataset: Thermal Surveillance Dataset comprises a rich collection of thermal images capturing individuals who walked-in into public premises such as offices, malls, and railway stations. It comprises a total of 902 thermal images, where each image within this dataset may feature more than one individual, providing a realistic representation of crowded public spaces.  A total of 1,567 individuals are part of this dataset.

(b) Augmented Surveillance Dataset: This dataset comprises thermal-like images generated from visual RGB images using data transformations. Overall, this dataset comprises 543 images with 434 people wearing masks and 109 people without masks.

(c) Lighting Dataset: This is a controlled dataset that captured both RGB and Thermal images at same time and only one person at a time under 4 different lighting conditions. Overall,  420 thermal images from 25 participants under 4 different lighting conditions were part of this dataset. For each of these thermal images, corresponding visual images are also captured.

All the above datasets comprise the bounding box regions corresponding to individual faces for all the individuals along with whether each person is wearing masks or not. This allows researchers to develop ML algorithms for below problem statements. For more information on these problem statements, read the blog Role of Thermal Imaging during COVID-19 Pandemic

  1. Identification of face bounding boxes from thermal images
  2. Detection of Mask from thermal images
  3. Synthesis of thermal-like images given a visual image


3. Thermal AI Datasets for COVID-19

Niramai Oncho Dataset

This is an open source dataset from Niramai Health Analytix that consists of thermal images and videos of different body parts consisting of palpable nodules along with their histopathological statuses confirming the presence of female adult worm.

Overall, the dataset comprised thermal data (images/videos) from 125 participants captured for different body parts. In total, the dataset encompasses thermal data for 192 distinct body parts with 101 corresponding to live female nodules and the remain 91 corresponding to dead nodules.

This dataset can be leveraged  to develop AI-driven solutions tailored to address the above discussed problem statements in onchocerciasis detection and management. By employing machine learning techniques such as convolutional neural networks (CNNs) for image classification and segmentation, researchers can train models to accurately classify thermal images based on the presence of live female nodules, localize nodules within thermal images, and even estimate the reproductive status of the parasites.





Using the above thermal AI dataset, researchers can work on different AI problem statements as listed below. if you want more information on these problem statements, read the Thermal Imaging & AI for Onchocerciasis
  1. Triage individuals for Onchocerciasis Detection
  2. Localization of Onchocerca Nodules
  3. Fecundity Estimation of Onchocerca Nodules

This concludes Part I of our blog series. Stay tuned for Part II, where we'll unveil additional thermal AI datasets tailored for diverse clinical applications. Together, let's harness the transformative potential of thermal imaging and artificial intelligence in healthcare.


Comments

Post a Comment