New AI Model Can Detect Diseases by Analyzing Visual Maps
New research has facilitated the development of an artificial intelligence model that can detect illnesses by examining medical images. This model can accurately pick out tumors and provide reasons for every diagnosis given via visual maps. This tool’s development marks an important advancement in the healthcare technology field, because it will simplify the process of detecting illnesses as well as the diagnosis process.
The study was led by Beckman Institute graduate research assistant Sourya Sengupta, who highlighted that the study’s focus was on detecting cancer in its early stages.
The visual maps produced by the model, dubbed equivalency maps, are basically an altered version of a mammogram, X-ray or other medical image medium. Each region of the equivalency map is assigned a number, with the researchers noting that the higher values represent medically interesting regions that could have an anomaly. Once this is done, the model adds up all the values assigned, then forms a diagnosis.
Sengupta explained that the values allowed the physician to see which regions needed extra attention and investigate those areas, noting that the result was a more transparent system between patients and their doctors.
For their study, the researchers trained their artificial-intelligence model on a trio of different illness diagnosis tasks. To start with, the model analyzed mammograms and learned to detect early signs of a tumor. It then assessed a retina’s optical coherence tomography images, where it flagged Drusen, which could indicate macular degeneration. Once this was done, the model examined X-rays and learned to identify a condition that causes heart enlargement, leading to disease.
After the model was sufficiently trained, the investigators compared how the model performed to other black-box artificial intelligence systems, noting that it performed in a way similar to existing systems.
In their report, the researchers stated that the high rates of accuracy observed were a product of the model’s neural network. Sengupta also noted that AI could be helpful in countries where physicians were scarce, noting that this technology could also decrease wait times, optimize allocation of resources and improve patient outcomes.
The investigators hope that future AI models will be able to identify and diagnose anomalies in different regions of the body, not just specific areas. Mark Anastasio, the study’s principal investigator, added that he was excited about how the tool could directly benefit society, particularly when it came to enhancing trust between doctors and their patients.
The research’s findings were reported in “IEEE Transactions on Medical Imaging.”
With other companies such as Astrotech Corp. (NASDAQ: ASTC) working to bring to market more next-gen medical diagnostic tools, it is likely to get a lot easier to diagnose a wide variety of diseases with increased accuracy.
NOTE TO INVESTORS: The latest news and updates relating to Astrotech Corp. (NASDAQ: ASTC) are available in the company’s newsroom at https://ibn.fm/ASTC
Please see full terms of use and disclaimers on the BioMedWire website applicable to all content provided by BMW, wherever published or re-published: http://BMW.fm/Disclaimer
New research has facilitated the development of an artificial intelligence model that can detect illnesses by examining medical images. This model can accurately pick out tumors and provide reasons for every diagnosis given via visual maps. This tool’s development marks an important advancement in the healthcare technology field, because it will simplify the process of detecting illnesses as well as the diagnosis process.
The study was led by Beckman Institute graduate research assistant Sourya Sengupta, who highlighted that the study’s focus was on detecting cancer in its early stages.
The visual maps produced by the model, dubbed equivalency maps, are basically an altered version of a mammogram, X-ray or other medical image medium. Each region of the equivalency map is assigned a number, with the researchers noting that the higher values represent medically interesting regions that could have an anomaly. Once this is done, the model adds up all the values assigned, then forms a diagnosis.
Sengupta explained that the values allowed the physician to see which regions needed extra attention and investigate those areas, noting that the result was a more transparent system between patients and their doctors.
For their study, the researchers trained their artificial-intelligence model on a trio of different illness diagnosis tasks. To start with, the model analyzed mammograms and learned to detect early signs of a tumor. It then assessed a retina’s optical coherence tomography images, where it flagged Drusen, which could indicate macular degeneration. Once this was done, the model examined X-rays and learned to identify a condition that causes heart enlargement, leading to disease.
After the model was sufficiently trained, the investigators compared how the model performed to other black-box artificial intelligence systems, noting that it performed in a way similar to existing systems.
In their report, the researchers stated that the high rates of accuracy observed were a product of the model’s neural network. Sengupta also noted that AI could be helpful in countries where physicians were scarce, noting that this technology could also decrease wait times, optimize allocation of resources and improve patient outcomes.
The investigators hope that future AI models will be able to identify and diagnose anomalies in different regions of the body, not just specific areas. Mark Anastasio, the study’s principal investigator, added that he was excited about how the tool could directly benefit society, particularly when it came to enhancing trust between doctors and their patients.
The research’s findings were reported in “IEEE Transactions on Medical Imaging.”
With other companies such as Astrotech Corp. (NASDAQ: ASTC) working to bring to market more next-gen medical diagnostic tools, it is likely to get a lot easier to diagnose a wide variety of diseases with increased accuracy.
NOTE TO INVESTORS: The latest news and updates relating to Astrotech Corp. (NASDAQ: ASTC) are available in the company’s newsroom at https://ibn.fm/ASTC
Please see full terms of use and disclaimers on the BioMedWire website applicable to all content provided by BMW, wherever published or re-published: http://BMW.fm/Disclaimer
