Singapore-based medtech startup FathomX has built an Artificial intelligence (AI) powered breast cancer screening solution to improve diagnoses and deliver results quickly.
The mammography support tool FxMammo is said to improve the screening procedure for mammograms and reduce false positives for interval cancers, or cancers that are diagnosed in between routine screening episodes.
FathomX CEO Stephen Lim told iTnews Asia that the tool cuts diagnosis times to a few minutes, helping clinicians provide better and faster health care.
Tested in top universities and hospitals across Japan, Taiwan, China, Hong Kong, Thailand, Malaysia, Australia and Singapore, Lim said, the AI assistant leads to fewer false recalls and unnecessary anxiety.
"These studies have shown the AI tool could significantly reduce the false positive rate by over 20 percent and false negative rates by over 38 percent, giving more opportunity for early diagnostics," he added.
Lim said FxMammo can analyse mammograms, generate a heat map of abnormal lesions, and estimate the risk of breast cancer through imaging and analysis of medical history records.
While traditional mammography remains the best tool to detect breast cancer, it is yet prone to human error. The new AI tool is found to be "effective" even in dense breast tissue, which is typically much harder to diagnose, he added.
FathomX is a health AI spin-off company from the National University of Singapore and the National University Health System. It aimed to offer a reliable, easy-to-access solution to support radiologists struggling to keep up with mounting screening needs.
The startup built FxMammo to detect, read and interpret mammograms in real-time through edge computing. It worked with the Diagnostics Development Hub of Singapore's Agency for Science, Technology and Research and Hewlett Packard Enterprise (HPE) to develop the tool into software as a medical device.
The solution is now regulatory approved in Singapore, Malaysia, Thailand, and Indonesia with more countries in the registration process.
Training algorithms
FathomX has built a secure production-level data infrastructure to train its AI model.
It uses HPE computing platforms at the Edge to accelerate data processing and enhance the visualisation experience to operationalise the model.
Mammography requires very high-resolution images to be able to capture microcalcifications, and minuscule deposits of calcium in the breast that might indicate an early stage of cancer.
Using a GPU-accelerated system with a large memory, Lim said the firm can process huge volumes of these medical images.
A HPE Apollo system is said to help the model learn faster and reach an optimal point to identify lesions in a mammogram.
The company is planning to move into HPE GreenLake, an open edge-to-cloud platform in future to enhance market expansions.
