Page 184 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

has shown its potential in detecting patients at the early stages of the pancreatic cancer with
high precision, enabling timely therapeutic interventions and significantly improving clinical
outcomes.

Use Case Status: Pilot

Category: Artificial Intelligence, Healthcare

Partners: Guangdong Provincial People’s Hospital; The First Affiliated Hospital of Zhejiang
University School of Medicine; Department of Radiology, Shanghai Institution of Pancreatic
Disease; The Affiliated People's Hospital of Ningbo University; Alibaba Foundation

2�2 Benefits of use case

1. safe, low-cost, high accuracy, early cancer detection using non-contrast CT scans.
2. detecting cancer before symptoms appear and thereby saving lives.
3. validated using pilot studies and expert feedback, the system enables timely intervention,
and significant benefits for prognosis.

2�3 Future Work

To achieve efficient and accurate early multi-cancer screening with a single non-contrast CT
imaging examination, we propose a deep learning-based early cancer screening solution that
is capable of 1) detection of eight major cancers (the top eight with the highest mortality in the
world); 2) differential diagnosis of cancers; and 3) measurement and station labelling of eight
major organs, in the early-stage with non-contrast CT scans. The proposed deep learning-based
early cancer screening solution maximizes the clinical value and patient benefits from a single
CT imaging examination, achieving accurate and efficient detection of multi-cancer at both
high sensitivity and high specificity levels, which has long been considered impossible due to
the low contrast of the tumor presented in CT imaging.

It is possible to build a more accurate AI-based cancer screening model by the integration of
routine non-contrast CT imaging with clinical history data in the future. The FDA’s recognition
of DAMO PANDA signifies a breakthrough in regulatory acceptance of AI technologies, paving
the way for their widespread implementation worldwide. We aim to establish collaborations
with new partners to validate our model across different regions. More importantly, AI-driven
screening should undergo rigorous evaluation equivalent to conventional methods—through
randomized controlled trials (RCTs) comparing its efficacy to validated comparators in terms
of all-cause mortality reduction. With global partnerships, we wish to achieve the real clinical
value of AI-based screening method for cancer in reducing all-cause mortality to address the
global cancer burden.

3 Use Case Requirements

REQ-01: It is required to have routine/non-contrast CT scan (chest or abdomen), DICOM format
for radiological examination. To use CT Scanner Manufacturers which comply with DICOM 3.0
protocol.

REQ-02: It is required to have 16 slices and beyond CT Scanners. CT Scan Parameters: tube
voltage 70-140kv, tube current 10-400 mA, slice thickness 1-5mm (> 2mm recommended),

148