https://datastudio.google.com/s/utRnltIRh3E
def capturing_lung_function_using_computer_vision():
description = [
**Biological Premise**: The process of respiration is made possible by creating a pressure gradient between the atmosphere and
the lungs that air will follow. When breathing in, the lungs and ribcage expand to allow air to come in. When breathing out,
they contract, pushing the air out. By taking advantage of this structural property of expansion and compression of the thorax,
we are able to track the movement of air during key breathing exercises and use them to assess lung function.
**The Problem It Solves:** Using a spirometer is inconvenient, prone to human error, and unable to capture significant
visual information, therefore making it a perfect candidate for computer-vision based methods.
**The Product:** For our prototype, we used the [Intel RealSense Depth Camera D415](<https://www.intelrealsense.com/depth-camera-d415/>). As shown in the video above, we place the
camera in front of the subject (in this case my co-founder Xavier) and record the depth information while he performs
traditional spirometry breathing maneuvers using a spirometer (a device which directly captures the flow of air through the mouth).
The algorithm isolates the adequate region of the upper body and records the changes in chest volume throughout the effort.
The algorithm then translates these fluctuations into transferred volume of air over time, generating a flow-volume loop from
which we extract the conventional measures of lung function (FEV1, FVC etc.) ]
return description
def automate_reading_and_interpretation_of_lung_function_results():
description = [
**Context:** After a patient performs their pulmonary function test, the results are handed over to a reading pulmonologist
who will examine the lung function values and the shape of the flow-volume curve. After comparing them to "normal" values
expected from someone similar with healthy lungs, the pulmonologist will interpret the test and write a diagnosis which will
determine the course of the patient's care.
**Problem It Solves:** Reading and interpretation of pulmonary function test results is repetitive, insufficiently standardized
and time-consuming, making it a perfect candidate for AI-driven automation.
**The Product:** Our algorithm extracts key visual features of the flow-volume curve and categorizes the abnormal components
into central categories such as obstructive or restrictive lung disease. The lung function metrics (FEV1, FVC...) are then
compared to established databases to identify patterns consistent with specific conditions. Finally, reports are generated
containing clearly annotated graphs and tables as well as pre-drafted reading and interpretation sections ready for physician
sign-off. ]
return description