That night, she wrote a new line of code. Not for the calculator. For herself.
It wasn't AI. It wasn't even particularly sophisticated. It was a weighted algorithm that took twenty-three physical markers—from wrist sign (the thumb and pinky overlapping around the wrist) to the ratio of upper to lower body segment, from lens dislocation to a family history of pneumothorax. Each marker had a value. Each value fed into a probability curve. marfan calculator
Dr. Marcus Tse at St. Jude's ran the calculator on a 41-year-old woman with chronic joint pain and a history of miscarriages. Her score was —well below the threshold. He sighed with relief and sent her to rheumatology. That night, she wrote a new line of code
She had written at the very top: "THIS IS A PROBABILISTIC TOOL. IT CANNOT REPLACE CLINICAL JUDGMENT. IT CANNOT SEE THE PATIENT. IT CANNOT HEAR THEIR VOICE." It wasn't AI
She called it the Marfan Calculator.
The autopsy showed cystic medial necrosis. The pathologist noted, almost as an afterthought: "Features suggestive of underlying connective tissue disorder."
The calculator didn't give a binary yes/no. It produced a single number: .
Subscribe now to keep reading and get access to the full archive.