ML model can classify sex from retinal photograph, clinicians can't

Re:ML model can classify sex from retinal photograph, clinic... by acituan on Saturday May 15, 2021 @04:26PM (#27168670)

Their use of an external validation dataset eliminates many, if not all, of those concerns.

Regarding external validation set:

> This dataset differed from the UK Biobank development set with respect to both fundus camera used, and in sourcing from a pathology-rich population at a tertiary ophthalmic referral center.

Regarding UK Biobank set (training set)

> UK Biobank dataset, which is an observational study in the United Kingdom that began in 2006 and has recruited over 500,000 participants—85,262 of which received eye imaging38. Eye imaging was obtained at 6 centers in the UK and comprises over 10 terabytes of data39. Participants volunteered to provide data including other medical imaging, laboratory results, and detailed subjective questionnaires.

[ Parent | on HN ]

Re:ML model can classify sex from retinal photograph, clinic... by nerdponx on Saturday May 15, 2021 @06:48PM (#27169682)

Still, until we have a better sense of what features the model is extracting, it's a surprising result and ought to be treated with caution.

[ Parent | on HN ]

Re:ML model can classify sex from retinal photograph, clinic... by slver on Sunday May 16, 2021 @02:27AM (#27171651)

I wish we treated our own conclusions this way.

[ Parent | on HN ]

Re:ML model can classify sex from retinal photograph, clinic... by ithkuil on Sunday May 16, 2021 @03:05AM (#27171818)

Science's secret weapon is the innate pleasure that many academics have in proving their colleagues wrong.

[ Parent | on HN ]

Re:ML model can classify sex from retinal photograph, clinic... by slver on Sunday May 16, 2021 @08:16AM (#27173150)

Unfortunately that also results in them trying to prove their colleagues are wrong about other of their colleagues being wrong.

So the bottom line is that ironically due to this science gets stuck on false beliefs and they can be a bit hard to dislodge.

We have even this principle: https://en.wikipedia.org/wiki/Planck%27s_principle

[ Parent | on HN ]

Re:ML model can classify sex from retinal photograph, clinic... by raihansaputra on Sunday May 16, 2021 @03:21AM (#27171878)

That same spirit is also what's behind shorts in the market to keep overly optimistic players in check. As with crypto-ransom operations against corporations to keep their system security on alert. Adversarial actions (within reasonable boundaries) is the most important balance mechanism.

[ Parent | on HN ]

Re:ML model can classify sex from retinal photograph, clinic... by slver on Monday May 17, 2021 @02:32AM (#27180592)

The problem is keeping the adversaries balanced. Otherwise you have one side overpowering everything (and typically not through the "correctness" of their argument, but scale and resources) and the adversarial system becomes detrimental to progress.

[ Parent | on HN ]