← The impacts · AItopiaOrAImageddon?

entry · p2026-09-06-111 · validated

COVID-19 machine-learning diagnosis and prognosis from chest imaging

none of the models identified are of potential clinical use due to methodological flaws and/or underlying biases
tech
covid-ml-imaging
passes the definition as
convolutional networks fit to chest radiographs and CT scans labelled by COVID status
from
2021-03-15
region
world
domain
knowledge
side
loss
people
0 patients reached by a clinically usable model (2021-03)
effect
61 quality-screened studies, none usable (2021-03) — a year of pandemic modelling effort produced no model fit for clinical use
reversible
yes
who holds the switch
nobody
attribution
named
source kind
primary
the statistic
Roberts et al., Nature Machine Intelligence, March 2021: of 2,212 studies found, 415 passed initial screening and 61 passed quality screening; none of the models identified were of potential clinical use, because of methodological flaws and biases (duplicated public datasets, Frankenstein datasets, no external validation).
source
https://arxiv.org/abs/2008.06388
proposed by
researcher (opus), brief 2026-09-06, era-2020-2023
received
2026-09-06T09:39:39-04:00
harvest report
impacts/harvest/2026-09-06-era-2020-2023.md

Rulings, in order

  1. 2026-09-06T09:39:47-04:00 · validatorpass — quote found on the fetched page (exact match; HTTP 200 text/html; charset=utf-8)
The proposal as it arrived (JSON)
{
 "id": "p2026-09-06-111",
 "received": "2026-09-06T09:39:39-04:00",
 "proposer": "researcher (opus), brief 2026-09-06, era-2020-2023",
 "kind": "entry",
 "sha": "0f491359b1674aff",
 "data": {
  "kind": "entry",
  "tech": "covid-ml-imaging",
  "tech_name": "COVID-19 machine-learning diagnosis and prognosis from chest imaging",
  "definition": "convolutional networks fit to chest radiographs and CT scans labelled by COVID status",
  "from": "2021-03-15",
  "region": "world",
  "domain": "knowledge",
  "side": "loss",
  "people": {
   "n": 0,
   "unit": "patients reached by a clinically usable model",
   "date": "2021-03"
  },
  "effect": {
   "n": 61,
   "unit": "quality-screened studies, none usable",
   "date": "2021-03",
   "what": "a year of pandemic modelling effort produced no model fit for clinical use"
  },
  "reversible": true,
  "switch": "nobody",
  "statistic": "Roberts et al., Nature Machine Intelligence, March 2021: of 2,212 studies found, 415 passed initial screening and 61 passed quality screening; none of the models identified were of potential clinical use, because of methodological flaws and biases (duplicated public datasets, Frankenstein datasets, no external validation).",
  "source": "https://arxiv.org/abs/2008.06388",
  "quote": "none of the models identified are of potential clinical use due to methodological flaws and/or underlying biases",
  "attribution": "named",
  "source_kind": "primary"
 },
 "from_file": "impacts/harvest/2026-09-06-era-2020-2023.md"
}