The limitation most likely to be misread in US20260221372A1 is the one at the end. The application, published 30 July 2026, assigned to Canon Kabushiki Kaisha and naming four inventors, carries seventeen claims across three independents: claim 1 a data processing system, claim 16 a method, claim 17 a non-transitory computer readable medium. All three recite the same operative sequence.

That sequence has two halves, and both are mandatory. First, obtain a series of captured images, provide them to at least one image recognition model configured to determine whether a subject of a user is present in the path of the X-ray beam, and on detection reduce the radiation provided to that subject by adjusting imaging conditions. Second — and this is the part that bounds the claim — subsequently, in response to a predefined condition being met, control the apparatus to further adjust the imaging conditions so as to increase the radiation supplied by the source to the detector.

A system that detects an operator and reduces dose, and stops there, does not practise these claims. The restoration step is not an optional refinement in a dependent claim; it is recited in every independent. Scope here is a closed loop, not a safety cut-off.

What the reduction may consist of

The reduction limitation is drafted in the alternative. Imaging conditions are adjusted "including one or more" of the collimation of the beam, the presence of a filter between the source and the object, and the imaging parameters. Only one is required. Claim 2 defines the parameter branch as reducing tube current or reducing frame rate, and claim 5 develops the collimation branch — the model outputs a location for the detected subject and the collimation is adjusted to prevent at least part of that location from being irradiated.

Claim 6 imposes the counterweight: adjusting the collimation is performed on a condition that the determined region of the object continues to be irradiated. Claims 7 and 8 supply the machinery for finding that region, using a further recognition model that identifies it by identifying an instrument in the image. Note the dependency shape — claim 6 hangs from claim 1, where collimation is only one of three alternatives, so "the collimation" arrives without having been positively required upstream.

A data processing system as claimed in claim 1, wherein the at least one image recognition model is configured to output an indication of a location of the detected subject within one of the at least some of the captured images, wherein the imaging conditions include the collimation of the X-ray beam, wherein the controlling the imaging apparatus to adjust imaging conditions comprises, based on the indication of the location of the detected subject, adjusting the collimation to prevent at least part of the location of the detected subject from being irradiated by the X-ray beam.— System and Method for Adjusting Fluoroscopy Imaging Conditions, US20260221372A1

Where the distinctive scope sits

The claim that carries the most technically specific position is claim 12. It ties the subject of the user to a hand, specifies that hand as ungloved, and requires the model additionally to determine whether a gloved hand is present — with a different action performed in response. Claim 13 identifies that action as preventing an automatic exposure control system from increasing X-ray intensity, and claim 14 splits the detection across two separate recognition models, one per case.

That is a narrow and concrete fallback. It is also the only place in the set where the recited subject is a hand at all; claim 1 says only "a subject of a user". Anyone characterising this application as hand-detection is describing claim 12, not the independent claims. The abstract, meanwhile, uses a third vocabulary — "an unintended part of an operator (e.g. a hand)" — which appears nowhere in the claims. Three descriptions, and only the claim language sets scope.

Claim 15 defines the predefined condition that triggers restoration, listing a failure to detect the subject in subsequent images, a change in capture angle, reception of user input, and expiry of a predefined time limit. Because the independents require restoration but do not specify its trigger, claim 15 is where that open term acquires content — and its breadth, particularly the time-limit option, means the restoration step can be satisfied without the operator having moved at all. For anyone mapping around this set, that asymmetry is the opening: the reduction half is tightly specified across three alternatives with concrete dependent support, while the restoration half is defined only by whichever condition claim 15 supplies.

Five defects appear as published and are reproduced rather than repaired. All three independent claims read "adjust imaging conditions including one or more:" with no "of". Claim 1 refers to "the series of fluoroscopy frames" having introduced "a series of fluoroscopy images" — a term mismatch with no antecedent for "frames". Claim 7 uses the permissive "may be" inside a claim, which sits awkwardly with definiteness, and also refers to "the series of X-ray images" where claim 1 introduced fluoroscopy images and optical images. Claim 14 reads "the least one image recognition model", missing "at". Claim 16 opens with "a series of captured images of a object".

None alters scope in a way a reader would act on, and all are of a kind routinely corrected during prosecution. That is the standing caveat: this is a published application, not an issued patent. It has no enforceable effect in this form, and the two-step structure of the independent claims — reduce, then restore — is broad enough in its second half that narrowing during examination should be expected rather than treated as a surprise.