Opinion for the court filed by Circuit Judge DYK.
Dissenting opinion filed by Circuit Judge NEWMAN.
DYK, Circuit Judge.General Protecht Group, Inc. (“GPG”), Wenzhou Trimone Science and Technology Electric Co., Ltd. (“Trimone”), and Shanghai ELE Manufacturing Corp. (“ELE”) appeal from a final determination of the International Trade Commission (“Commission”) that the importation into the United States, sale for importation, or sale within the United States of certain ground fault circuit interrupters (“GFCIs”) violated section 337 of the Tariff Act of 1930, as amended, 19 U.S.C. § 1337. The Commission issued limited exclusion orders against the importation of GFCI products from each of the three appellants. See In re Certain Ground Fault Circuit Interrupters & Prods. Containing Same, Inv. No. 337-TA-615, 2009 WL 962585 (Int’l Trade Comm’n Mar. 9, 2009) (“Final Determination”). The Commission found that devices manufactured by appellants infringe U.S. Patent Nos. 7,283,340 (“the '340 patent”), 5,594,398 (“the '398 patent”), and 7,164,564 (“the '564 patent”), and that none of these patents is invalid or unenforceable.
We hold that the Commission erred in three respects: (1) GPG’s 2003 and 2006 GFCIs and ELE’s 2006 GFCIs do not infringe the '340 patent, because they do not have a “detection circuit” as claimed in the patent; (2) Trimone’s 2006 GFCIs and ELE’s 2006 GFCIs do not infringe the '340 patent, because the “load terminals” of the patent do not include receptacle outlets; and (3) GPG’s 2006 GFCIs do not infringe the '398 patent, because GPG performs the function of the “latching means” in a substantially different way than the structure disclosed in the patent. We remand for further proceedings in these respects. We affirm the Commission’s determination in all other respects.
BACKGROUND
GFCI receptacles are the electrical outlets found commonly in bathrooms and kitchens. Typically, they can be identified by the “test” and “reset” buttons positioned between the two electrical sockets. GFCIs are designed to protect people from potentially fatal electrical shocks by cutting off the flow of electricity — or “tripping” — when the device detects a “ground fault.” A GFCI detects a ground fault when the electrical current flowing from the GFCI to a connected device on the “hot” prong of the socket does not match the current flowing from the connected device back to the GFCI on the “neutral” prong of the socket. This indicates that *1306electrical current is leaking out along an unintended path, possibly through a person. This may be due to an exposed wire or the connected device’s being dropped in water, for example.
Pass & Seymour, Inc. (“Pass & Seymour”) is the assignee of various GFCI patents. In September 2007, on a complaint filed by Pass & Seymour, the Commission initiated an investigation to determine whether violations of section 337 of the Tariff Act of 1930 (19 U.S.C. § 1337) had occurred by the importation into the United States, the sale for importation, or the sale within the United States after importation of certain GFCIs that allegedly infringe some of Pass & Seymour’s patents. GPG, Trimone, ELE, and others were named as respondents.
On September 24, 2008, the Administrative Law Judge (“ALJ”) issued an initial determination finding violations of section 337 by each of the appellants. See In re Certain Ground Fault Circuit Interrupters & Prods. Containing Same, Inv. No. 337-TA-615 (Int’l Trade Comm’n Sept. 24, 2008) (“Initial Determination”). Appellants petitioned the Commission for review of the ALJ’s decision, and the Commission determined that it would review certain of the ALJ’s findings.
On March 9, 2009, the Commission issued its final opinion. With respect to the devices and claims involved in this appeal, the Commission, while modifying the ALJ’s claim constructions in a few respects, affirmed the findings of infringement. GPG, Trimone, and ELE appealed. This opinion addresses that appeal. In certain other respects, the Commission reversed the ALJ’s findings of infringement. Pass & Seymour appealed. In a separate opinion released today we address that appeal.1
Insofar as is pertinent here, the Commission issued a limited exclusion order prohibiting entry into the United States of GPG GFCIs found to infringe one or more of claims 1 and 7 of the '398 patent and claims 14 and 18 of the '340 patent; Trimone GFCIs infringing one or more of claims 14 and 18 of the '340 patent; and ELE GFCIs infringing one or more of claims 1, 7, and 8 of the '398 patent, claims 14, 18, and 30 of the '340 patent, and claims 1 and 15 of the '564 patent. The Commission’s determination became final on May 8, 2009, at the conclusion of the sixty-day presidential review period. See 19 U.S.C. § 1337(j)(4). As noted, GPG, Trimone, and ELE timely appealed, and we have jurisdiction under 28 U.S.C. § 1295(a)(6).
Discussion
GPG, Trimone, and ELE raise numerous issues challenging the Commission’s final determination, arguing that their accused devices do not infringe or that the asserted patents are invalid. We have considered appellants’ arguments, and find most of them unpersuasive. Therefore, we affirm the Commission’s determination in most respects, and we think an extended discussion of those points is unnecessary. We focus our discussion on only those issues as to which we conclude that the Commission was in error.
We review the Commission’s final determination of a violation of section 337 under the standards of the Administrative Procedure Act (“APA”). See 19 U.S.C. § 1337(c). Under the APA, this court reviews the Commission’s legal determinations de novo and its factual findings for substantial evidence. See 5 U.S.C. § 706(2)(A), (E); Honeywell Int’l, Inc. v. Int’l Trade Comm’n, 341 F.3d 1332, 1338 *1307(Fed.Cir.2003). Claim construction is an issue of law and is subject to de novo review. See Cybor Corp. v. FAS Techs., Inc., 138 F.3d 1448, 1451 (Fed.Cir.1998) (en banc).
I The '340 Patent
The Commission found that GFCI devices from each of the three appellants infringe the '340 patent. GPG’s 2003 and 2006 GFCIs were found to infringe claims 14 and 18; Trimone’s 2006 GFCIs were found to infringe found to infringe claims 14 and 18; and ELE’s 2006 GFCIs were found to infringe claims 14, 18, and 30.
The '340 patent is directed to a GFCI receptacle that “detects the wiring state of the device and inhibits operation if the device is miswired.” '340 patent col.2 11.29-31. When properly wired, the electrical source is connected to the GFCI’s “line terminals,” from which power flows into the rest of the device. However, there is a chance that an installer may accidentally miswire the electrical source to the device’s “load terminals,” which are normally intended for connection to downstream outlets that receive ground fault protection through the GFCI. If miswired, the devices do not protect against a ground fault. To effect the miswiring protection, the '340 patent has a “detection circuit” to detect whether the GFCI device is properly wired to an electrical circuit, and “four sets of interrupting contacts” configured to make or break an electrical circuit between the line terminals and the load terminals, depending on the signal from the detection circuit. Claim 14 is representative:
14. An electrical wiring device comprising: line terminals and load terminals; at least one detection circuit including a circuit segment coupled between the line terminals and configured to generate a predetermined signal in response to detecting a proper wiring condition, the predetermined signal not simulating a fault condition, a proper wiring condition being effected when the line terminals are connected to a source of AC power; and
an interrupting contact assembly coupled to the at least one detection circuit, the interrupting contact assembly including four sets of interrupting contacts that are configured to provide electrical continuity between the line terminals and the load terminals in a reset state and configured to interrupt the electrical continuity in tripped state, the interrupting contact assembly being substantially prevented from effecting the reset state absent the predetermined signal being generated by the at least one detection circuit.
'340 patent col.10 11.7-25 (emphases added).
A “detection circuit”
ELE and GPG argue that their GFCIs do not infringe the '340 patent because their devices do not have a “detection circuit.” The ALJ construed “detection circuit” to mean “at least one detection circuit having a circuit segment connected between the line terminals and configured to generate a predetermined signal in response to detecting a proper wiring condition, which occurs when the line terminals are connected to a source of AC power.” Initial Determination, slip op. at 85.2 Ap*1308pellants did not petition the full Commission for review of the ALJ’s construction. Thus, the Commission did not review the ALJ’s construction, and therefore adopted it. Neither ELE nor GPG challenges on appeal the appropriateness of the ALJ’s original claim construction. GPG, however, argues that the ALJ effectively modified the construction or misapplied it with respect to the GPG GFCIs.
ELE argues that the Commission erred in adopting the ALJ’s finding that its GFCIs infringe the '340 patent, because its devices do not have a circuit that generates a “predetermined signal” when the circuit detects that the GFCI device is properly wired.3 Rather, ELE claims that its GFCIs are designed to electrically isolate the load terminals, so that the devices simply only have power when an AC power source is properly wired to the GFCIs’ line terminals, and lack power when the power source is miswired to the load terminals.
The ALJ based his finding that the ELE 2006 GFCIs meet the “detection circuit” limitation on the testimony of Pass & Seymour’s expert, Dr. Tom Harman (“Dr. Harman”). At the hearing before the ALJ, Dr. Harman presented the circuit diagram for ELE’s device and identified a portion of the device as the “detection circuit.” See J.A. 40,396. He then identified the “predetermined signal” as electrical current flowing in from the hot line terminal, through the identified circuit, and out through the neutral line terminal. Id.
This testimony, however, is plainly inconsistent with the asserted claims of the '340 patent and the ALJ’s construction. The construed claims require the “detection circuit” to “generate a predetermined signal in response to detecting a proper wiring condition.” Initial Determination, slip op. at 85 (emphasis added). But as Dr. Harman’s own testimony makes clear, his “predetermined signal” is merely the “current flow” originating from the hot line terminal. J.A. 40,396. The identified circuit does not generate this current; it is the current that comes from the AC power connection. So instead of detecting a proper wiring condition and generating a signal in response, as the properly construed claim requires, the accused GFCIs simply have power to operate, or not. Dr. Harman’s testimony is not substantial evidence to support a finding that ELE’s 2006 GFCIs infringe claims 14, 18, and 30 of the '340 patent.
While the focus of GPG’s “detection circuit” argument is somewhat different than ELE’s, GPG’s 2003 and 2006 GFCIs also do not satisfy the “detection circuit” limitation for the same reason as argued by ELE. The circuits identified by Dr. Har-man in the accused GPG devices do not generate any signal in response to detecting a proper wiring condition. Rather, the GFCIs either have power if properly wired, or they do not if miswired. AC current from the hot line terminal flows into each identified circuit when properly wired, but that AC current cannot be the “predetermined signal” because it is not generated by the circuit. The Commission acknowledged that both ELE’s and GPG’s GFCIs work in the same way, by isolating the load terminals so that power only flows when the devices are wired properly to the line terminals. Oral Arg. at 21:31 — :46. Therefore, GPG’s 2003 and 2006 GFCIs also do not infringe claims 14 and 18 of the '340 patent.
*1309B “load terminals”
Trimone argues that its 2006 GFCIs do not satisfy the “four sets of interrupting contacts that are configured to provide electrical continuity between the line terminals and the load terminals in a reset state and configured to interrupt the electrical continuity in tripped state” limitation of the '340 patent. The ALJ construed “four sets of interrupting contacts” to mean “four pairs of electrical contacts that can separate from each other to interrupt the flow of electricity.” Initial Determination, slip op. at 89. The partial figure below, taken from Figure 1 of the '340 patent, shows a circuit interrupter, indicated by label number 120, having four pairs of electrical contacts — two on the line terminal side of the circuit (line terminals to the left side of the figure not shown) and two on the load terminal side of the circuit.
[[Image here]]
'SI0 Patent, Figure 1 (partial)
The other disclosed GFCI embodiments in the '340 patent all have the same basic circuit interrupter.
Trimone argues that its 2006 GFCI devices do not satisfy the “four sets of interrupting contacts that are configured to provide electrical continuity between the line terminals and the load terminals in a reset state and configured to interrupt the electrical continuity in tripped state” limitation of the '340 patent because its GFCIs have only two sets of contacts configured to make or break the circuit between the line terminals and the load terminals. The figure below shows Trimone’s circuit interrupter (the wires in the lower left of the figure lead to the line terminals).
*1310[[Image here]]
Trimone 2006 GFCI Circuit Interrupter
Trimone asserts that it was error for the ALJ to count the two sets of contacts in Trimone’s GFCIs between the line terminals and the GFCI receptacle outlets — the electrical sockets on the face of the GFCI device — in finding the limitation satisfied, because receptacle outlets are not load terminals. The Commission and Pass & Seymour respond that receptacle outlets are also known as “user load terminals”; that someone of ordinary skill in the art would understand receptacle outlets to be a type of load terminal; and that any devices connected to a GFCI, including those connected to receptacle outlets, are considered “loads.” The issue is thus whether the term “load terminals” in the '340 patent includes receptacle outlets.4
We start with the '340 patent itself. A patent’s specification “is the single best guide to the meaning of a disputed term.” Phillips v. AWH Corp., 415 F.3d 1303, 1315 (Fed.Cir.2005) (en banc) (quoting Vitronics Corp. v. Conceptronic, Inc., 90 F.3d 1576, 1582 (Fed.Cir.1996)) (quotation mark omitted). The '340 patent never describes receptacle outlets as load terminals. In fact, the circuit diagram drawings in the '340 patent all clearly label the load terminals separately from the receptacle outlets, as seen in the portion of Figure 1 above.
We have held that expert testimony can be useful in assisting a court “to establish that a particular term in the patent ... has a particular meaning in the pertinent field.” Phillips, 415 F.3d at 1318. However, we have cautioned that “conclusory, unsupported assertions by experts as to the definition of a claim term are not useful to a court.” Id. The testimony relied on by the ALJ does not reveal that the term “load terminals” had a particular meaning in the art that included receptacle outlets. None of the experts identified a particular meaning in the art, and an expert’s subjective understanding of a patent term is irrelevant. See Howmedica Osteonics v. Wright Med. Tech., 540 F.3d 1337, 1347 & n. 5 (Fed.Cir. 2008) (holding that an inventor’s subjective intent in using a term is irrelevant, but that the inventor may testify as an expert on the established meaning of particular terms in the relevant art); Symantec Corp. v. Computer Assocs. Int’l, 522 F.3d 1279, 1291 (Fed.Cir.2008) (disregarding expert testimony that “simply recite[d] how *1311each expert would construe [a] term ... based on his own reading of the specification” because it did “not identify the ‘accepted meaning in the field’ to one skilled in the art”); Sinorgchem Co., Shandong v. Int’l Trade Comm’n, 511 F.3d 1132, 1137 n. 3 (Fed.Cir.2007) (“We attribute no weight to that testimony because the experts did not identify any evidence that those skilled in the art would recognize ‘controlled amount,’ or any term used in the specification, has an accepted meaning in the field of chemistry. Under such circumstances, testimony as to how one skilled in the art would interpret the language in the specification is entitled to little or no weight.”).
Moreover, contrary to the ALJ’s unsupported conclusion, the experts here never suggested that receptacle outlets, which they also called “user load terminals” or “user accessible load terminals,” could be referred to as simply load terminals. Rather, they were fastidious in referring to them as separate elements. See J.A. 40,377 (testimony of Pass & Seymour witness Jim Osterbrock) (“The G4 design separates the line terminals from the load terminals and user accessible load terminals.”); id. at 40,378 (“[W]e differentiate by calling them load terminals and user accessible load terminals.”); id. at 40,938 (testimony of Shanghai Meihao Electric Co. expert Dr. Mark Horenstein) (agreeing that “receptacles are also referred to as user accessible load terminals”). The testimony by Dr. Harman relied on by the ALJ does not mention receptacle outlets or user load terminals at all. See id. at 40,408. Furthermore, usage of the term “user load terminals” to refer to receptacle outlets in other patents does not help the patentee here. There is no evidence that a person of ordinary skill in the art would read the term “load terminals” to also include receptacle outlets, or user load terminals, and there is nothing in the extrinsic evidence that in any way contradicts the meaning of the term “load terminals” apparent from the face of the patent.
Therefore, the ALJ erred in construing the term “load terminals” to also include receptacle outlets. As a result, under the proper construction of “load terminals,” Trimone’s 2006 GFCIs and ELE’s 2006 GFCIs have only two “sets of interrupting contacts that are configured to provide electrical continuity between the line terminals and the load terminals in a reset state and configured to interrupt the electrical continuity in tripped state,” and do not infringe the asserted claims of the '340 patent.
II The '398 Patent
The Commission found that GPG’s 2006 GFCIs infringe claims 1 and 7 of the '398 patent. GPG argues that the structure in its 2006 GFCIs for performing the function of the “latching means” in claims 1 and 7 is substantially different from that disclosed in the '398 patent.
The '398 patent discloses a new mechanical architecture for a GFCI receptacle, with detailed descriptions of an improved contact system that can move between a circuit-making position and a circuit-breaking position. The invention involves fixed contacts on the line terminal side of the device and fixed contacts on the load terminal side of the device, separated by a break in the device circuit that can be completed by a moveable conducting member when the device is in the reset, or circuit-making, state. The claims use means-plus-function language to define several elements, pursuant to 35 U.S.C. § 112 ¶ 6. Claim 1 is representative:
1. A ground fault interrupter (gfi) wiring device for connection in an electrical circuit, said device comprising:
a) housing means defining an enclosed space;
*1312b) at least one pair of electrical terminals fixedly supported in spaced relation within said enclosed space;
c) a unitary, electrically conducting member carrying a pair of spaced electrical contacts;
d) mounting means for said conducting member to permit movement thereof between a first position, wherein said pair of contacts are in respective, circuit-making engagement with said pair of terminals, and a second position, wherein both of said pair of contacts are in spaced, circuit-breaking relation to said pair of terminals;
e) biasing means urging said conducting member toward movement to said second position;
f) latching means releasably retaining said conducting member in said first position; and
g) actuating means for releasing said latching means to permit said biasing means to move said conducting member to said second position in response to a predetermined fault condition in said electrical circuit.
'398 patent col.13 11.24-46 (emphasis added). The element at issue here is the “latching means” element. Claim 7 is a dependent claim that also requires the same “latching means” element.
“Literal infringement of a means-plus-funetion limitation requires that the relevant structure in the accused device perform the identical function recited in the claim and be identical or equivalent to the corresponding structure in the specification.” Applied Med. Res. v. U.S. Surgical Corp., 448 F.3d 1324, 1333 (Fed. Cir.2006). A structure in the accused device constitutes an equivalent to the corresponding structure in the patent only if the accused structure performs the identical function “in substantially the same way, with substantially the same result.” Id.-, see Kemco Sales, Inc. v. Control Papers Co., 208 F.3d 1352, 1364 (Fed.Cir.2000) (stating that in order to literally infringe, “the accused structure must either be the same as the disclosed structure or be a section 112, paragraph 6 ‘equivalent,’ i.e., (1) perform the identical function and (2) be otherwise insubstantially different with respect to structure”).
The ALJ concluded that the function of the latching means is “releasably retaining the conducting member in the first position,” and that the structure described in the '398 patent corresponding to that function is “a pin passing through a hole in the block having a shoulder that cooperates with a hole in the latch member and a spring biasing the pin to retain the conducting member in the first position, and equivalents thereof.” Final Determination, slip op. at 10. Under this construction, the ALJ found that “because [GPG’s 2006 GFCIs] contain a magnet capable of retaining an armature that is attached to the GFCI’s ‘mounting means,’ ” and because “[w]hen the armature is retained by the magnet, the conducting member is in a first (circuit-making) position with respect to the pair of terminals,” the GPG GFCIs satisfy the function of the latching means and contain a structural equivalent to the latching means structure disclosed in the '398 patent. See Initial Determination, slip op. at 66.
The Commission did not modify the ALJ’s definition of the function, but it did modify the identified structure “to include the entire ‘latch member,’ and not only the hole therein.” Final Determination, slip op. at 11. Despite modifying the structure corresponding to the latching means, the Commission upheld the ALJ’s finding that GPG’s 2006 GFCIs contain structural equivalents to the structure identified in the '398 patent. See id. at 12.
*1313GPG argues that the structure in its 2006 GFCIs performs the function of the latching means in a substantially different way than the structure disclosed in the '398 patent. GPG’s structure, as the ALJ noted, uses only a magnet in order to retain the contacts of the conducting member in the first, circuit-making position. In contrast, the disclosed structure in the '398 patent employs a mechanical solution requiring the interaction of a number of separate parts, including a spring latching member, a pin with a shoulder extending through a hole in the spring latching member, and a spring biasing the pin to pull against the force of the spring latching member.
As we held in Toro Co. v. Deere & Co., 355 F.3d 1313 (Fed.Cir.2004), one system that accomplishes a function mechanically and another system that accomplishes the same function using magnetic force “function in fundamentally different ways.” Id. at 1324. The Commission and Pass & Seymour argue that the ALJ’s decision was supported by substantial evidence because of the testimony of expert witnesses that one of ordinary skill in the art would have considered the magnetic matching structure to be interchangeable with the mechanical structure of the '398 patent. See Initial Determination, slip op. at 67. However, the experts merely testified that magnets were well known as latches, not that they performed the latching means function in substantially the same way as the mechanical latch disclosed in the patent. See J.A. 40,524 (testimony of Pass & Seymour expert Dr. Fred Stolfi) (“The use of a magnet as a latch is fairly common in other devices. You know, for example like a door latch often has a magnet catch. And there [are] patents that I have looked at that speak about a magnetic latch in a GFCI ....”); id. at 40,750 (testimony of ELE expert Kenneth Eugene Haynes) (agreeing that “permanent magnets were known in the art of latching structures and GFCIs in the late 1980s and 1990s”); id. at 40,979 (testimony of GPG expert Dr. James Roberge) (agreeing that permanent magnets had been used in “latching relays” in the electrical industry to perform latching functions). This testimony “goes to the function or result of these systems, and begs the issue of the way in which [the mechanical] systems and [magnetic] systems actually work.” Toro, 355 F.3d at 1324. We therefore hold that substantial evidence does not support the Commission’s finding that the magnetic latching structure of GPG’s 2006 GFCI is equivalent to the mechanical structure disclosed in the '398 patent corresponding to the latching means.
Conclusion
For the aforementioned reasons, we reverse the Commissions findings that GPG’s 2003 GFCIs infringe claims 14 and 18 of the '340 patent; that GPG’s 2006 GFCIs infringe claims 1 and 7 of the '398 patent and claims 14 and 18 of the '340 patent; that Trimone’s 2006 GFCIs infringe claims 14 and 18 of the '340 patent; and that ELE’s 2006 GFCIs infringe claims 14, 18, and 30 of the '340 patent. We remand to the Commission to modify its limited exclusion order in accordance with this opinion.
AFFIRMED IN PART, REVERSED IN PART, and REMANDED
Costs
No costs.
. The Commission’s determination and limited exclusion order also involved other respondents and products, and additional patents not included in either appeal.
. Claim 18 is dependent from claim 14, and thus has the same limitation. The language of independent claim 30 is slightly different from that of claim 14 with respect to this element. However, none of the parties has suggested that the difference in language represents any difference in scope. The ALJ treated the elements as identical. See Initial Determination, slip op. at 92.
. The dissent suggests that the parties did not raise this argument, but ELE clearly argued that "[b]y treating the 'predetermined signal’ as including whatever AC power happens to arrive at a building from the local power plant, the Commission has broadened this limitation beyond all recognition and effectively reads it out of the claim.” Br. of Appellant Shanghai ELE Mfg. Corp. 42.
. This issue also applies to ELE’s 2006 GFCIs, which we have already concluded do not infringe the '340 patent. Contrary to the dissent, ELE raises the load terminal issue. See Br. of Appellant Shanghai ELE Mfg. Corp. 44-45 (“Notably, the '340 patent does not contemplate separating the downstream load terminals from the surface ('face') load or receptacle terminals as is done in ELE's 2006 GFCIs.”).