University of Queensland v Siemens Magnet Technology Ltd & Anor

This is an action for infringement of UK patent number GB 2 308 451. The Defendants (“Siemens”) deny infringement and allege that the patent is invalid. The patent is concerned with magnets, and in particular in the design of magnets for use in providing the magnetic field for both nuclear magnetic resonance analysis and for magnetic resonance imaging (“MRI”). Although the claim is not so limited, the patent is largely concerned with MRI and most of the evidence was directed to this field of application. Furthermore, although at the priority date (1995) the magnetic field for these applications might be provided by permanent magnets, ordinary current-excited electromagnets (“resistive magnets”) or by superconducting coils, most of the discussion before me centred on superconducting windings. These have the advantage that the high currents which superconducting windings are capable of carrying can induce intense magnetic fields over substantial volumes with comparatively small apparatus and low power supply.

MRI is an ingenious application of a phenomenon known for many years called nuclear magnetic resonance. In a strong magnetic field the nuclei of many atoms line up like little spinning magnets and absorb and re-radiate radio waves at a well-defined frequency defined by the strength of the applied magnetic field and the nature of the atom in question. In MRI, the nuclei of interest are hydrogen nuclei (protons) which are omnipresent in living tissue.

In MRI the frequencies emitted by re-radiation (during the process called relaxation) are studied. The precise frequency and phase of the radiation is affected by the chemical environment of the protons. The objective is to form an image of a section through the whole or a large part of the patient’s body in which different types of tissue are differentiated visually. The nuclei are lined up in a highly uniform strong magnetic field and subjected to a pulse of radio waves at an appropriate frequency. The frequencies emitted during relaxation are studied, and in order to locate their source in space small magnetic field gradients are superimposed on the principal field. These gradients affect the frequency and phase of the relaxation signals and a picture showing the proton density in space can be computed.

Central to this technique is the accurate measurement of the frequency and phase of the relaxation signal. A non-uniform applied magnetic field would mean that frequency and phase would no longer accurately convey the necessary spatial information. In 1995, the need for high spatial uniformity in the area of interest was well understood. The patent is concerned with another problem, that of size. To obtain a field of high uniformity and of sufficient strength, superconducting coils about 1 metre in diameter need to be used. The “tube” in which the patient is placed is much smaller in diameter than this, because of the space needed for the cryostat in which the main coils, the liquid helium surrounding them, and the liquid nitrogen and vacuum insulation surrounding that, and for other non-superconducting coils, including the gradient coils, the transmission and reception coils and the so-called “shimming” coils (of which more later). Many patients, and particularly those whose upper body is being examined, do not like being confined in a narrow tube which is far from quiet, and those with a touch of claustrophobia may find it impossible. Designs of coils had been directed at this problem, but the evidence was that there are technical advantages to generally tubular configurations, and so a corresponding pressure to make them as short as practicable. By 1995, Philips had released an MRI scanner with a cryostat length of 1.57m (coil length ~1.37m). In any event, a shorter coil represents an advantage in materials and in cooling.

A number of other facts need to be taken into account. In 1995, it was well recognised that to obtain a useful section through a human body, the region of interest in the bore of the magnet in which the field is highly uniform needs to be generally spherical or oblate (Footnote: ¹) and have a diameter ~40 cm or a volume of 35-45 litres. The magnetic “field” B is very large—from 1 Tesla on up. For some sort of comparison the earth’s magnetic field is of the order of 50 microTeslas. The windings of the coils in a superconducting magnet carry currents of the order of 200 Amps on up, and the niobium/tin wire is very long. In such a field, the force on the windings is very substantial—of the order of tonnes. All these facts are relevant to understanding the issues in the action.

In general, the two experts, Professor Müller and Professor Saint-Jalmes, gave helpful evidence. Neither had English as a first language, but both were willing to express complex matters in that language, for which I was very grateful, but this does mean that it is not possible to find meaning in every nuance of their words. Professor Müller indicated that he understood a question, and then answered it. Caution must be used when he merely said yes. Professor Saint-Jalmes tended to indicate any misunderstanding in his answers. The principal criticism which was made of his evidence was that his reading of the document called Ohta was not sensible. He did also tend to repeat his views. My conclusions on this are below. Professor Müller was criticised for the employment of hindsight in his treatment of the same document, but there was contemporary evidence that at least one other person had extracted from the document just the same message as Professor Müller had. I also deal with this below. Both experts were helpful.

Entitled “Magnets for magnetic resonance systems” the specification begins with the statement that the invention relates to magnetic resonance systems and to methods for designing and producing such magnets. As will become clear, the principal concern of the specification is to describe a new method of magnet design. The claims, which are product claims, are in no case limited to this new method of design. An extensive list of the patent literature, none of which I was referred to, is set out, and the “basic components of a magnet system useful for performing magnetic resonance investigations” are said to be shown (page 2 line 15) in Figure 1.

Although very schematic, this figure shows the essentials of the apparatus. The superconducting magnet 13 provides the uniform field in the homogeneous region, and the following is said of figure 1:

In fact, what is said of the prior art magnet here is true on the evidence, so far as it relates to the level of uniformity of field that was thought desirable in 1995. What is difficult to interpret is the phrase “prior to shimming”, which is found in claim 1 and which is better discussed in that context. It is to be noted, however, that the reference to shimming is on the face of it a reference to a procedure standard in 1995 that was undertaken when a new magnet was installed. Although the dimensions of the windings had been carefully computed to achieve the kind of uniformity described as a theoretical design, manufacture would introduce errors in the shape of uneven winding, shrinkage on cooling to the temperature of liquid helium and slight distortion of the windings and former when the magnet is activated by starting current flow. Prof Müller (for Siemens) thought that some attempt was made to compensate for these effects in design, while Prof Saint-Jalmes (for the University of Queensland) was more doubtful, suggesting that some components would be appropriately strengthened. I have little doubt that (for example) the radius of the former would be calculated as the radius after cooling. In any event, the evidence was that, however carefully designed the magnet might be, it would be an order of magnitude out of specification for uniformity once activated, and for this purpose the uniformity of the field in the homogeneous region would be adjusted by the use of iron shims, shimming coils or magnets to achieve the desired degree of uniformity.

Figures 2A and 2B are said to show typical constructions for conventional horizontal and vertical superconducting magnets. The distinction between the two is that the “multiwinding primary coils” of figure 2A are of common inner radius, while the “multiwinding primary solenoids” of Figure 2B do not have a common inner radius. The invention is said to be applicable to either type, providing they have a common inner radius. This is the next problem, which will again be discussed in its context in the claim: what is meant by a “common inner radius”?

After describing the various reasons why shorter magnets are desirable, (including the problem of patient claustrophobia and inaccessibility, cost and weight) the patent sets out the statement of the problem it seeks to solve:

The prior art method of magnet design is then described.

Laukien, referred to in this passage, is a very high level reference to all aspects of magnet and cryostat design. The evidence established that the classic reference was “Thick Cylindrical Coil Systems for Strong Magnetic Fields with Field or Gradient Homogeneities of the 6^th to 20^th Order” Garrett, J Appl Phys 38, 2563 (1967), which I am satisfied was part of the common general knowledge as probably is Garrett’s earlier 1951 paper, at least to judge by citations in the other papers in the case. For example, Professor Saint-Jalmes had written a paper in the field which did not cite Garrett (1967), which was returned to him by the referee for the citation to be added. Professor Crozier and Professor Doddrell, the inventors of the patent in suit, describe Garrett’s 1951 and 1967 papers as “seminal” in their own article “Compact MRI Magnet design by Stochastic Optimization”, J Magnetic Resonance 127, 233-237 (1997). Garrett gives a mathematical treatment of the general problem of how to compute the central field produced by a number of coils symmetrically spaced along the axis, like Figure 2A of the patent. A number of approximations are made, principally that each coil is represented by a single loop carrying a variable current. The analysis is carried out for the central field (in the middle of the coils). The field B in the magnet is given by B=V where V is a solution of Laplace’s equation ²V=0. V is something called the magnetic scalar potential, and in the axially symmetric case these solutions take the form of infinite convergent sums of so-called spherical harmonics. Garrett identifies the desired error-free field as B₀, that is, the zeroth order axial spherical harmonic, and the error which he wishes to eliminate as the higher-order terms in the axial component and all radial components. With a great deal of work, he identifies a number of theoretical designs in which the error term is minimised and states the degree of homogeneity that they possess. He also shows that if the axial position, base radius and ampère-turns (the product of the current flowing in the winding and the number of turns, or magnetomotive force) are allowed to vary independently, in a magnet of n windings he can eliminate all harmonics up to (but not including) 2n and so he calls a magnet with 4 windings an eighth-order magnet, meaning that the first spherical harmonic term which has not been eliminated or minimised (“nulled”) is the eighth. Professor Müller used one of these designs (the 12^th order magnet with uniform inner radius a in Table IV) for his experiments.

To apply the teaching of Garrett to a real life magnet, something like the process undertaken by Professor Müller in Experiment 1 (Fact 1) can be done. What Professor Müller does is to take the Garrett design, and then correct it for finite sizes of coil. He then optimizes it, using a deterministic technique which was not criticised and which he explains in his report. The design which he produces has 6 coils, a spherical homogeneous region of diameter 44.5cm, better than 20 ppm homogeneity on the surface of the sphere and is about 128cm long. In my judgment, this represents a common general knowledge magnet in 1995.

It is to this technique as elaborated by Garrett that the passage from the specification quoted above refers. It will be observed that the one criticism made of it is its inability to “identify coil properties which will produce the desired B₀ homogeneity for short magnets”.

At this point the specification introduces a new technique for the minimization of errors. This is called “simulated annealing” and it is a different technique from the linear deterministic optimization used by Professor Müller. It is a stochastic technique, depending upon assessing the effect of random small perturbations in specified design parameters subject to specified constraints on an overall error function. In effect, one can specify the maximum length of the magnet, and let the algorithm wander about until it thinks it has found the best mixture of parameters to minimise the error. It appears from the article by Crozier and Doddrell to which I have already referred that it was in the course of such work that they discovered that if the direction of current flow were allowed to vary then the algorithm would reverse the flow in certain coils:

Returning now to the patent, the crucial aspects of the method are described at page 7 lines 13ff. The difference from prior art simulated annealing methods is said to lie in the employment of the number of coils, their radial size and their winding direction as parameters to be perturbed, and in the limited usefulness of the error functions previously employed, which, it is said, should include weighted sums of the field harmonics (i.e. weighted spherical harmonic terms). There immediately follows a series of so-called consistory clauses (page 9 ff). It is convenient to set out the claims alleged to be infringed and/or independently valid here. I have underlined phrases which give rise to difficulties in interpretation. Claim 18 is proposed to be amended by the addition of certain features, which are shown in italics.

Claim 1:

Claim 10:

In the course of argument I suggested that this claim was plainly intended to relate to some method of design of a magnet, rather than any steps carried out on a real magnet. The Claimant had originally suggested that it would be infringed during the testing of magnets by Siemens, but I considered that this approach was untenable and in the event the Claimant abandoned any reliance on this claim.

Claim 18:

Claim 19:

The same objection to reliance on claim 10 must be made in respect of this claim. It relates to the design process alone.

I should add that claims 1 and 10 have dependent on them two other claims adding the requirement that the length of the magnet be less than 1.2m.

The specification then proceeds with a conventional discussion of the solution of Laplace’s equation in the interior of the magnet in a region which does not include any magnetic source (i.e. winding). While this does not call for any comment as a whole, there are occasional remarks that may throw light on the meaning in context of the phrases underlined above. I can gather them together, because they are not extensive.

Page 13 line 2:

Page 14 line 13:

Page 27 lines 4-12 is said to indicate that in the simulated annealing process the sizes of the coils are perturbed radially. I am satisfied that it does not. It is explicit only about axial perturbations and number of windings.

In respect of the error function, at page 28 line 13:

The general simulated annealing method is said to be appropriate for the design of active shim magnets. There was not a great deal of evidence about this, but an active shim magnet can be viewed as “targeted” at a particular harmonic. The object of the design is to minimise all harmonics other than that targeted by the magnet, and the resulting windings are rather complex: see for example Example 4 and the Z₀ shim there described. The magnets do not produce a strong field. It is of the order of milliTesla, appropriate for adjusting a harmonic which may be 2000ppm after manufacture and activation of the magnet.

With this introduction, I can turn to the troublesome phrases.

These words are apt to refer to any system used for producing and detecting magnetic resonance signals for the purpose of measuring the properties of some sample. Obviously, the primary area of interest is whole-body imaging. There is little point in having such a large volume of interest otherwise. But the field at the sample must be uniform so that the resolution is as high as possible, even for the small sample volumes normally used for NMR. The other components in such a system are the cryostat (the claim is only concerned with superconducting magnets) the gradient coils (essential only for image formation) shim coils and the radio-frequency coils together with the associated electronics.

I am not sure why it was thought that this phrase caused any difficulty. It merely identifies an essential feature of any magnetic resonance equipment: the field over the region of interest, whether small or large, must be as uniform as possible so that subject to the imposed gradients the range of frequencies radiated on relaxation of the excited nuclear spins is as well-defined as possible.

That this phrase should cause a problem is not a tribute to the drafting of the specification. Does it mean that the radii of the various coils in the magnet should be the same, or something else? Is it there to distinguish prior art? Is there some problem with simulated annealing and non-common radii which means that simulated annealing will not work, or cannot be shown to work? I must give the phrase a meaning that its context requires (see Kirin Amgen v Hoechst Marion Roussel[2005] RPC 9, Technip’s Patent[2004] RPC 46) remembering that the starting point is, as always, the words that the patentee has freely chosen to use.

I think this phrase might have two contrasting meanings. The first is a general meaning, ‘generally cylindrical’. This covers really all tunnel-like MRI magnets. There are other shapes—cones, for example, or pancakes. Figures 10A and 10B of the specification show pancake-shaped coils which can be designed using the simulated annealing technique and produce a field satisfying the requirements of the claim, but are excluded from the invention because they lack a common inner radius. See Example 3 in general and page 38 lines 20-1 in particular. The same goes for the prior art coils of Figure 2B, which are said (page 4 line 10) not to have a common inner radius. It is difficult to see why, and, indeed, this is not a feature of the same coils identified at page 4 line 10 of the application for the patent as filed.

The narrow, or ordinary, meaning of ‘common’ is perhaps rather more natural. It is suggested that it is nevertheless excluded by clear indications in the specification. The first is Figures 6 and 7, which it is suggested show coils which do not have a common inner radius. Figure 15 is said to be the same and I was shown an enlarged version of it. More significant is the “annealing schedule” in Table 2C, which sets out the steps by which the selected parameters will change between each change of the simulated annealing process, which refers to 1.5% for radial position, and table 2B, which lists initial guesses for the principal parameters for each coil. There is only a little relevant commentary in the body of the specification, which merely says that “the magnet coils were circular in shape and had a specified minimum inner diameter of about 90cm” but these table entries, at least, suggest that the radius of the coils is variable to some extent. Indeed, the first guess, quite apart from not having the 20 coils specified in the annealing schedule, seems to have coils of inner diameter 86 cm, which I suppose is “about 90cm”. It is not possible to draw any firm conclusions from Figure 7, because during the simulated annealing the step size will have decreased. Certainly, a 1.5% step would have shown clearly.

The next point is the passage at page 13 line 2 quoted above: “…the designs may employ primary coils of varying radii.”

A particularly unhelpful point was raised on the passage at page 30 line 4, also relating to the process of perturbation of parameters. This does not refer to the inner radius of the coils as a variable parameter. It is of no assistance.

Finally, Professor Saint-Jalmes points out that it was common in 1995 to wind all the coils on a common former provided with machined grooves for their reception, the position of the bottom of the grooves varying by a few millimetres. The reason he gives is that small differences in radius can make large differences in field purity. This, it is suggested, provides at least some explanation for the “common radius” as referring to the generally cylindrical layout. But the point obviously cuts both ways, because if small variations were likely to be important, their exclusion by the use of the word ‘common’ might with reason be thought to be deliberate.

I must also take into account the fact that the shielding coils, which are present to reduce the field outside the magnet, have a substantial effect on the field in the bore. So far as the patent is concerned, the shielding terms are part of the error function and are apparently parameters adjusted during optimization – see Table 2B, where the two shielding coils in the half-magnet appear at the foot of the table. I cannot think that it can be convincingly suggested that all the coils having an effect on the purity of the field and adjusted during the simulated annealing process have a common radius. The radius is anything but common. In Figure 15 the radius of the shielding coils is 22 cm greater than the primary coils, and by way of example it may be noted that in one of the magnets alleged to infringe, the OR76, ‘the inner coils would produce a field with very little homogeneity…a 20ppm volume of only 50cm³’ without the shielding coils.

The experts did not identify any relevant technical considerations relating to this feature of the claim, apart from the comparatively banal one of being able to place the primary windings on a common former. There is no relevant prior art, and at least some indication in the specification that strict compliance was not intended. In other words, this limitation appears to be arbitrary, and possibly technically wrong, if interpreted strictly, and can be given some meaning if interpreted to mean generally cylindrical. I take it to have that meaning.

I should add a short footnote to this discussion. The interesting short passage in Crozier and Doddrell (1997) to which I have already referred in [16] above appears to suggest that the simulated annealing algorithm does not so willingly produce reverse turns when the primary coils are not ‘on essentially the same diameter’. There was no evidence about this, and I do not think it would anyway be relevant in assessing the meaning of the corresponding phrase in the specification and claims. So it is just a curiosity, but it might explain the limitation.

This phrase gave rise to a very substantial dispute. It colours the whole claim, because Siemens maintain that “shimming” is a well-understood process which takes place after installation and activation of a magnet. What matters (and what mattered in 1995) was the homogeneity of the field in the relevant volume after installation. Siemens say that even though a magnet was designed to have high purity the processes of manufacture, installation and activation would have the inevitable consequence that until the magnet was shimmed its purity might be two or three orders of magnitude worse than the design purity. This is inevitable. Certainly actual purity could not be better than 20ppm until the magnet had been shimmed. So it directly raises the question what is to be looked at when the question ‘Is this better than 20ppm over the homogeneous region?’ is asked.

The first stage is to ask what was meant by ‘shimming’ in 1995. It certainly meant adjusting the magnetic field produced by the primary coils using other windings, permanent magnets or iron shims. So much can be gleaned from the discussion of the use of simulated annealing to design shimming coils in the patent itself (Example 4, page 40). Such coils are specific to particular harmonics and are used to null those harmonics. There was an unsatisfactory attempt to pray in aid something called “theoretical shimming” on the basis of a late witness statement from Professor Crozier, one of the inventors. He had produced a magnet design for a particular magnet in which the Z₈ harmonic had been difficult to reduce (the magnet length limit was set very short) and so it was proposed to incorporate a suitable shim coil to remove it specifically.

The specification itself proposes that higher orders of harmonics (i.e. up to order 2(n-1) where n is the number of coils) might be “nulled” in the simulated annealing process in preference to the lower orders by suitable adjustment of the error function, leaving excessively large low order harmonics to be dealt with by shimming. Obviously a coil specifically designed to remove a Z₄ harmonic in a particular arrangement of primary coils will itself be specifically designed with that end in view: the patent says how to do it. But this is an optional step, and the claim is not directed to a combination of primary and shim coils. It is assumed that shimming may be accomplished by other means. That shimming was a well established and understood technique may immediately be seen from a very general book recording the state of things in about 1988 entitled “Imaging Systems for Medical Diagnostics” (Krestel) published by Siemens and exhibited by Professor Müller which has pictures of the coils used to remove both the zonal harmonics (zeroth degree harmonics), that is, the harmonics which are nulled in the design of the principal magnet, and the so-called tesseral harmonics, which correspond to first (and higher) degree harmonics solutions to Laplace’s equation. At the date of this book, it appears that coils were used to eliminate the harmonics up to A₄.

These considerations lead me to the conclusion that feature (vi) of the claim relates to the process of design of the magnet. The words ‘prior to shimming’ should be taken as a clumsy attempt to indicate that what is being considered is the theoretical magnet design. The contrary argument, that this feature is directed to the field actually produced by the magnet before it is shimmed, seems to me to be improbable, if only because the common general knowledge is that purity may be two or three orders of magnitude out before actual shimming takes place. I accept that is an unsatisfactory construction because avoiding the claim would merely involve somewhat more shimming of the actual magnet rather than further optimization of the design, and the specification itself admits that these are alternative routes to a useful magnet.

I take refuge in the observation that the claim is anyway badly constructed, since its opening words (‘…which produces a magnetic field which is substantially homogeneous over a predetermined region…’) plainly introduce an additional feature over and above feature (vi). Attention was concentrated at trial on the word ‘produces’ to show that the whole claim was talking about the magnet in use. I think this is right, but I think that features (i) to (vi) are design features of the magnet. The words ‘substantially homogeneous’ describe the field achieved after shimming in the final apparatus, using the specified magnet.

Accordingly, the claim is directed to an actual magnetic resonance system incorporating a magnet whose theoretical design has the stated qualities. Given that the claim is directed to the physical system, I consider that the system must incorporate a magnet that in use produces a magnetic field which is substantially homogeneous over a predetermined region of volume greater than 40  10³ cm³. That magnet must also have the design features (i) to (vi).

These words are sought to be introduced by amendment into claim 18. They mean what they say: the magnet must be big enough, and the homogeneous volume big enough, for it to be suitable in 1995 as a whole-body MRI machine, but ignoring improvements in measurement technique etc. since 1995 which might loosen the requirements.

Two magnets are in issue, the OR76 and the OR122. The First Defendant (‘SMT’) used to be called Oxford Magnet Technology and it makes magnets, consisting of cryostats, primary coils and superconducting shielding coils. There are no RF coils, shimming coils or gradient coils, and no electronics. The Second Defendant (‘SPLC”) sells complete MRI scanners.

So far as the OR76 is concerned, only SMT is accused of infringement. Direct infringement under section 60(1)(a)Patents Act 1977 is complained of, as is indirect infringement under section 60(2) by supply of means, the magnet, relating to an essential feature of the invention when SMT knew, or it was obvious to a reasonable person in the circumstances that the magnets were suitable for putting, and were intended to put, the invention into effect in the United Kingdom.

Taking the case on direct infringement first, I think that it is obvious that the magnets manufactured and sold by SMT are not “a magnetic resonance system”. This appears to be accepted, but the Claimant points to the tests undergone by the magnet, which is cold and activated before despatch. These tests include a field test in which a probe is swept over the surface of a notional sphere in the bore of the magnet. The field is measured by examining the radiation frequency of the protons in a small sample, no doubt of water, held in the probe. Is the magnet with the test equipment a “magnetic resonance system” in the sense of the claim ([32] above)? I think it is not, because the assemblage is merely measuring its own magnetic field: it is not measuring anything outside itself. It is really the antithesis of a useful system: it can neither elucidate the structure of crystals or molecules, and it cannot produce an image of something. This is not what the claim is talking about when it refers to a magnetic resonance system.

If I am right in my view as to the meaning of the words “common inner radius” then the magnet as built satisfies this requirement, subject to one point. The primary coils in the OR76 are carried by a 3-part former, the outermost coils having their own supports. The arrangement appears to have been designed to have comparatively direct transmission of the axial forces on the outer coils to the former carrying the remainder of the coils. I do not think that this matters.

So far as ‘20ppm over 40  10³ cm³ prior to shimming’ is concerned, the problem is more difficult. I have interpreted this phrase as intended to ensure that the magnet claimed is characterised by its theoretical design: but the OR76 has two relevant design stages. The initial stage, in which only coils without formers are optimized, satisfies this requirement. In order to compensate for anticipated distortions in the magnet structure as built and activated, the ‘calculated manufacturing design’ has a 20ppm volume of 4 litres, about a tenth of what is required by the claim, and when it is cooled and activated this volume falls further, to less than 100 cm³. Only when it is shimmed and in service does the 20ppm volume rise to 66 litres, well within the limits of the claim.

The manner in which the OR76 design was arrived at is described in paragraphs 16-28 of the Product and Process Description relating to that magnet (paragraphs 19 to 28 of which are confidential to the Defendants). Two things are to be noted. First, the negative coils ‘appeared’ as a result of the optimization technique used, which was not that of the patent.

Second, it does appear that the axial and radial positions of the coils were not set until a second round of optimization, when the diameter of the former (and so the internal diameter of the coils that would be wound on it) had been set. The whole operation was recursive, including parameters such as electromagnetic load, stress, superconducting properties and cost.

Having regard to the complexity of the real optimization operation, I feel very considerable doubt about taking any theoretical design other than the one from which the actual magnet will be manufactured into consideration when considering whether it has a homogeneity of ‘20ppm over 40 10³cm prior to shimming’. On the other hand, the experts seemed to accept that it was a meaningful way of characterising a magnet design to talk of its initial theoretical homogeneity, whatever happened to the design when it was made and installed. With very considerable doubt, and in the absence of any argument from either side directed to this point, I have come to the conclusion that this magnet does satisfy this requirement of claim 1.

For the reason given in [52] above, I do not think this magnet infringes the claim directly during the test phase.

So far as section 60(2) is concerned, infringement by way of supply of means relating to an essential feature of the invention can only occur if the ultimate use was known to be, or was obviously, an infringement in the United Kingdom. There is no doubt that these magnets were incorporated into magnetic resonance systems abroad, and so there will be infringement of claim 1 on re-importation to that extent.

So far as the OR122 is concerned, there is no infringement of claim 1. The volume of the 20ppm region is too small at all stages of the design. The allegation is made in respect of claim 18 as unamended and as proposed to be amended. Infringement by SPLC by importation of a MRI apparatus called the “Magnetom Espree” incorporating the magnet is also complained of.

The magnet is developed from the OR76. It contains a single superconducting shim winding underlying each of the primary coils, but apart from that feature (which does not permit full shimming of the magnet for the purposes already discussed) is not distinguishable for present purposes from the OR76. A subsidiary issue, introduced by the proposed amendment, is whether this device is suitable for whole-body MRI. It plainly is. So claim 18 is infringed directly by SPLC by the supply of the Magnetom Espree, and indirectly infringed to the extent of sales etc. in the United Kingdom by SMT.

Two principal objections are taken, added matter and obviousness. It is also suggested that the specification is bad for what is usually called Biogen insufficiency. This last is an important objection and raises difficult issues, but loomed less large at trial than the others.

By Section 76(2) of the 1977 Act

Siemens say that there was no disclosure in the application as filed that the magnets in question had a common inner radius, and that all magnets, including the pancake-shaped magnets of Figure 10, were to be treated equally, so far as the disclosure went. It is said that not only was the phrase not used in the application, but its use to characterise the invention adds subject matter.

The test in the UK context has been described by Aldous J in Bonzel v Intervention (No 3)[1991] RPC 553 at 574 as follows:

to ascertain through the eyes of the skilled addressee what is disclosed, both explicitly and implicitly in the application.

To do the same in respect of the patent as granted.

To compare the two disclosures and decide whether any subject matter relevant to the invention has been added whether by deletion or addition. The comparison is strict in the sense that subject matter will be added unless such matter is clearly and unambiguously disclosed in the application either explicitly or implicitly.’

So far as the disclosure of the application as filed is concerned, Siemens are right. There is no disclosure of the coils of the magnet possessing a common inner radius. Professor Saint-Jalmes says nothing about it. The furthest that the Claimant is prepared to go by way of submission is that the coils disclosed, with the obvious exception of Figure 10, do have a common radius, and, I infer, accordingly provide support for this feature of the claim. As it happens, I doubt this, since the specification is silent as to the significance of the feature. But I am in no doubt that the fact that the coils of the magnet have a common inner radius is not disclosed in the application for the patent, and is disclosed in the granted patent.

The test can really be answered by reference to Figure 2 of the drawings. Figures 2A and 2B show different coil conformations. The application says:

By way of contrast, the patent says this:

The reason for the introduction of this feature to the disclosure is still unclear to me, but its consequence is incurable invalidity. None of the claims in which this feature appears can be amended to remove it, since such an amendment would increase their scope and so offend against section 76(3)(b) of the 1977 Act. So the patent must be revoked.

Only one citation is relied on, US patent 5,343,182 (‘Ohta’). Ohta took up a lot of time at the trial, it being suggested that Professor Saint-Jalmes had approached it in an unfair and inappropriate way, and that Professor Müller had approached it with what might be called 20-20 hindsight. The dispute became very involved.

When obviousness on the basis of a documentary disclosure like Ohta is being considered, the question for the Court is whether the alleged invention is obvious to the skilled person at the priority date of the claim in the light of the matter disclosed in the document. Windsurfing International v Tabur Marine [1985] RPC 59 at 73 is taken as providing a checklist of things the court must do, remembering that the ultimate question is, was it obvious? I have amended the passage slightly to reflect the terms of the 1977 Act:

I must now examine the common general knowledge in a little more detail. At the priority date in 1995, there is no doubt that the manufacture of high-homogeneity superconducting magnets with an inner diameter of the order of 1m was well established. Field purity requirements rather stricter than the claim (20ppm over 47-65  10³cm³, i.e. 45-50 cm diameter) were conventional for whole body MRI. Table 10.4 in the Krestel book exhibited by Professor Müller compares superconducting, resistive and permanent magnets and gives these sorts of values as typical. There is no doubt that quite apart from the usual quest for image quality, one of the things which was of interest to the manufacturer of MRI machines, and so his team of designers, was the length of the magnet, for the reasons I have already stated. In such magnets, it is the outermost pair of windings which have the highest number of ampère-turns.

All the knowledge necessary to construct a superconducting magnet with a 1 m diameter bore, 2 m length and a 0.5-2 Tesla field was available and used. This includes such matters as the nature of the windings (very fine niobium/titanium or niobium/tin filaments in a copper sheath or matrix) the former carrying the windings, the accommodation of the very substantial forces on the windings and the construction of the cryostat which contains liquid helium or other means to keep the windings at superconducting temperatures.

A common type of MRI machine in 1995 used a tunnel shaped magnet, equipped with shimming magnets both passive and active, gradient magnets and RF coils.

At the priority date, the skilled man understood the theory of the field in the bore of the magnet and used Garrett as an aid to calculating it. In particular, the skilled person knows that there can be a region of high homogeneity at the centre of multi-coil arrangements in which the coils are correctly adjusted for position, radius, and current-turns, and that the approach taken by Garrett was to investigate the fluctuations in the field at the surface of a suitable volume (sphere or ellipsoid) centred at the centre of the magnet. Having decided on the dimensions of this shape, the contributions of the various coils to the variations of the field over the surface of the shape can be calculated with very high precision using methods available since James Clerk Maxwell. Knowing that the variations inside the shape will be smaller than the variations on the surface, the homogeneity available from such-and-such a magnet can be stated in terms of a peak to peak variation in total magnetic field over the surface of a shape of such-and-such a size. The precision was always available, but until the advent of the digital computer tiresome to calculate, as Garrett (1967) makes clear.

In principle, therefore, the skilled man would have understood that the field from each of the coils in a magnet design at the surface of a shape such as a sphere centred at the centre of the magnet and having a particular radius could be represented by a sum of the functions called spherical harmonics. The analysis of these functions would be known, so far as relevant, and the basic message, which is that harmonics of all orders and degrees are impurities except the harmonic of zeroth order and degree, which is the desired pure field, would be very well known to him, as was the mandatory requirement to minimise, so far as possible, the impurities by adjusting the position, radius and current carried by the coils.

Also by 1995, the skilled person knew how to optimize a design and to use a digital computer for the purpose. The process of optimization consists of identifying parameters to be varied and the constraints upon them, and devising an error function which reflects the desired result. One then varies the parameters in a systematic manner to investigate their effect on the error function with a view always to minimising it. There are many techniques. The results will be heavily dependent, obviously, on the constraints placed on the parameters. For example, an optimization in which cost is a parameter may well have a different outcome from one in which cost is disregarded — Professor Saint-Jalmes made this quite clear. Particular methods of optimization were well known: Both professors refer to Newton-Raphson root-finding and Professor Müller to linear programming. These are deterministic techniques which use the result of the previous approximation as the starting point for the next. They are not guaranteed always to work (a reason for starting with a known design) and they may miss the true optimum.

For this reason, it is not practicable to start every design from a blank sheet of paper. The skilled person will start with a known design before specifying his own parameters and constraints. There was agreement that Garrett’s 12^th order design, to which I have referred above at [13], was a standard, common general knowledge design; ‘well-known’ as Professor Saint-Jalmes described it. Professor Müller still had the copy of the paper he had obtained in 1980 and had used as a reference for the magnets he had designed. The example from that paper which he chose had three pairs of coils, and using it as a theoretical starting point the designer could minimise up to the 10^th order of impurities in a corresponding real magnet. The level of homogeneity achievable with a 12^th order magnet is fine for MRI, as I understand the evidence, subject to the problem of length.

Finally, I should add that by the priority date simulated annealing was a known method of optimization of a stochastic type, in which random variations in the parameters are used in an attempt to find a true optimum, and had been applied to the extent that the specification acknowledges for optimizing magnet designs, but it it is not suggested that it was obvious to employ it as the specification describes.

The inventive concept is what is in claim 1, stripped to the bare essentials. Mr Prescott QC submitted that this was a magnet which included coils having a configuration, i.e. configurations in which the current in at least one primary coil flowed in the opposite sense to at least two adjacent remaining coils. He said that the discovery of this possible configuration opened up a ‘rich seam’ of possible designs, shorter than designs in which the current flowed in the same sense in all the primary coils. I think it may be expressed rather differently. The concept is the addition of at least one opposite sense coil to a magnet of at least two coils, so reducing length while maintaining homogeneity. There was surprisingly little evidence that the invention has been widely used, but I assume it made an advance.

The source of the inventive concept — a product of an optimization algorithm unconstrained so far as current flow is concerned — is in principle irrelevant. Nor is it relevant that once a magnet of this type has been designed, it is possible to build on the design using deterministic optimization algorithms to obtain other magnets. It is, however, relevant that the configuration will necessarily be present in a magnet with four coils or above to which a pair of reverse windings has been added.

There was a serious dispute as to what the prior art, Ohta, discloses. Ohta is a US patent, assigned to Toshiba, and is concerned with magnet devices for MRI systems. The abstract is a helpful summary:

Passing over the figures (I shall refer to them in their place) Ohta states (column 1 line 1) that it is concerned with magnets for MRI. It says that a high static magnetic field is used which is highly uniform. It acknowledges the employment of even and odd axisymmetric shim coil pairs for providing a ‘correction field’ for an MRI magnet, but objects that the use of such coils increases the radial size of the device and necessitates additional power supplies. It then acknowledges a vertical magnetic device (col 1 line 42) in which a pair of sub-coils are arranged ‘ together with main coils producing the main magnetic field, within a fixed angle range with respect to the central point of the static magnetic field.’ It states the problem with which it is concerned at column 1 line 54:

The object of the invention disclosed is said to be ‘to provide a magnet device which can be produced into a compact form in both the axial and radial directions and which can generate a uniform diagnostic space in a static magnetic field’. The description of the invention is substantially the same as the abstract quoted above ([82]). It follows that the currents in the sub-coils will be reversed, and at col 2 line 50, various preferred arrangements are described: (line 50) the sub-coils should be disposed so as to suppress high-order components of a series expansion of the axial component of the field; (line 55) that when the main coil assembly has three or more main coils and the sub-coil assembly comprises one pair of coils, the latter should be arranged to be next to and in board of the outermost pair of main coils; and (line 60) when the main coil assembly has five or more coils and the sub-coil assembly two pairs of subcoils, the latter should be arranged with one coil of the pair between the first and second main coils and the other coil of the pair between the first and third main coils.

Turning to the drawings, Figure 1 is said to be an axially-cut sectional view of an upper half part of a superconducting magnet device according to the invention. This diagram was used by Professor Saint-Jalmes in one of his unsuccessful attempts to ‘get Ohta to work’. What this attempt involved is described in paragraphs 108 ff of his report. He observed that the diagnostic space was depicted by a circle in this diagram: and so he scaled it, being told the diagnostic space was 30-50cm in diameter. He mixed in coil data from Figure 7, which is not said to relate to Figure 1 at all. Having carried out this operation, he drew the magnet out (report para 116) in a form which enabled him to add the ‘z-derivative cancellation lines’ of Fig 4 of the patent, which I should now describe.

Figures 3 and 4A to C of the patent are respectively ‘a graphical representation showing behaviour of axial components, corresponding to error terms, of a static magnetic field’, and ‘graphical representations showing necessity of sub-coil assembly’—see col 3 lines 58-63. Professor Saint-Jalmes does not explain his interpretation of these figures expressly, but Professor Müller gives a straightforward explanation. Starting with a simple diagram of the forms of the spherical harmonics along the axis

The idea of the change of sign in the harmonics is crucial. You cannot add harmonics having the same sign and hope that they cancel out. The coils must be spaced accordingly. Figure 4A is a diagram of a system in which the field is uniform at the centre.

In addition, the specification states in respect of Figure 4C that the coils may be wound without overlap, that Lc may be composed of four ‘small width coils’ shown (touching) in Figure 5, and that the Figure 5 arrangement has a current density distribution (it is agreed that this is loose terminology and ampère-turns is meant) as shown in Figure 6. This figure shows two schematic winding layouts, one a 6 coil arrangement with all coils carrying current in the same direction (labelled CVa) and the other (labelled CVb) showing 6 forward coils and 2 reverse. The CVb arrangement is shorter. The field distribution of this arrangement is shown in Figure 8 (see col 7 lines 12-32), which shows the 1, 3 and 5 ppm contours. This Figure suggests that the components to Z⁶ have been nulled, the magnet having being 8^th order.

In seeking to put the teaching of Ohta into effect, Professor Saint-Jalmes, as I have said, squeezed such dimensional information as he could out of it, and constructed a magnet. He found that the coils were not in the same position vis-à-vis the sloping zero lines as shown in Figure 4C (compare his Figure 11) so he reduced the radius until the pictures were rather similar (his Figure 12). Using, therefore, a diameter too small to admit a human and the magnetomotive force (mmf, units ampère-turns) specified by Figure 7 for the arrangement of Figure 6, he obtained a magnet which would not work to produce a uniform field. This establishes that Ohta does not anticipate claim 1.

Taking this approach, by his second optimization (paragraph 137 of his report) he had produced a winding within the claim, at least if the arrangement shown on page 15 of Appendix 1B of the Claimant’s notice of experiments has the necessary number of windings. His objection, that there is coil overlapping, may take it outside the claim. But his second, that it ‘is not Ohta’, deserves closer attention. Professor Saint-Jalmes approached Ohta as if it were to be emulated in dimension and appearance as closely as possible. So he regarded Figure 1 and Figure 5 as disclosing different designs, and moved to Figure 5.

In his Figure 5 work, he came quickly within the claim. Using data from Figures 6 and 7, and merely scaling the drawings he obtained very nearly Ohta’s quoted figure for the homogeneous field (0.33 Tesla as opposed to 0.35 T) but the uniformity was very poor. Recognising that accuracy was not available by scaling the drawings in the patent specification, he optimized the design while insisting that the four small coils remained together. Uniformity was still very poor. Allowing a constraint of ±150mm on each of the small coils produced a result which converged, with coils with a common radius, to values well within the claim. Professor Saint-Jalmes’s objection to this magnet, and with his second successful result on the basis of Figure 5 of Ohta, was that they did not look like Figure 5 of Ohta (see appendix 1b of the Claimant’s Notice of Experiments, pages 33 and 35).

There are a number of observations about what Professor Saint-Jalmes did. First, he tried to implement Ohta on the basis of its own drawings. So far as Figure 1, a general sectional layout, was concerned I think this was misguided. So far as Figure 5 was concerned, he produced magnets within the claim. He had worked through a number of possibilities by the time he allowed the small coils to move independently, but then success appears to have come rapidly. Although he used his very extensive knowledge of the relevant variables, step sizes and constraints during optimization, he effectively created each design from scratch.

Professor Müller took a rather different line to Ohta. He said that Ohta taught that ‘by incorporating reverse polarity coils (referred to by Ohta as sub-coils) into the primary magnet, one can produce a magnet that is shorter but yet has a useful homogeneous volume.’ He was cross-examined on this and accepted that the words ‘according to the principle he teaches’ could be inserted in this statement. I do not know what this was intended to convey to the witness—it conveyed little to me—unless it was supposed to suggest that Ohta taught a principle related to the algorithm referred to at column 7 line 51. But it does not matter: this was Professor Müller’s approach. It was also that of Professor Saint-Jalmes (paragraph 102 of his first report):

In paragraph 5 of his second report, Professor Saint-Jalmes seems to change the emphasis:

In other words, Ohta seeks to keep the radius as compact as possible while reducing the length. This does not materially affect the “take-home message”. Professor Müller cannot be convincingly condemned for hindsight if what struck Professor Saint-Jalmes first was the reduction in length made possible by the negatively wound coils.

Accordingly, Professor Müller did not attempt to build a Ohta magnet from the ground up. He says, and I accept, that Ohta does not give sufficient dimensional information to start constructing a magnet, and so he did what it was agreed magnet designers generally do, which is to start with an existing design which is known to work. He started with the Garrett 6-coil 12^th order design from the 1967 paper, which was agreed on all sides to be a well known design. He added two negative coils inboard of the outermost coils. He optimized it, and it came out shorter and within the claim. His work, compared with that of Professor Saint-Jalmes, was not extensive and the results set out in the spreadsheet he constructed. His results are not challenged.

The Claimants submit that as a matter of disclosure the skilled man would start with Figure 9 of Ohta, which lacks the feature. It was never explained to me, and I do not understand, why in the light of the passages in column 2 I have quoted above ([84]) it was not perfectly obvious if one wanted to use two coils, to use them in the manner stated, whether or not it was also obvious to emulate Figure 9. The teaching is quite general. Siemens are not obliged to show that everything obvious in the light of Ohta must fall within the claim, but merely that the claim covers some things that Ohta makes it obvious to do. Indeed, most of Ohta’s explanation is given in respect of the 2 sub-coil embodiment, and that might be a slight reason for starting with two coils. In the absence of a worked out magnet design, surely to start with a common general knowledge design is an obvious thing to do, and motive is provided for such a step by Ohta’s suggestion that the magnet may be shortened, a result he considers helpful for claustrophobic patients. There is no question of combining the teaching of two unconnected documents. Garrett is common general knowledge and the design in question was well-known to Professor Saint-Jalmes.

Mr Prescott QC submits that the approach taken by Siemens misunderstands the design process, with the consequence that I am in danger of seeing it as merely playing around with coils. On the contrary, it is because the design process is so highly sensitive to small dimensional changes that it is plausible that the skilled man would start with an existing, working, design. Secondly (a minor point) he submits that it was not obvious to ‘divide up’ the central solenoid into the small coils referred to by Ohta and permit those coils to float during the optimization process. This is not an objection to the allegation of obviousness as it has been developed, which is to take the broad teaching and apply it to a known working design.

Finally, I should refer to Mr White’s work. Mr White was an employee of SMT from 1985 to 2005. He was a designer of cryostats, and up to 1999 he acted as the head of magnet design. In 1995, it was suggested to him that he investigate the effect of incorporating a negative pair of coils to two magnets. He thinks it was suggested to him by Frank Davies of SMT, but he knows that he was referred to Toshiba’s European patent application 0496368, which is the same as Ohta. His report, entitled ‘shortened OR70 with a negative coil’ begins with the sentences

Of course, this is exactly what Siemens say is the obvious thing to do with Ohta. The system length was reduced somewhat (by 2.1cm) and the wire cost increased. The changes to the OR70 brought it within the claim of the patent in suit, Mr White having used the optimizing program at SMT to recalculate the magnet. So Siemens say that I should at least draw comfort from what Mr White did in saying that Professor Müller’s approach was reasonable and that it gives rise to obvious designs of magnet falling within the claim.

It is important not to rely on Mr White’s work as primary evidence of invalidity. It is not. Because Mr White was not cross-examined, I cannot be sure that he was representative of the skilled person, although Professor Saint-Jalmes could not point to anything he did that was not routine. Nor can I be certain that he did not possess excessive skills. Nevertheless, it is secondary evidence against which it is legitimate to test the evidence of the experts, at least for the purpose of satisfying oneself in a complex art that the suggestions made are not outlandish or, perhaps more importantly, tainted with hindsight.

The balancing operation which must be carried out in assessing obviousness is difficult, the more so when the evidence in favour of obviousness and of non-obviousness has the unusual features of this case. Professor Saint-Jalmes’ experiments, and Mr White’s evidence, are both unusual, and really inconsistent. It is necessary always to guard against the temptations of simplicity: and that is what Professor Müller’s approach amounts to. What seems simple now is attractive, being easier to understand, than the complexities which might have encircled the skilled person at the date. On the whole, I consider that the claim does cover what is obvious, and none of the allegedly independently valid claims can survive. The action fails and the patent must be revoked for this reason also.

It is not necessary to consider Biogen insufficiency in detail. Given that the criteria of the claim, with the exception of the reverse current coils, were common general knowledge at the date, and since the evidence (that of the experiments and of Mr White in particular) shows that it was possible to design a magnet at the priority date without using simulated annealing and arrive within the claims, the claims must either be insufficient or obvious unless limited to the non-obvious method of design represented by simulated annealing. Siemens have been careful not to suggest that the teaching of the specification was obvious in this respect, merely that the claims were too broad.

The action fails and the counterclaim succeeds.