Cancer Detection by NMR in the Living Animal

The purpose of this paper is to review in vivo NMR experiments [1, 2] on a transplantable tumor in mice and to discuss the feasibility of using noninvasive NMR for cancer detection in humans.


Introduction
R ece ntly, pulsed nuclear magne ti c meas urements have been mad e on biological tissues , with th e observation first b y Da ma dian [3,4]\ and subsequently by others [1, 2, 5-9] that a variety of neoplasms display differe nt spin -lattice (T\) and s pin-s pin (T2 ) relaxation time s than corres ponding normal tissue. Thes e differe nt relaxation times occur with tumors of diverse histologic typ e, with tumors of human and animal origin, a nd for tumors transplanted in the live animal or for tumors which have bee n excised before measure me nt. Although the physical mechanism of these T\ and T2 values has not yet been determined and is still a matter of co ntroversy, these findings raise the possiblity that the principles and techniques of pulsed NMR might be adapted to detection and diagnosis of cancer without the need for surgical intervention.
The technique utilizes the magnetic resonan ce properties of the atomic nucle us (in this case protons) when subjected to a steady polarizin g magneti c field and radio frequency e xcitin g pulse at the appropriate frequency. Two important processes, spin-lattice relaxation and spin-spin relaxation, characterize the reso nant nu cle us and are meas ured in a pulsed NMR experim e nt. Spin-lattice relaxatio n is the process in whic h th e spin of the nucle us reac hes a n eq uilibrium orientation in the polarizing field after bein g perturbed by the radio frequency pulse. Spin-spin relaxation is the process in which resonant nuclear spin s, initially precessing co herently in the polarizing field , lose s ynchronization with eac h other. Both processes are mediated by interactions betwee n the nuclear spin and its static and dynamic atomic environment, which includes motion of other nuclear spins. In simple cases the relaxation processes are each charac-terized mathematically by a single exponential decay or recovery, in which case a single characteristic time , T" describes spin-lattice relaxation, and a single tim e , T2, describes spin-spin relaxation.
Experience suggests that it is often possible to associate a single relaxation time with a particular tissue or tumor. For spin-lattice relaxation measurements using a 1Tt -n/2 pulse sequence, as indicated in figure 1, the recovery of the nuclear magnetization M (t) as a function of the time t between the two pulses IS (1) Frequently the point at which the recovery curve crossed the 0 axis is used for determining T,. This is often referred to as the null-point method.
For spin-spin relaxation, using the Carr-Purcell-Meiboom-Gill (CPMG) pulse sequence, the magnetization M (t) decays with time t according to (2) In general, the stronger the interaction between the nuclear spin and other nuclear spins, the shorter the relaxation times. Pure water is an example where eqs (1) and (2) apply and in which a specific value of T, and of T2 are appropriate. Strong diffusional motion of the protons in water leads to a relatively weak interaction between protons and thus long values of T, and T2 (2 to 4 s). As discussed below, it is important that the full relaxation curve be measured in both spin-lattice and spin-spin measurements to ascertain if there is a unique T, or T2 before final conclusions are drawn with regard to the significance of the numbers measure. [1,2,10-12]'

2_ 1_ Spin-Lattice Relaxation
For biological tissue more than one time constant may be involved because of the complex nature of the specimen. It is thus desirable to generalize eq (1) as follows. n 11 to a time long enough to reduce the signal M (t) until it is comparable with the background noise. The initial signal-to-noise ratio must be high at least 100 to 1. In the case that there is only a small amount of tissue with a long relaxation time, then even this criterion may not be satisfactory. Such a situation would require even weaker signals to be measured and higher signal-to-noise for the combined tissues. Then, using semilogarithmic plots it should be possible to extract the true values of Tt . This involves, of course, the corresponding amplitudes that are also unknown a priori. It is believed that this type of analysis can be successfully carried out for n = 2 and n = 3 and possibly for higher values provided that the T, constants are sufficiently separated. Analysis could result in the detection and identification of one or possibly more neoplasms existing in a host or environment of normal tissue, provided that the individual tissues had different relaxation times, T,. Furthermore, repetitive measurements taken as a function of time can reveal the growth of the tumor as a relative change in the volume of tumor which shows up aE; an increase or decrease in amplitude Mi associated with the particular (T,) i and normal tissue. A change from benign to malignant might be revealed as a change or appearance of a T, associated with the tumor.
The type of response that would be obtained by the use of 1Tt -1T/2 pulse sequence illustrated in figure 1 has been calculated for single and multiple spin-lattice relaxation times, as shown in figure 2.
The assumed values of T, were 0.3 sand 0.7 s in the various proportions indicated. It is diffic ult to distinguish, by visual observation, the single-exponential curves (a and e) from the double-exponential ones It is obvious that a double-exponential curve cannot be appropriately described by a single spinlattice relaxation time, but if it is insisted that a single ~ o ;:: In the event of two or more terms in eq (1) it is necessary to determine the various corresponding values of T,. First it may not be immediately clear as to the value of n and the range of Tt's encountered.    (curve e) is reached. In contrast to figure 2b, c, d, the departure from a single exponential response can be discerned in figure 3b, c, d by a simple visual observation. Furthermore, by the use of a straight edge, an approximate estimate can be obtained of the "slow" time co nstant involved in this figure along with its relative amplitude. It is clear in the experimental data which we present below in this chapter that there is at least an additional faster relaxation occuring than that associated with the normal tissue. For this reason additional curves have been synthesized to indicate the effect on the spin-lattice relaxation behavior. It would require exceedingly careful measurements, high sensitivity, and data acquisition over a wide time scale to distinguish between relaxation of the types shown in figures 4 and 5.
In figure 4 more than two time constants are introduced, as shown, but still restricting the calculations to two exponentials. The slow component maintains an unchanging value of T t = 0.7 s. The fast component, however, was made to vary linearly from  Figure 5, as a result, illustrates the relaxation behavior to be expected for tissues with three different time constants and the difficulty in resolvin g them into their separate exponentials. The similarity of the 3-exponential curves of figure  5 to the 2-exponential curves of figure 4 is striking.

Spin-Spin Relaxation
In analogy with the spin-lattice relaxation case the data for spin-spin relaxation might be analyzed in terms of exponentials of the form ---L.,,; ie 21 Mo i= t (4) in which (T2 ); are the component spin-spin times and Ai the corresponding fractional amplitudes which depend upon the amounts of particular tissue with time constants (T2k However. there is evidence [2] that even a tissue that is characterized by a single Tt, can not be characterized by a single T2 • As in the case of spin-lattice relaxation, it is mathematically possible [10] to analyze the nonexponential part of spin-spin relaxation into n parts each with amplitude Ai and time constant (T2 )i. The analysis to be followed is then similar to that described above for spin-lattice relaxation. It is not clear that the (T2 ); thus obtained are physic ally meaningful, but such an analysis into parame ters may be useful for monitoring tumor growth. Further work is necessary in order to correlate tumor growth with these parameters.

Brief Review of In Vitro Experiments
Damadian [3] (See also this volume) compared relaxation measurements that he made in vitro on various normal and malignant tissues from rats. He reported that the Tl and T2 values obtained by the null and lIe methods respectively were larger in the malignant than in the normal tissues. In some cases these differences were of the order of 10-20 percent, in others factors of 2 or 3. For example, values of Tt for normal tissue from rats varied from 0.3-0.6 s, whereas the Tl values for two typical malignant tissures was in the neighborhood of 0.7-0.8 s. Similarly, the T2 value was about 0.05 s for normal tissue and was 0.1 s for malignant tissue. Note that it was assumed that relaxation was a single exponential in these cases and the values reported were measured using the null method.
Further work has been reported [4] in humans in which Tt measurements on excised normal and malignant tissue have been compared for breast, lung, muscle, skin, and intestine. Again Tl was found to increase in the tumor tissue relative to corresponding normal tissue by factors of 1.5 to 3, with breast showing the biggest difference. A singular exception was found for melanomas where the Tt values were depressed relative to the host tissue (normal lymph node).
Comparisons of Tt and T2 for mammary glands removed from mice by Hazlewood et al. [6] have shown a lengthening of both relaxation times in the pre neoplastic nodules and neoplastic tissue relative to normal tissue. In this case full spin-lattice relaxation curves were plotted and found to be described by a single exponential decay. Spin-spin relaxation was separated from diffusion effects using a measurement technique tha t did not show th e full relaxation curve_ The diffusion co nstants were found to in crease along with TI and T2 in preneo plasti c and neo plasti c tissues co mpared to the diffu sion co ns tant in normal tissue. This observation supports the mod el proposed by Damadian in whic h tumor ti ss ue has less intracellular water structure than normal tiss ue (he nce more diffusion and less inte rac tion be twee n protons and their s urroundin gs res ultin g in lo nger relaxation tim es). Th e s trong inte raction s be twee n proton spin s in normal tis s ue as e vide nc ed by th e meas ured TI and T2 bein g s horter than th e corres pondin g para me ters in pure wate r had bee n attributed to a majority fraction of bound water molec ules [3-5 , 13]. Other work [14][15][16][17][18] is not in agreement with this theory of relaxation in normal cells. Some recellt work [19] indicates a lac k of co rrelation be twee n diffu sion tim es and s pinlatti ce and s pin-s pin relaxation times in normal rabbit le ns ti ss ue s ugges ting that stru c ture or c rystallinity of a large fra c tion of cell wate r do es not account for the observ ed s hort e ning of relaxation times in tissu e relati ve to ordinary distille d water.
Pulsed NMR studie s have also bee n made on cell s in terins of diffu s ion [20] a nd of diffu sion and molecular exc han ge [21].
Hollis a nd co work ers [8 , 9] have also found th at s pin -latti ce re laxation tim es, as meas ured usin g the null me th od , in mali gnant tum ors are in general lon ger than th e corres pondin g norm'll tissu e. In one study th ese a uthors associated differe nces in tu mor growth rates (slow versus fas t) with th e relative magnitude of TI • For exa mple, th e more rapidly growin g tumors have the lon ger TI values whereas the slower growin g tumors te nd to overlap the TI values for normal tissu e. Hollis e t al. [8] have carri ed out a more exte nsive study on a large number of tumors in animals and human s. While th e animal tumors s howed s pinlatti ce relaxation tim es well in excess of those for normal ti ss ues the differe nces in four human tumors relati ve to normal tissu e was less cl ear c ut. Accordin g to th ese authors it is necessary to explore further the optimization of fac tors which enhance the relaxation time differe nces be tween normal and malignant tissues such as temperature and frequen c y dependences and a more complete study of the full relaxation curves. They point out that there are alternatives, such as the effect of bound paramagnetic impurities [10], to the mechanism propose d by Damadian (i.e., crystallinity of cellular water) to explain the differences between normal and malignant tissues.
The value of in vitro determination is clearly evident from the correlation of the NMR data with neoplasms and non-neoplasms. In addition, such information is then available for in vivo studies where applicable. This is obviously important for its use for noninvasive diagonosis of can ct"!f.
Ce rtain limitations exists, however, for in vitro applications. An obvious one is co ncern ed with what ne w and possibly unknown variables are introduced by the incision and th e removal of the specim e n for e xamination. In what manner, if any, is the water co nfi guration and behavior altered by this process? It is necessary to make measurements on the same tissu e, normal or tumor, before and after removal from the livin g a nimal to determine whether in fact there is a diffe ren ce.

Neoplasms In Vivo
We now re port on th e inves ti gati on which s hows th at it is pos sible to d etect a nd monitor the growth of cancer in a live animal by means of nuclear spin-lattice and spin-spin relaxation measure ments. Dele te riou s effects introduced by the removal of the specimen from the host including special preparation of the sample are obviously absent.
Studies on tissue removed from the body have th e advantage of selectivity in that the investi gator can carefully dis sec t out the ti ss ue of inte res t and observe reso nances from only its protons. It is a mu c h more diffic ult tas k to pe rform and analyze nuclear magneti c relaxation e xperim e nts on live animals because in ge neral many kinds of ti ss ues, only a fe w of whic h are of inte rest) co ntribute to th e resonance signal. It is pos sible, howe ver , to obtain useful s pin-lattice and spin-sp in relaxation data from protons in th e livin g animal. For exa mpl e, such data was o btained by transplantin g a Cloudman melanoma into the tails of DBA/2 mice. [1 ,2 ].
One proble m is that of sensitivity, i.e. detecting in th e human body a relaxation tim e differe nce from a relatively s mall number of mali gnant cell s. It is im-pOl·tant to determin e, in detail, th e tim e-de pe nd ent c han ges in relaxation that may distinguis h betwee n normal and mali gnant tiss ues in live animals. For thi s purpose, it is absolutel y esse ntial that the full re laxation c urves be examin ed . Although the motion of the animal contributed to the nois e, it was possible to obtain significant data without the use of a nesthesia.

. Experimental Conditions
Because of the size limitations of the available magnet, it was most convenient to perform this initial experiment on the tail of a mouse, which being lon g, cylindrical, and narrow, is easily inserted in a small probe. A schematic diagram of a mouse being confined to a small plastic cage is shown in figure 6. The tail is taped to an extension to be inserted in the rf coil. The rf coil is in a probe assembly placed between the poles of an electromagnet, shown in figure 7.
An example of the neoplasm studied is illustrated in figure 8. A Cloudman 591 malignant melanoma was transplanted into the tails of DBA mice. The NMR response of the tumor, as well as the be havior of the immediate and adjoining tissue in the tail, co uld there· fore be ascertained as the tumor grew. Histologically, the Cloudman melanoma is a pleomorphi c anaplastic tumor whi ch exhibits extensiv e areas of necrosis. The necrosis tends to occur ce ntrally while peripherally the tumor co ntain s viable cells whic h proliferat e and expand the tumor mass ( fig. 9). Melanin pigment is not present in the tumor cells. Occasionally in necrotic areas, viable tumor cells remain circumjacent to vascular channels. Clearly, the NMR curves represent contributions from both the viable and necrotic parts of the tumor. In large tumor masses, greater than 60% of the tumor becomes necrotic. Measurements were made at ambient temperature with a phase·coherent and pulse·coherent 5·k W spectrometer using the laboratory magnet illustrated in figure 7. The amplified nuclear signal voltage fed to the phase detector, was maintained at a value much less than the detector rf reference voltage in order to insure linearity. The magnet was a water· cooled 4 in laboratory electromagnet. Measurements were made at field strengths corresponding to proton frequencies ranging from 8 to 24 MHz. A rotating frame HI ~ 25G was sufficient to saturate the protons in the mouse tail with a single rr turning-angle pulse. Either a rrt -rr/2 sequence as shown in figure 1 or a rrt -echo sequence in which the echo·forming pulses had a spacing~ T!, was used to measure TI • For the measurements of T2 , a Carr-Purcell-Meiboom-Gill sequence conslstmg of a rr/2 pulse followed by a chain of rr pulses (phase shifted by rr/2 with respect to the first pulse) was employed. In order to avoid nonexponential behavior associated with diffusion in a magnetic field gradient, the rr pulses were kept closely spaced (~ 1 ms apart). Signal averaging was accomplished with a gated integrator and with a digital signal averager. In many experiments the data gathering and storage was automated. In this arrangement recording of spinlattice relaxation could be accomplished quite simply. The pulse sequence programmer would sequentially advance the delay between the rr saturating pulse and Actual coil used in s tudy must be shielded against RF leakage and hen ce would be unobse rvable whe n probe is in place.  echo monitorin g signal a small in cremen t of time !1t with each repetition of the 7T -t -echo sequence. The echo amplitude correspondin g to each spacing twas stored synchronously in an appropriate channel of a signal averager. By re petitively recycling the programmer, many sweeps through the full spin-lattice relaxation (with 100 point resolution) co uld be added togethe r and appropriate relaxation times derived.
In the case of spin-spin relaxation the signal averager was externally advanced in synchronism with the pulse programmer so as to sequentially store the echo amplitudes appearing between 180 0 pulses in the Carr-Purcell chain. Many repetitions of the chain were averaged together to obtain the transverse relaxation curve.

Spin-Lattice Relaxation
There exists a distinc t difference between the spinlattice relaxation behavior of normal tail and of melanoma tissue in the mice. The spin-lattice relaxation was measured on the tails of a number of normal mice and in every such case was found to be characterized by a single time T1• The value of TJ , depended on frequency or magnetic field, as shown in fi gure 10.
Similar frequency dependence was reported for in vitro measureme nts by Outhred and George [I8], and for hi gher frequen cies by Damadian et aL [4]. The latter results are shown in figure 11 , where an increase in T, was found for in creasing frequency, both for neo plasms and for co ntrol tis s ue.
Th e scatter s hown for our data ( fig. 10) may be in part instrumental , since it includ es res ults obtained before automation of the data gatherin g and storage, but it also may be due to biological variability.
In protein solutions TI was found [22,23] to be dependent upon frequency as well as other parameters such as temperature and existing paramagnetic proper· ties. TI was found to decrease markedly with decreasing frequencies below 2 MHz.
Just as the spin·lattice relaxation was exponential for normal tissue, so it was exponential for a well· developed melanoma, as is evident from figure 12, where the data extends over two orders of magnitude. The relaxation time associated with the tumor is about twice that of the normal tail, for each frequency (see fig. 10).
The relaxation behavior was investigated for several growing tumors. The time for tumor development in the tails was from 1 to 3 months. If the data were treated as if it were a single exponential, then a "TJ" would be obtained, whose value increases as the size of the tumor increases. This method of obtaining a "T," would correspond to the pro cedures illustrated in table 1 and might be a weighted average. It is better however to analyze the relaxation behavior into two, or perhaps three, relaxation times, as ex· plained in co njun ction with figures 3-5. This preferred procedure leads to a constant TI for the tumor as it grows.  for tim es > 0.3 sec es tabli shes th e s low relaxation at about 0.95 ± 0_10 s. The fast relaxation can th en be obtained by s ubtractin g out th e slow relaxation co mpone nt a nd re plottin g the diffe re nce. Th e res ult is a bout 0.20 ± 0.05 s. The data are not good e nough in thi s case to attempt to a nalyze th e fast co mpon e nt furth er into two tim es. Th e fast tim e th e n is either a chan ge in th e " normal " ti ssue near th e tumor or an average be twee n th e normal ti ss ue valu e of 0_35 to 0.40 s and a faster relaxation time associated with th e tumor. W e favor thi s latte r inte rpre tation, and a valu e of about 0.11 s at 23 MHz appears to be appropriate from da ta w e ha ve shown elsewhere [2] for th e "fast" co mpon e nt associated with the tumor. W e di sc uss below the experim e nts necessary to relate the relaxation times to th e appropriate cells. For example, can the slow relaxation compone nt of the tumor be associated with th e necrotic cells of fi gure 9, whil e the fast component corresponds to the viable cells, or is some other explanation correct?

Spin-Spin Relaxation
Spin-spin relaxation curves, in vivo, measured for normal and for tails with transplanted tumors are shown in figure 15. The data in figure 15a were obtained from the tail of a normal mouse. The tumor in the tail was monitored at regular intervals from the time that tumor growth was visible to the naked eye (67 days after transplanting) until the tumor was 2 cm in diameter (compared to the normal tail diameter of -0.5 c m). In contrast to the spin-lattice relaxation case, the s pin-s pin relaxation is see n to be more co mplex: a fit to th e normal mou se data of fi gure 15a requires a s uperposition of 3 or more expone ntials of the form of eq (4). W ell-d eveloped tumor relaxation re main s co mplicated although fewer term s are required for fittin g than for a normal tail. For example, in  The effect s of magnetic fi e ld gradie nt s we re minimi zed by usin g closel y spaced (2 ms) pulses . Solid curve in (a) is fit to data from normal tail. In (b)-(f) the ope n c irc les a re dat a taken fro m tail with melanom a a s it incre ased in size. The solid c urve represe nt s superimposed no rmal d ata fr om (a) for vis ual co mpari son. figure 15 (d, f) the initial decay of the well-developed tumor is less rapid than the corresponding part of the normal curve whereas the slope of the tumor curve at long times is more rapid than the same portion of the normal curve. Taking the lie point of the normal curve gives the number shown in table 2, but this sheds little light on the nature of the normal relaxation. The intermediate c urve , re presenting tumor growth at 70 days after trans plantation deviates from simple exponential behavior even more than the normal curve. Spin-spin relaxation actually again resembles the normal curve at about 74 days (not shown)_ Although the curves of figure 15 are not single exponentials, even by the normal tissue, table 2 was prepared to shown the loss of information accompanying conventionaille data. For early stages of the tumor the lie value was slightly less than the normal while for the latter stages the II e became greater than the normal.
The usual larger difference found between tumor and normal is clearly absent. This is not surprising in view of the complexity of the responses observed.

Further Work Needed on Mice
The studies on the tails of mice discussed above should be extended to transplantable sarcomas, lymphomas, and carcinomas. The correct assignment should be made of relaxation times of the appropriate tissue in the tails of live mice. This would involve measuring spin-lattice and spin-spin relaxation of protons in the tails of the live animals as the tumor develops. Then, at va,rious stages of development, animals would be sacrificed and the tumors should be taken for pathological and further NMR examination. All NMR measurements should be accompanied by pathological examination and perhaps by measurements of the water content. If it is then determined by following this procedure that the behavior in the live animal is, in fact, the sum of the behavior for representative tissues examined by NMR outisde the body, then the information available from experiments on tissues removed from the animal may be utilized to detect and monitor many tumors in the live animal, without detailed repetition of the in vitro standardization procedure for each case.
The Cloudman melanoma S91 contains both necrotic and viable cells which introduces the question as to whether the relaxation times, both T, and T2 , are different for these cells. In fact are the data given here for the tumor representative of one or the other or some average of the two? This introduces a new scope to NMR investigation to study both the necrotic and viable regions of the tumor separately and to compare the results between the two regions and with normal cells.

Probe Design
A surface type of NMR probe could be used for monitoring, in live animals and humans, tumors that are not accessible to the conventional geometry NMR probes, such as general body cavity tumors (including breast and liver tumors)_ A surface type of probe is not a standard item in NMR spec troscopy. A flat coil may serve as a probe for NMR de tection on a surface layer near the coil. The sensitivity would be very low because of the gross inhomogeneous distribution of radio frequenc y magnetic fields over the region of interest from a single planar coil. It should be possible, however, to design the winding so as to focus a homogeneous rf magnetic field into a relatively small region ( < 1/2 cubic centime ter) whic h would produce a considerable improveme nt in sensitivity. If suc h a s urface probe design is s uccessful , con sid era tion s hould then be given to designing a scannin g type of s pectrometer (wh er e the prob e is slowly moved over the volume of interest) to searc h for tumors perh a ps a s deep as an inch b elow the s urface of the s kin.
We feel that ultimately, th e use of NMR as a diagnostic tool de pe nds on th e s uccessful use of a s urface probe.

. NMR Zeugmatography
In the area of techniqu es of meas ure me nt , th ere is a new de velopm e nt involvin g NMR in a magneti c field gradi e nt whi c h offer s pro mi se in its ability to map our th e s pa ti al exte nt of diffe rin g nuclear e nvironm e nts in vari ous ma terials, in cluding biological tiss ues. This technique has been called NMR " diffraction " [24] or NMR ze ugm atogra phy [25,26]. Whe th er thi s tec hnique can s upple me nt the d e velopm e nt of a surface probe, as discussed above , requires · furth e r inves tigation.

Discussion
There ha ve bee n severa l expla nati ons give n to acco unt for th e diffe re nces in pro ton s pin-lattice a nd s pin-s pin relaxatio n be tween livin g ti ss ue (norm al a nd tumor) a nd in wa ter. W e have not fo cused on mec ha nis ms here because we we re more interested in utilizin g of the J'elaxation tim e difference be tween norm al and tumor tiss ue fo r possible in vivo di agnosis of cancer in hum an s.
An important q uestion re mains to be answered for both norm al and mali gnant tiss ue: Does ti ss ue s pin-la ttice a nd s pin-s pin relaxation be have the sa me in the live anim al as in th e biopsied tiss ue [27] ? .
It is importa nt to establish the corres pondence between relaxation in the component ti ssues a nd relaxation in the tissues actin g collectively in the body because thi s knowledge simplifies the searc h for subtle changes in the relaxation tha t may signal early signs of tumor growth. It may also be possible to monitor tumor beha vior after treatments s uc h as r adio-or che motherapy.
Our ultimate goal is to a pply NMR for th e diagnosis of primary and me tas ta ti c ca ncer in hum ans without the necessity of s urgical intervention. For example, NMR might be used to de termine if me tastatic tumor exists in the axillary lymph nodes upon di scovery of a cancer in the breas t. F or thi s reason so me precautions and thought s hould be give n to the possibility of a ny adverse effects tha t mi ght a ri se durin g conventional use of NMR on hum a ns of all ages_ The influ e nces of the D. C. a nd A.C. magnetic field s are of possible concern _ It has ne ve r bee n shown th a t constant D. C. magne ti c fi elds e ve n to high fie ld s trengths, greate r a nd of longer duration th a n that use d for NMR, are injurious. There are some reports [28] howe ver , of de tecta ble effects of D. C. magne ti c fields on biological organisms but e ve n these re ports are not ge nerally acce pted.
Of gr eate r co ncern ar e the possible side effects of the alte rn a tin g electromagne ti c fi elds. Th e harmful effects (e .g. heatin g) of rf fi eld s [28] a re greater at hi ghe r frequ e ncies, and thu s it is favorable that the frequ e ncies at whi c h th e NMR would be perform ed are definitely below the mi crowav e ra nge where so me da mage has bee n shown to occ ur. Anoth e r fa vorable aspect of the pulsed NMR is tha t the a verage power expende d is low. In critic al regions , s pec ial s hieldin g co uld be used to protect nearby parts from th e A.C. radiation.
We are grateful to Dr. D. Burk and Dr. M_ W. W oods of th e N ational Cancer In s titute for their participation in the in vivo s tudies re ported in thi s cha pte r , to C. M.
Mlad e n for hi s assis tan ce in the synth esis and analysis of d ata and to R. L P a rke for important tec hni cal aid with th e meas ure me nts.