TwinTree Insert

04-07 Measurements in Medical Diagnosis

ifteen years after the first description of different relaxation behavior in tis­sues by Erik Odeblad [⇒ Odeblad 1955], other researchers started pos­tu­lat­ing that relaxation ti­mes differentiate tumors from normal tissue since most T1 (and in a similar way T2) values of pathologic tissue can differ markedly from the T1 of the similar nor­mal tissue [⇒ Damadian 1971] (cf. Chapter 20: History of MRI).

However, the ability to discriminate, type, or even grade tumors using re­la­xa­tion time values has remained a dream, despite the sophisticated multi-point fits in­­tro­­du­c­ed over the years. Figure 04-24 shows that there are differences between, in this case, T2 of normal and diseased tissues [⇒ Rinck 1985]. Although values of T2 are more ac­cu­ra­te than those of T1 because more points are used for their calculation, these dif­fe­ren­ces are not significant between T2 values of, for in­stan­ce, tumors and edema or infarction.

Figure 04-24:
T2 values of normal and pathological hu­man brain tissues measured at 0.15 Tesla, based upon 24 echoes. The standard deviation (SD) is given in yellow. The SD of nor­mal tissues can reach 20%, that of pathological tissues 30%.

Every year, the literature announces new attempts to exploit relaxation-time mea­sure­ments in vivo. There are some positive reports about its successful use. Many concern follow-up of therapy, with patients being their own reference.

Pub­li­ca­tions include, for instance, the report that relaxation times from leukemic bone marrow can be used for the differential diagnosis of this disease (Figure 04-25) [⇒ Jensen 1990]. Similar results in high-grade gliomas have been published by another research group [⇒ Boesiger 1990].

Figure 04-25:
T1 measurements. Follow-up of treatment of acute myeloblastic leukemia. Responder: green; non-re­sponder: red.

Yet, the follow-up of treatment based upon relaxation-time values is difficult and in most instances dubious (Figure 04-26). A rise and subsequent decline of relaxation time values after a local intervention might rather indicate edema and in­flam­ma­tion than successful treatment [⇒ Zhang 2014].

Figure 04-26:
Relaxation-time measurements of iden­ti­cal samples under identical measurement conditions can reveal great standard de­via­tions, as shown in this example.
Re­ly­ing on in vivo measurements to evaluate the out­come of treatment is dubious. Only in some instances do massive changes allow a positive assessment.

Several other studies dealt with pixel-by-pixel mapping of relaxation times of nor­mal appearing white brain matter in multiple sclerosis (MS) patients. The results sug­gest­ed minute invisible changes in the white matter which might ex­plain brain func­tion deficits that cannot be explained by the size and location of visible MS pla­ques [⇒ Barbosa 1994, ⇒ Lacomis 1986, ⇒ Rinck 1987].

However, also these measurements are not clinically applicable.

Availability of databases of in vivo relaxation-time measurements is very limited. A large collection of data was published by Bottomley et al. [⇒ Bottomley 1984, 1987].

A comparison between in vivo and in vitro relaxation measurements is quite dif­fi­cult because many T1 re­la­xa­tion time values change rapidly after excision. Only brain tissues reveal a re­la­ti­ve­ly stable relaxation behavior after they have been re­mov­ed from the body [⇒ Fischer 1989, 1990].

After absolute T1 and T2 values had been used unsuccessfully by researchers, com­­bi­­na­­tions of T1 and T2, histogram techniques, and sophisticated three di­men­sio­nal display techniques of factor representations were used (‘fingerprinting’, bio­mar­kers) [⇒ Skalej 1985].

04-07-01 Rapid Relaxation Constant Estimation Techniques

Precise measure­ments require long acquisition times; the repetition time, TR, should be equal to or greater than 5×T1. At 0.15 T, the T1 of my­ocardium is around 380 ms, at 1.5 T it has climbed to around 1000 ms. Measure­ments at low fields take ap­pro­xi­ma­te­ly 5 minutes, at high or ultrahigh field more than 10, perhaps 15 mi­nu­tes. Thus, makeshift faster acquisition methods were sought and devel­oped.

Fast acquisition of quantitative T1 maps can, e.g., be based on a series of snap­­shot fast low-angle shot (FLASH) images after inversion of the magnetization [⇒ Deich­mann 1999].

Such techniques were for instance used for estimat­ing the con­cen­tra­tion of pa­ra­mag­ne­tic con­­trast agents in an organ.

Since the acquisition of quantitative tis­sue data from a beating heart has to be very fast, lately much research is focused on modifications of a pulsed NMR se­­quen­­ce proposed by David C. Look and Donald R. Locker in 1969. MRI did not exist at that time, and Look and Locker used their time-saving one-shot method for NMR spec­­tro­scopy instead of the conventional meth­ods to measure the T1 re­la­xa­tion time. The spectroscopic “LL” method was within 10% of the conventionally pre­cise­ly cal­cu­­la­ted value [⇒ Look 1969].

In the 1980s, the method was further de­veloped for MRI by Graumann and his col­­lea­gues [⇒ Graumann 1987].

Others followed and precision was waived for speed. Among the modified se­quen­ces for car­diac and, e.g., brain MRI experimentally (and some­times cli­ni­cal­ly) used today, one finds PURR [⇒ Lee 2000], MOLLI [⇒ Messroghli 2004], and ShMOLLI [⇒ Piechnik 2010]. They all suf­fer to varying extent from er­rors, re­sulting in an un­der­esti­ma­tion of the true T1. ‘Ap­parent’ T1 values of MOLLI and ShMOLLI mea­su­re­ments of, e.g., nor­­mal myocardium have an error range of 30% or higher and always are shorter than true T1 values.

A number of different pulse se­quences, e.g., SASHA, SAPPHIRE, DESPOT and many others were also introduced and tested. Unfortunately, none of these values are re­li­able or reproducible (Figure 04-27) [⇒ Bojorquez 2017]. From sci­entific and me­di­cal points of view, these mea­surements are un­sound because the margin of error is huge.

Figure 04-27:
T1 values at 3 Tesla of human gray brain matter (25 different collectives) and myocardium (9 different col­lec­ti­ves) measured in vivo on different MR machines with accelerated data acquisition algorithms (com­pil­ed from different sources). The estimated values are imprecise and spread across several hundred milliseconds.

The sharp fall of T2* values at high and ultrahigh fields is related to the drastic rise of magnetic sus­ceptibility effects which grow linearly with magnetic field strength; pure T2 is not af­fected in the same way. Imperfect spoiling of transverse mag­ne­­ti­za­tion at higher flip angles in gradient echo sequences has a negative effect on a pre­cise estimation of signal intensity and other parameters, such as relaxation times [⇒ Zur 1991].

spaceholder redCritical Remarks. Yet, in the end, it is not even the most elaborate data ac­qui­si­tion that makes typing of normal and pathological tissues (‘fingerprinting’) or grad­ing of diseases impossible but rather the complexity of tissue composition and the overlapping of relaxation time values of heterogeneous volume elements exa­min­ed and processed into a single number or number range.

It is helpful to once look into a microscope and to see how complex and com­pli­cate tis­sue struc­tu­res are, both in normal and in pathological tissues — and in ab­nor­mal, but not (yet) patho­lo­gi­cal tissues.

Even following a trend of changing T1 and T2 acquired with the same equipment and the same imaging parameters to calculate the relaxation constant values during and after the treatment of a patient can be like 'fishing in troubled waters.'

From a scientific point of view, MR imag­ing is a crude and not very exact tech­no­l­o­gy. However, to be imprecise in medicine does not preclude specific use.

One example is the measurement of cardiac iron overload, which according to a num­ber of cardiological publications is, so far, one of the most useful diagnostic ap­­proa­­ches in patients with thalassemia major; however, many papers about the topic are dubious and the methods used lack scientific background. Myocardial damage in tha­las­se­mia is induced by iron deposition: free unbound iron catalyzes the for­ma­tion of cell-toxic hydroxyl radicals. Thus, monitoring of myocardial iron content would be useful and could be done by estimating T2 or T2*.

To be able to discriminate ‘normal’ myocardium and pathological tissue alteration the approach requires massive tissue changes, and it cannot distinguish between fi­bro­sis, in­flam­ma­tion, and infiltrative cardiomyopathies, myocardial edema and pos­sib­le other tissue changes. However, the pathological T2* values seem to be highly reproducible on different MR equipment [⇒ Auger 2016].

Another area of application of relaxation times measurements might be the fol­low-up of massive T1 changes after the injection of a targeted contrast agent, such as Mn-DPDP and the comparison of plain and contrast-enhanced tissue, e.g., in heart diseases (cf. cardiac applications of manganese).

Here, too, imprecise measurements might be of diagnostic value.

04-07-02 Biomarkers

To add confusion to complicated science, chan­ges of terms and ter­mi­no­lo­gy are com­­mon in contemporary bioscience.

Thirty years after the description of relaxat­ion times and re­la­xa­tion ra­tes as pos­sib­le or, rather, questionable biological indicators they were re-rank­ed among bio­­lo­­gi­­cal mark­ers or biomarkers.

In general, biomarkers are bio­logical in­dicators of any kind; there are thousands of them. They are not spe­cific for MR imaging or MR spectroscopy. Typical bio­mar­­kers are mea­sure­ments or scores such as blood pressure, body tempera­ture, the body mass index, or clinical signs such as external manifestations of disease. Many of them are helpful, others of limited and questionable value — still widely used.

In MR imaging, biomarkers break down into numerous subgroups where they can be applied standing alone or several combined, relaxation times being only one of them.

Aside of T1 and T2, there are other possible indicators for the detection, dia­gno­sis, and monitoring of treatment, i.e., of particular physiological or disease states.

Quantifica­tion of MR parameters is also dis­cussed in Chapter 15. Biomarkers ex­­tract­­ed through image segmentation and multispec­tral analy­sis and the basics of artificial intelligence are also de­scrib­ed in Chapter 15, those acquired with the help of contrast agents in Chapter 13, and by dy­na­mic imag­­ing in Chapter 16.