From Brownian Motion to Virtual Biopsy: A Historical Perspective from 40 years of Diffusion MRI

From Brownian Motion to Virtual Biopsy: A Historical Perspective from 40 years of Diffusion MRI, by Denis Le Bihan

Denis Le Bihan recently published an open access review article in the Japanese Journal of Radiology titled “From Brownian Motion to Virtual Biopsy: A Historical Perspective from 40 years of Diffusion MRI” (https://doi.org/10.1007/s11604-024-01642-z). The article explores in depth several of the concepts that Russ Hobbie and I describe in Section 18.13 (Diffusion and Diffusion Tensor MRI) of Intermediate Physics for Medicine and Biology. The introduction begins (references removed)

Diffusion MRI was born in the mid-1980s. Since then, it has enjoyed incredible success over the past 40 years, both for research and in the clinical field. Clinical applications began in the brain, notably in the management of acute stroke patients. Diffusion MRI then became the standard for the study of cerebral white-matter diseases, through the diffusion tensor imaging (DTI) framework, revealing abnormalities in the integrity of white-matter fibers in neurologic disorders and, more recently, mental disorders. Over time, clinical applications of diffusion MRI have been extended, notably in oncology, to diagnose and monitor cancerous lesions in almost all organs of the body. Diffusion MRI has become a reference-imaging modality for prostate and breast cancer. Diffusion MRI began in my hands in 1984 (I was then a radiology resident and a PhD student in nuclear and particle physics) with my intuition that measuring the molecular diffusion of water would perhaps allow to characterize solid tumors due to the restriction of molecular motion and vascular lesions where in circulating blood “diffusion” would be somewhat enhanced. This idea was to become the cornerstone of diffusion MRI. This article retraces the early days and milestones of diffusion MRI which spawned over 40 years.

I knew Le Bihan when I worked at the intramural program of the National Institutes of Health in the late 1980s and early 1990s. To me, he was mainly Peter Basser’s French friend. Peter was my colleague who worked in the same section as I did (his office was the second office down the hall from mine), and was my best friend at NIH. Le Bihan describes the start of his collaboration with Basser this way:

During the “NIH Research Festival” of October 1990 I met Peter Basser who had a poster on ionic fluxes in tissues while I had a talk on our recent diffusion MRI results. Peter appropriately commented that the correct way to deal with anisotropic diffusion was to estimate the full diffusion tensor , not just the ADC [apparent diffusion constant], as the approach of the time provided. Basically, ADCs are not sufficient in the presence of diffusion anisotropy, except in particular cases where the main diffusion directions coincide with those of the diffusion MRI measurements. To solve this issue Peter and I came with a new paradigm, the Diffusion Tensor Imaging (DTI) framework. By applying simultaneous diffusion-sensitizing gradient pulses along the X, Y and Z axes the diffusion MRI signal would become a linear combination of the diffusion tensor components. From the diffusion MRI signals acquired along a set of non-colinear directions, encoding multiple combinations of diffusion tensor components weighted by the corresponding b values, it would be possible to retrieve the individual diffusion tensor components at each location.

In Le Bihan’s Figure 3, he includes a photo of Basser, Jim Mattiello, and himself doing an early diffusion tensor imaging experiment. Le Bihan was the diffusion MRI expert and Mattiello (who worked in the same section as Basser and I did at NIH, and who I’ve written about before) was skilled at writing MRI pulse sequences. When they started collaborating, Basser knew little about magnetic resonance imaging, but he understood linear algebra and its relationship to anisotropy, and realized that by making the “b vector” a matrix he could obtain important information (such as its eigenvalues and eigenvectors) that would determine the fiber direction. 

Denis Le Bihon (left), Peter Basser (center) and Jim Mattiello (seated), circa 1991.

Diffusion MRI works because spins that are excited by a radiofrequency pulse will then diffuse away from the tissue voxel being imaged, degrading the signal. The degradation is exponential and given by e^–bD, where D is the diffusion constant and b is the “b-factor” that depends on the magnetic field gradient used to extract the diffusion information and the timing of the gradient pulse. I had always thought that this notation went way back in the MRI literature, but according to Le Bihon’s article he named the “b-factor” after himself (“B”ihon)!

Le Bihon describes how the clinical importance of diffusion MRI was demonstrated in 1990 when it was found that stroke victims showed a big change in the diffusion signal while having little change in the traditional magnetic resonance image. In fact, Le Bihon claims that the other big advance in MRI of that era—the development of functional MRI based on the blood oxygenation level dependent (BOLD) imaging—has not yet led to any clinical applications, while diffusion imaging has several.

Le Bihon’s article concludes

Diffusion MRI, as its additions, DTI and IVIM [IntraVoxel Incoherent Motion] MRI, has become a pillar of modern medical imaging with broad applications in both clinical and research settings, providing insights into tissue integrity and structural abnormalities. It allows to detect early changes in tissues that may not be visible with other imaging modalities. Diffusion imaging first revolutionized the management of acute cerebral ischemia by allowing diagnosis at an acute stage when therapies can still work, saving the outcomes of many patients. Diffusion imaging is today extensively used not only in neurology but also in oncology throughout the body for detecting and classifying various kinds of cancers, as well as monitoring treatment response at an early stage. The second major impact of diffusion imaging concerns the wiring of the brain, allowing to obtain non-invasively images in 3 dimensions of the brain connections. DTI has opened up new avenues of clinical diagnosis and research to investigate brain diseases, revealing for the first time how defects in white-matter track integrity could be linked to mental illnesses.

If you want to learn more about diffusion MRI, I recommend Le Bihon’s article. It provides an excellent introduction to the subject, with a fascinating historical perspective.