fMRI in analytical vs non-analytical clinical reasoning

It is now widely accepted that human thinking involves two functionally and
probably anatomically separated systems.
System 1 processes are unconscious, rapid and automatic, while system 2
processes are conscious, slow and deliberative. It has been suggested that
System 1 partly consists of automated System 2 responses, but System 1 in
itself probably is able to learn implicitly.
System 2 thinking requires access to limited working memory systems, whereas
System 1 does not. There is some evidence that System 1 is related to (or
identical with) the X-system (refleXive) and System 2 with the C- system
(RefleCtive). The X-system is composed of the amygdala, basal ganglia and
lateral temporal cortex, while the C-system involves the anterior cingulate
cortex, prefrontal cortex and the medial temporal lobe. These systems have been
constructed based on activation studies using relatively simple decision
processes. As yet, no studies have focussed on higher order decision making
(i.e. clinical reasoning) using fMRI and at present it is unknown whether non
analytical and analytical models of clinical reasoning can be distinguished
using fMRI. The present study for the first time will address neural correlates
of the X- and C-system in the field of medical expertise. Experienced
neurologists will solve typical NAR- and AR cases in the scanner, allowing the
study of activation occurring in different parts of the brain related to
different thinking strategies.

Can clinical experience be related to activation differences using fMRI? On the
basis of previous studies by Schmidt cs. (Erasmus University Rotterdam and one
of the investigators in the present study) it has been demonstrated that
clinical reasoning involves 2 different reasoning processes, an "automatic',
fast, non-analytical form of reasoning which can deal with the majority of
patient problems, and a slower, analytical, deliberative way of reasoning that
has to be used when clinical problems are perceived to be difficult or unusual.
Then, and only then, explicit pathophysiological knowledge will have to be
accessed. There is some evidence concerning the situations in which analytical
reasoning has to be used instead of the default non-analytical form of clinical
reasoning. In some experiments case material and situations have been
manipulated in order to study the circumstances in which doctors trust on their
efficient non-analytical reasoning and in which situations even expert doctors
have to rely on effortful and time consuming analytical forms of reasoning.
There is circumstantial evidence (correctness of diagnoses, speed of reasoning
processes) and more direct forms of evidence (analysis of thinking aloud
protocols, recall studies) for the assumption that this dualism really exists,
but as yet direct proof is lacking. We hope to be able to show that functional
imaging using fMRI is a useful way to study this dualism from a totally
different perspective. We want to study whether non-analytical reasoning (the
method of the expert) results in a different activation pattern than analytical
reasoning (the method of the novice and the method of the expert for difficult
or ambiguous clinical presentations). We also want to study whether these
higher order thinking processes can be related to X- and C-systems as described
by others.

In this study the subjects (20 experienced neurologists) are their own
controls. Following a validation study, which at present is in the process of
completion, subjects will solve 20 cases, of which 10 will (as has been
demonstrated in the validation study) stimulate non-analytical reasoning and 10
that will lead to analytical reasoning. The cases consist of 10 sentences,
65-80 words in total. The 10 sentences will be projected by a beamer on a
screen in 2 parts of 5 sentences. Every case has 2 variants, a NAR- and an
AR-version. The AR-version has been constructed by inserting ambiguous
information (1-3 noise items), each time transforming a straightforward case
into a difficult one. The third condition consists of a neutral, non-medical
stimulus, in order to neutralise the effect of reading. After the scanner
session, a semistructured interview will be conducted in which questions as
confidence in the diagnoses that were made, the effort with which the solution
came and the perceived degree of complexity of each case will be addressed. The
number of 20 subjects is in accordance with accepted standards in fMRI studies.
In the validation study, NAR cases were solved in an average of 48 seconds
(including typing of the final diagnosis), while the mean reasoning time for
analytical cases was 70 seconds.

Participants will be exposed to a 3 T magnetic field. No side effects have been
described so far. On rare occasions a peripheral nerve (abdomen) is stimulated
by the changing magnetic gradients. This will cause an itchy feeling, but is
not harmful.
It is a single session of two hours (of which max. 1 hour research in the
MRI-scanner, preceded by an instruction and followed by a review of max. 30
minutes).
It has been found in our earlier studies that experienced neurologists are
usually very interested in their own thought processes.