594
Points of View
SCIENCE IN RADIOLOGY*
explanation for a time. The useful, fruitful ideas are in fact just those that expose themselves to being tested, to being found out false. The more a good hypothesis excludes-the more it risks-the more testable it is, and the greater its worth and value. The same goes for X-ray reports. ,
THOMAS SHERWOOD
University of Cambridge Department of Radiology, Addenbrooke’s Hospital, Hills Road, Cambridge CB2 2QQ
MEDICAL students in the 1950s saw a stark contrast between the clinician of old, making brilliant diagnoses by means of an apparently chaotic method called "in my experience", and the new medical scientist, observing and measuring facts; at 1’ast, it seemed, medicine might become part of orderly, respectable science. But of these two it is my belief now that the first is in fact the scientist. Many of us still think that the process that distinguishes proper science, laid down by Francis Bacon, goes something like this: you first observe the facts, watch nature innocently at play, or you might do experiments ("contrived experience" in Francis Bacon’s phrase - i.e., you eavesdrop on nature). If repeated observations hang together, their coherence and consistency establish a law, describing how certain things happen,
always. FALLACY OF INDUCTION
The problem about the neat course of events linked with this great man’s name is that it involves induction, arguing for proof from the particular to the general, from observation to.conclusion, from information-gathering to the truth. If we classify facts, if we read the book of nature patiently, says induction, we are scientists, and will arrive at scientific truth. Unfortunately, it seems that in the first place there is no good logical rhyme or reason to the notion of induction, and also that it does not work very well. It is a nice idea, of course, that numbers must prove something true: after all, one patient is a case-report, two are a series. But all that numbers can do is tell us what is probable, and probability can lead us into terrible mistakes-like Bertrand Russell’s chicken. Day in, day out, throughout its life, the chicken got breakfast as the farmer arrived in the morning. Thus it had clearly discovered a highly probable natural law: farmer=food-until that is, the morning when the farmer very naturally arrived to wring its neck instead. As Popper has pointed out, 1000 facts will not tell you what is true, but one good fact will show you what is not true. The business of science, in this view, is to generate a testable idea, and then try to prove it false. This criterion of testability, of possible refutation, as the hallmark of science is really a delightfully sharp and honest tool. Scientific method in medicine, for example when looking at an X-ray, involves at once the generating of hypotheses, or making up stories, which are open to being knocked down by critical observation or experiment. If we are not wrong, our story might be a good *Based on Professor Sherwood’s inaugural lecture in the new chair of radiology, given in Cambridge on March 13.
BIAS OF THE INNER EYE
hope it is now clear why in my view it is the clinician trying to make a diagnosis, and not the data enthusiast collecting facts alone, who is practising scientific method. But the way we see things-perception, or the inner eye-also plays an important part in diagnosis, especially for the radiologist. We come with a preconceived idea, or schema, to all our observations; we have all learnt to see in schemas, and indeed they are very I
to our success and survival. One of our rockhard schemas is always to assume a firm spatial context for whatever we see; it is as if we would believe anything of the actors on stage, but never that the scenery might change. And it is just because in radiology X-rays are part of the framework of our own familiar everyday world that we use schemas of our own making in judg= ing radiographs. There is no observation, no data-collecting, without interpretation. We are biased observers, and can help ourselves greatly by recognising these mechanisms of bias of the inner eye, which are built into all our judgments. This ties in with my contention that the old ideas about scientific method---data first, interpretation later-are untenable. Science works by trial-and-error, not, it should be noted, by trial-and-success. We try our scientific ideas by observations and experiment to see if they are in error: we can never tell whether they are true, which is a futile pursuit. If they are not in error, they may do for the present.
necessary
TYRANNY OF INFORMATION-GATHERING
On both fronts, therefore, on the physiology of the ineye, and on the logic of science, we can never start off with a clean slate, along the lines of the model: look at the patient or the X-ray first, describe what you see, then interpret. To try to learn or teach radiology in this way might look orderly and tidy, but it is a sterile, doomed exercise. Knowledge means power not because it gives us measurable facts, but because it generates testable ideas. This needs saying, in radiology in particular, because current orthodoxy encourages a tyranny of information-gathering--embodied in that awful notion, the diagnostic work-up. "We can’t talk about this problem because we haven’t yet gotten ourselves a firm database", a professor of radiology said to me recently, and I think he was wrong. Examiners are wrong when they show a candidate for the F.R.c.R., as they tend to do without clinical comment, a hand X-ray with an expanded bony lesion, expecting by observation and inference the answer that this is a hxmophiliac who has knocked his thumb. And most grievously, we ourselves as teachers are wrong, because we teach, not what we do, but what we think we ought to do.2 For an example, let us take the discovery of a lump in the kidney. This is now most commonly a’ chance finding on an intravenous urogram done for some quite ner
595 very heavily weighted towards information-gathering. Second, we might be less ashamed of teaching
different reason, perhaps because of bladder outflow obstruction in a man. The mass may be an early car-
ent are
curable and well worth accurate It could be a simple, unimportreatment. and diagnosis tant renal cyst. So let us do a good, complete diagnostic work-up, you might say, with an ultrasound scan following the intravenous urogram, a renal arteriogram and venogram, a radioisotope scan, lymphogram, and computed tomography scan. We then assemble these data, and draw a conclusion as to the likely diagnosis. The whole process may strike you as rather unkind to thepatient, not to say extravagant of the doctor. It is true that the method has an air of scientific respectability, but I have tried to show that it is spurious. The alternative is to seek a critical diagnostic pathway through this ;ungle. We will then use ultrasound scanning, as the simple outpatient test for dividing patients into those with probable simple cysts and probable complex solid masses. The likely cyst can be needled at once as an outpatient procedure. The rather few patients with complex lesions will need inpatient arteriography. Because ultrasound scanning is not 100% accurate, mistakes will occur, and we must be ready for them. Occasionally a simple cyst is misdiagnosed on ultrasound, and the patient then has an unnecessary arteriogram. This is of course costly, and not entirely without hazard, but not a disaster. The false ultrasound diagnosis of cyst is more alarming, since a solid mass, a possible tumour, will then be needled. However, careful studies have shown that fine-needle puncture of renal tumours does not impair the patient’s chances by tumour spread. Most importantly, we know that the patient has a probable carcinoma, and needs
what
cinoma,
potentially
actually do, rather than what we think we ought to be doing.2 Perhaps best of all, research and teaching can be linked enterprises, so that, for instance, we
on a laboratory problem, or on a critical diagnostic pathway through a particular clinical difficulty, might make teaching a more active, stimulating procedureachieved more by example than by precept.
work
,
CASTING NETSs
Much of what I have said might be taken as being against classical Baconian science. But I dare to think that the appeal tc imagination and criticism, to inquiry and experim- ent, offered by an applied science in 1978, would have found favour in Bacon’s sharp eyes. The key quotation is: "Hypotheses are nets: only he who casts will catch". That is what science is about;I think we should take note of this thinking in medicine and radiology, and recognise that we would be much better off in our lives and work, even at a very mundane everyday level, if we behaved as scientists. That is, as trial-anderror men, boldly casting those nets.
REFERENCES 1. Popper, K. R. The Logic of Scientific Discovery. London, 1959. 2. Moran Campbell, E. J. Lancet. 1976,i, 134. 3. Sherwood, T. Br. med. J. 1975, IV, 682. 4. Abercrombie, M. L. J. The Anatomy of Judgement. Harmondsworth, 1969. 5. Medawar, P. B. Induction and Intuition in Scientific Thought. London,
1969.
3
urgent surgical attention.
A critical diagnostic pathway like this is therefore a trial-and-error exercise. It aims at simplicity, elegance, and economy. Its justification is scientific method and kindness to the patient. Indiscriminate informationgathering is not science. When we look at X-rays, it is absurd, and very expensive, to try to train ourselves or others into starting off with a blank mind, a clean canvas which we then paint with unbiased observations-the whole inductive apparatus. Our usefulness as doctors and radiologists depends on the likely stories we can tell about a particular patient. We will tell better stories the more our hypotheses are specific, exclusive, testable. That of course is why we go mad with irritation in X-ray departments if we don’t’get clinical information on the request form: our ability to generate tight useful stories around the film is much impoverished.
Objections can naturally be raised. I havepresented deliberate view which may be considered unbalanced, for I have had to move quite a little way along the seesaw to try to lever up that top-heavy lump of current orthodoxy sitting astride the other end (though most of my ideas are derived from the writings of others1,2,4,5). One obvious objection is that I am encouraging slapdash guesswork, instead of orderly learning. But the argument is concerned with an informed throw of ideas and their criticism. There is nothing facile about that; on the contrary, it is usually a tough exercise, demanding 2 lot of traditional background knowledge, and requiring a skill not easily taught or learned. I would like to suggest three initial lines of attack toward,a better balance of the see-saw. First, there are the textbooks which at presa
Occasional
Survey
USE OF CHRONIC CEREBELLAR STIMULATION FOR DISORDERS OF
DISINHIBITION I. S. COOPER
A. R. M. UPTON
Westchester County Medical Center, Valhalla, New York 11374, U.S.A.; and McMaster University Medical Center, Hamilton, Ontario, Canada L85 4J9
Center for Physiologic Neurosurgery,
Summary
A review of the clinical results- from 200
patients and the neurophysiological results from 42 patients suggest? that chronic cerebellar stimulation (C.C.S.) can Improve cerebral palsy and reduce intractable seizures. The therapeutic effects of stimulation of the cerebellar surface may not be due to activation of Purkinje cells. There is evidence that stimulation of brainstem structures, particularly the reticular formation, may be associated with thalamic inhibition; such effects would explain the clinical results of C.C.S. as well as the reduction in amplitude of reflexes, evoked potentials, and parexysmal discharges in the electroencephalogram. This hypothesis would explain the prolonged, rebound, paradoxical, and cumulative effects of C.C.S. No clinical disturbance or significant tissue damage has resulted from C.C.S. over 5 years. The technique is an example of the therapeutic manipulation
Points of View
SCIENCE IN RADIOLOGY*
explanation for a time. The useful, fruitful ideas are in fact just those that expose themselves to being tested, to being found out false. The more a good hypothesis excludes-the more it risks-the more testable it is, and the greater its worth and value. The same goes for X-ray reports. ,
THOMAS SHERWOOD
University of Cambridge Department of Radiology, Addenbrooke’s Hospital, Hills Road, Cambridge CB2 2QQ
MEDICAL students in the 1950s saw a stark contrast between the clinician of old, making brilliant diagnoses by means of an apparently chaotic method called "in my experience", and the new medical scientist, observing and measuring facts; at 1’ast, it seemed, medicine might become part of orderly, respectable science. But of these two it is my belief now that the first is in fact the scientist. Many of us still think that the process that distinguishes proper science, laid down by Francis Bacon, goes something like this: you first observe the facts, watch nature innocently at play, or you might do experiments ("contrived experience" in Francis Bacon’s phrase - i.e., you eavesdrop on nature). If repeated observations hang together, their coherence and consistency establish a law, describing how certain things happen,
always. FALLACY OF INDUCTION
The problem about the neat course of events linked with this great man’s name is that it involves induction, arguing for proof from the particular to the general, from observation to.conclusion, from information-gathering to the truth. If we classify facts, if we read the book of nature patiently, says induction, we are scientists, and will arrive at scientific truth. Unfortunately, it seems that in the first place there is no good logical rhyme or reason to the notion of induction, and also that it does not work very well. It is a nice idea, of course, that numbers must prove something true: after all, one patient is a case-report, two are a series. But all that numbers can do is tell us what is probable, and probability can lead us into terrible mistakes-like Bertrand Russell’s chicken. Day in, day out, throughout its life, the chicken got breakfast as the farmer arrived in the morning. Thus it had clearly discovered a highly probable natural law: farmer=food-until that is, the morning when the farmer very naturally arrived to wring its neck instead. As Popper has pointed out, 1000 facts will not tell you what is true, but one good fact will show you what is not true. The business of science, in this view, is to generate a testable idea, and then try to prove it false. This criterion of testability, of possible refutation, as the hallmark of science is really a delightfully sharp and honest tool. Scientific method in medicine, for example when looking at an X-ray, involves at once the generating of hypotheses, or making up stories, which are open to being knocked down by critical observation or experiment. If we are not wrong, our story might be a good *Based on Professor Sherwood’s inaugural lecture in the new chair of radiology, given in Cambridge on March 13.
BIAS OF THE INNER EYE
hope it is now clear why in my view it is the clinician trying to make a diagnosis, and not the data enthusiast collecting facts alone, who is practising scientific method. But the way we see things-perception, or the inner eye-also plays an important part in diagnosis, especially for the radiologist. We come with a preconceived idea, or schema, to all our observations; we have all learnt to see in schemas, and indeed they are very I
to our success and survival. One of our rockhard schemas is always to assume a firm spatial context for whatever we see; it is as if we would believe anything of the actors on stage, but never that the scenery might change. And it is just because in radiology X-rays are part of the framework of our own familiar everyday world that we use schemas of our own making in judg= ing radiographs. There is no observation, no data-collecting, without interpretation. We are biased observers, and can help ourselves greatly by recognising these mechanisms of bias of the inner eye, which are built into all our judgments. This ties in with my contention that the old ideas about scientific method---data first, interpretation later-are untenable. Science works by trial-and-error, not, it should be noted, by trial-and-success. We try our scientific ideas by observations and experiment to see if they are in error: we can never tell whether they are true, which is a futile pursuit. If they are not in error, they may do for the present.
necessary
TYRANNY OF INFORMATION-GATHERING
On both fronts, therefore, on the physiology of the ineye, and on the logic of science, we can never start off with a clean slate, along the lines of the model: look at the patient or the X-ray first, describe what you see, then interpret. To try to learn or teach radiology in this way might look orderly and tidy, but it is a sterile, doomed exercise. Knowledge means power not because it gives us measurable facts, but because it generates testable ideas. This needs saying, in radiology in particular, because current orthodoxy encourages a tyranny of information-gathering--embodied in that awful notion, the diagnostic work-up. "We can’t talk about this problem because we haven’t yet gotten ourselves a firm database", a professor of radiology said to me recently, and I think he was wrong. Examiners are wrong when they show a candidate for the F.R.c.R., as they tend to do without clinical comment, a hand X-ray with an expanded bony lesion, expecting by observation and inference the answer that this is a hxmophiliac who has knocked his thumb. And most grievously, we ourselves as teachers are wrong, because we teach, not what we do, but what we think we ought to do.2 For an example, let us take the discovery of a lump in the kidney. This is now most commonly a’ chance finding on an intravenous urogram done for some quite ner
595 very heavily weighted towards information-gathering. Second, we might be less ashamed of teaching
different reason, perhaps because of bladder outflow obstruction in a man. The mass may be an early car-
ent are
curable and well worth accurate It could be a simple, unimportreatment. and diagnosis tant renal cyst. So let us do a good, complete diagnostic work-up, you might say, with an ultrasound scan following the intravenous urogram, a renal arteriogram and venogram, a radioisotope scan, lymphogram, and computed tomography scan. We then assemble these data, and draw a conclusion as to the likely diagnosis. The whole process may strike you as rather unkind to thepatient, not to say extravagant of the doctor. It is true that the method has an air of scientific respectability, but I have tried to show that it is spurious. The alternative is to seek a critical diagnostic pathway through this ;ungle. We will then use ultrasound scanning, as the simple outpatient test for dividing patients into those with probable simple cysts and probable complex solid masses. The likely cyst can be needled at once as an outpatient procedure. The rather few patients with complex lesions will need inpatient arteriography. Because ultrasound scanning is not 100% accurate, mistakes will occur, and we must be ready for them. Occasionally a simple cyst is misdiagnosed on ultrasound, and the patient then has an unnecessary arteriogram. This is of course costly, and not entirely without hazard, but not a disaster. The false ultrasound diagnosis of cyst is more alarming, since a solid mass, a possible tumour, will then be needled. However, careful studies have shown that fine-needle puncture of renal tumours does not impair the patient’s chances by tumour spread. Most importantly, we know that the patient has a probable carcinoma, and needs
what
cinoma,
potentially
actually do, rather than what we think we ought to be doing.2 Perhaps best of all, research and teaching can be linked enterprises, so that, for instance, we
on a laboratory problem, or on a critical diagnostic pathway through a particular clinical difficulty, might make teaching a more active, stimulating procedureachieved more by example than by precept.
work
,
CASTING NETSs
Much of what I have said might be taken as being against classical Baconian science. But I dare to think that the appeal tc imagination and criticism, to inquiry and experim- ent, offered by an applied science in 1978, would have found favour in Bacon’s sharp eyes. The key quotation is: "Hypotheses are nets: only he who casts will catch". That is what science is about;I think we should take note of this thinking in medicine and radiology, and recognise that we would be much better off in our lives and work, even at a very mundane everyday level, if we behaved as scientists. That is, as trial-anderror men, boldly casting those nets.
REFERENCES 1. Popper, K. R. The Logic of Scientific Discovery. London, 1959. 2. Moran Campbell, E. J. Lancet. 1976,i, 134. 3. Sherwood, T. Br. med. J. 1975, IV, 682. 4. Abercrombie, M. L. J. The Anatomy of Judgement. Harmondsworth, 1969. 5. Medawar, P. B. Induction and Intuition in Scientific Thought. London,
1969.
3
urgent surgical attention.
A critical diagnostic pathway like this is therefore a trial-and-error exercise. It aims at simplicity, elegance, and economy. Its justification is scientific method and kindness to the patient. Indiscriminate informationgathering is not science. When we look at X-rays, it is absurd, and very expensive, to try to train ourselves or others into starting off with a blank mind, a clean canvas which we then paint with unbiased observations-the whole inductive apparatus. Our usefulness as doctors and radiologists depends on the likely stories we can tell about a particular patient. We will tell better stories the more our hypotheses are specific, exclusive, testable. That of course is why we go mad with irritation in X-ray departments if we don’t’get clinical information on the request form: our ability to generate tight useful stories around the film is much impoverished.
Objections can naturally be raised. I havepresented deliberate view which may be considered unbalanced, for I have had to move quite a little way along the seesaw to try to lever up that top-heavy lump of current orthodoxy sitting astride the other end (though most of my ideas are derived from the writings of others1,2,4,5). One obvious objection is that I am encouraging slapdash guesswork, instead of orderly learning. But the argument is concerned with an informed throw of ideas and their criticism. There is nothing facile about that; on the contrary, it is usually a tough exercise, demanding 2 lot of traditional background knowledge, and requiring a skill not easily taught or learned. I would like to suggest three initial lines of attack toward,a better balance of the see-saw. First, there are the textbooks which at presa
Occasional
Survey
USE OF CHRONIC CEREBELLAR STIMULATION FOR DISORDERS OF
DISINHIBITION I. S. COOPER
A. R. M. UPTON
Westchester County Medical Center, Valhalla, New York 11374, U.S.A.; and McMaster University Medical Center, Hamilton, Ontario, Canada L85 4J9
Center for Physiologic Neurosurgery,
Summary
A review of the clinical results- from 200
patients and the neurophysiological results from 42 patients suggest? that chronic cerebellar stimulation (C.C.S.) can Improve cerebral palsy and reduce intractable seizures. The therapeutic effects of stimulation of the cerebellar surface may not be due to activation of Purkinje cells. There is evidence that stimulation of brainstem structures, particularly the reticular formation, may be associated with thalamic inhibition; such effects would explain the clinical results of C.C.S. as well as the reduction in amplitude of reflexes, evoked potentials, and parexysmal discharges in the electroencephalogram. This hypothesis would explain the prolonged, rebound, paradoxical, and cumulative effects of C.C.S. No clinical disturbance or significant tissue damage has resulted from C.C.S. over 5 years. The technique is an example of the therapeutic manipulation
