What Do I Need to Know for Image Analysis

Is the Quality of the Study Technically Adequate?

Checklist:

Determining Study Quality

• Has the best method/modality been chosen in the given clinical context?

• Have the right body parts been imaged and completely so?

• Was the study performed properly or was imaging compromised owing to the patient's state or the situation during the examination?

• Was the study performed properly or was imaging compromised owing to the patient's state or the situation during the examination? This includes not only study technique (e.g., exposure voltage; the use of a scatter grid in projection radiography; choice of window and filter settings in CT or frequency and probe type in ultra-sonography; adequate selection of coil, sequence, and projection in MR imaging [MRI]), but also positioning (Did the patient stand upright?) and the degree of cooperation of the patient (Could the patient keep still [Fig. 4.3]? Did he hold his breath? Did he inhale deeply?). Having finished reading this book you should be able to answer the first two questions in 95 % of all cases. If the study is of satisfactory quality, you're on your way to the correct diagnosis.

How Do I Analyze an Image?

If you track the eye movements of an experienced radiologist, you will find that they appear to be rather unsystematic or even chaotic. Their perception of findings actually takes place in the sub-second range. The neophyte, however, has to adhere to a rigid sequence to accommodate the need for longer average observation time, which conflicts with the individual's momentary attention span. You will find suggestions for such sequences in the individual chapters.

Every experienced radiologist first eyes the "quality of the study" before starting a thorough image analysis. The objective of a first technical check is not to discard the study or ignore a finding (although this is also necessary in a few cases) but to establish a sound basis for the perception and decision-making process. This is a precious habit that you should also stick to. Obviously first and foremost one must assure that the images at hand indeed belong to the patient in question. To determine the quality of a study we ask the following questions:

• Was the right kind of diagnostic or interventional method selected considering the clinical indication at hand? The indication lists in this book will give you an orientation on what studies are done for what problem based on science and clinical experience.

• Was the right body region imaged and is it completely represented? This is a crucial question. As an example, check Figure 4.1a-c for confirmatory evidence. In addition, for example, a true second projection is an absolute must in skeletal radiography (Fig. 4.2).

Tissue Characteristics on Radiographic Images

Speaking in terms of characteristic density on plain radiography, the human body consists of different basic components, specifically fat, water, soft tissue, and bone. These four basic densities are recognizable when a structure contains an overwhelming amount of one particular kind of tissue. Obviously that is not always the case, and, to make matters worse, since conventional radiographs are projectional images we often look at a composite shadow made up of the density of several characteristic tissues. Some additional components such as calcifications or metal are introduced by disease or by external events. Body functions can also be observed with imaging modalities; for example, the flow in blood vessels or the cerebrospinal fluid spaces or the uptake of contrast media into a specific types of tissue.

Every imaging modality has its specific characteristics, strengths, and weaknesses with respect to the depiction of these components and functions. Some modalities

I Attention, This Is Your Wake-up Call!

Oblique Wake Calls

Fig. 4.1a Have a good look at the lateral cervical spine of this patient who fell off his bicycle. Can the stiff neck collar be taken off safely? •juajDj.yns jou si

M9jA |BJajB| SIL|}—3|qiSIA 9JB S9jpoq |Bjqa}jaA |BD|AJ9D aAjJ A|UO b The additional oblique view (two sturdy trauma surgeons

Fig. 4.1a Have a good look at the lateral cervical spine of this patient who fell off his bicycle. Can the stiff neck collar be taken off safely? •juajDj.yns jou si

M9jA |BJajB| SIL|}—3|qiSIA 9JB S9jpoq |Bjqa}jaA |BD|AJ9D aAjJ A|UO b The additional oblique view (two sturdy trauma surgeons pulled the shoulders caudally) shows the dislocation of C6 with respect to C7 (arrow). c The anterior-posterior radiograph proves the distortion in the C6/7 segment. Had the collar been taken off without further stabilization, a spinal cord contusion could have resulted.

I Diagnosis at Second Glance

Radiographic Distortion

Fig. 4.2 a The anterior-posterior radiograph shows no obvious abnormality.

b Only the lateral projection demonstrates the fibular fracture (arrow).

I No, Doctor

Fig. 4.3 After having been asked whether he could please hold still for just one more scan, this patient only shook his head. Not everything is possible.

I What Looks How with What Modality?

a Imaging of body components b Test yourself!

a Imaging of body components

b You see a part of a study of a patient who has suffered an injury. Try to c l assifythe study as precise l y as possib l e and then come up with the c l in-ica l diagnosis. Use Fig. 4.4a, a l itt l e anatomy, and your gray matter in the process.

A: air; O: oil; W: water; L: liver; M: muscle; C: calcium; M: metal b Test yourself!

b You see a part of a study of a patient who has suffered an injury. Try to c l assifythe study as precise l y as possib l e and then come up with the c l in-ica l diagnosis. Use Fig. 4.4a, a l itt l e anatomy, and your gray matter in the process.

A: air; O: oil; W: water; L: liver; M: muscle; C: calcium; M: metal

Fig. 4.4 a Here you see the most important body components as they are depicted by the different imaging modalities. The samples are surrounded by air. Gas, fluids, and tissues are contained in rubber glove fingers. For the ultrasound, the samples were dipped in freshly drawn tap water—the little bubbles are caused by gas in that water. The calcium tablet, of course, could not be dropped into water without dissolving immediately and producing gas, so my own (G.W.E.) radius had to take its place. By the way, which metal did we choose? Copper, lead, or iron? And why?

•(punj L|sn|s s.juaujjjBdap jno 'asjoM s,}Bi|M pue) luajsAs ai|} paBewep 3abl| p|noM pue aujipBiu y|/\| ai|} jo AjjubB ai|} oju| }i|Bu UMOU 3ABL| p|noM UOJJ :/\>| 00LJ3H3CI ipnuj pajBjjauad uaaq 3abl| p|noM jaddo^ 'pBa| s| }j asjnoD jo uB|s „ße|j ipjna,, ai|} sim ll63 °1 sa>l!l '3'M'D '3U!I sjnjdbjj ai|} jo ábm Áq mojjbuj auoq ai|} luojj jujoi ai|} ojuj pasBa|aj s| jbj ai|i 'ajnjDBjj psaq |B|q|} bá|a>||| jsouj 'aau>| ai^jo ajnpBjj Bu|A|oau| jujoi bjo djjsou -Bb|p s| ajnpjd a|oi|M ai|i •uojjBjnjBS jbj ai|} jaqiuaiuaj—ujBBB IBj s| |buB|smo| m!mjaáb| dojaqi 'poo|q i|}jM ||9moB p|noM}BL|:( |buB|s ajBjpaujjajuj ub sbl| j3Ab| 'juapuadap 'j3M0| 3i|i 'ssBeiui pajL|B|aM-2i uo pjnu jujoi snojas i|}jM a|qnBdujoD—iqBijq s| jaÁB| 3|pp|uj am 'saDBjjajui pin ij—pin o/w( i|}jM pajBpajddB ajB Bsjnq JB i i ajBdBjdns ai|} ujLpiAA sj3Ab| aajqi 'jelh ujb | dx3 p | noM uojjBjnjBS }B.(—>|jBp â| aAjjBJBdujoD 3jb snoauBjnDqns aqj pus auoq sno| | bdubd aqi 'bBeiui y|/\| ub s| }j saaojd }bl|}—>|db| q ajB suopuaj ai|} pus xajjoD Auoq ai|i 'a| qjssodujj snqj s| punos -Bjj i n UB—qai| i ai|} jo jnojuoD aja | doioD ai|} aas ubd no/, 'B | | ajBd áq} o} i BjUBJD jsnf aau>| aqj jo Apnjs | Buojpas | bjxb ub s| s|L|i

"Anatomical Noise"

Fig. 4.5 a Analyze the chest radiograph of this volunteer, who has an arrangement of wax spheres fixed to his back: large spheres are overlooked in the hilar region. b Compare the radiograph of the wax spheres alone: How many did you miss?

may have excellent spatial resolution (such as high-frequency ultrasound) but poor tissue depth penetration. Some have good spatial resolution (the ability to discern two small objects/points in space) but inferior soft tissue contrast resolution (the ability to discern two different types of soft tissue, such as gray and white matter of the brain). Such an example is the choice between CT of the brain versus MRI. In addition there are, in the individual modalities, special ways to enhance one aspect or another. For the beginner it is difficult—if not impossible—to get a comprehensive overview. Figure 4.4a is intended to help you a little. It shows the relevant components and how they are depicted by the pertinent imaging modalities. Figure 4.4b puts your new knowledge to the test. Please keep in mind that in projection radiography it is not only the density (Fig. 4.4) but also the thickness of an exposed object that determines the signal intensity (that is, the radiation attenuation in this case).

What Is a Normal, What Is a Pathological Finding?

The objective of imaging is the detection and localization of relevant disease or the exclusion of significant findings. What we consider to be pathological depends a great deal on the patient, on the societal context, and sometimes even on the immediate political or socioeconomic situation (a period of clinical depression following a catastro phe like 11 September 2001 in New York may be considered a normal reaction rather than disease). Calcifications in the wall of the abdominal aorta or a vertebral disk herniation without symptoms are not pathological in a 90-year-old unless their shape suggests a large aneurysm but would certainly lead to additional diagnostic work-up and possibly therapy in a young adult. Incidental findings like a closure defect of the vertebral arch of S1, a pulmonary azygos vein lobe, or a circumaortic renal vein are anatomical variants that have no relevance at all except under extraordinary circumstances: for instance, when surgery to the region is planned, e.g., laparoscopic resection of a left kidney from a healthy living donor. As has been mentioned in the previous chapter, the normal anatomy can be so confusing that detection, localization, and classification of findings can be extremely difficult. The vascular tree of the lung, for example, with its intertwined veins and pulmonary and bronchial arteries, is so complicated that large nodules can be completely concealed (Fig. 4.5). This interference of the normal anatomy with the detection of pathological findings is also called "anatomical noise" (analogous to the bothersome noise you hear in your Dad's old stereo system).

Where is the Pathology?

To assign a lesion to a certain location we need three dimensions, just like in stereoscopic viewing. In sectional

I Let Me Have Another Slice!?

Cranial Caudal Dimension

Fig. 4.6 The magnified view of a chest CTshows a number of little round structures (b). Only the review of the next cranial (a) and the next caudal slice (c) indicates that the central spot in b represents a nodule while the other structures are tubular and thus represent pulmonary vessels. If a thick slab reconstruction is used, things become easier: In d a single CT slice shows a few round suspect lesions. It is the axial (e) and the horizontal (f) thick slabs that differentiate clearly between the tubular vessels and the metastatic nodule (arrow).

f imaging it is the neighboring slices, reconstructions or so-called thick slab reconstructions, that convey the third dimension. Interactive review of thin CT images on an electronic viewing system such as a workstation or PACS system is becoming increasingly popular; scrolling through stacks of contiguous images allows the interpreting radiologist to form a 3D impression in his or her head. Looking at the third dimension and the neighborhood is the only way to differentiate between a sphere (such as a lung no dule) and a cylinder (such as a vessel section in a chest CT) (Fig. 4.6). In projection radiography, especially of the skeleton, it is the obligatory second oblique or perpendicular projection that gives us this information in conjunction with the initial radiograph (Fig. 4.7)—if we see the abnormality on the other projections, which is by no means always the case. However, a simple single projection performed with a sagittal x-ray beam can also give us hints as to the localization of a lesion. We use this information

I A Valid Second Point of View

I Radiologists Just Love Fat!

I A Valid Second Point of View

Humerus Xray Obilque View

Fig. 4.7 a What you see is a fracture of the humerus that has been stabilized with a number of rush pins. Looking at this projection, all looks well.

b The second projection brings a rude awakening: The rush pins lie outside of the proximal fracture fragment.

Fig. 4.7 a What you see is a fracture of the humerus that has been stabilized with a number of rush pins. Looking at this projection, all looks well.

b The second projection brings a rude awakening: The rush pins lie outside of the proximal fracture fragment.

I Radiologists Just Love Fat!

Fig. 4.8 The collimated view of an abdominal radiograph shows the oblique course of the iliopsoas muscle as it extends from the spine into the pelvis, made visible by its interface with the surrounding fat (arrow). The renal contour is also obvious. Sitting on top of the kidney one can, with a little imagination, appreciate the adrenal gland. The patient's breast is superimposed over it. All these retroperitoneal structures are perceptible owing to the their interface with surrounding fat, which is of different radiopacity. The dark, irregular areas represent intestinal air.

whenever a second projection is not present or does not clearly show the lesion in question. Radiographically perceptible interfaces exist between tissues whenever their density (or signal strength) in the given imaging modality is different enough to be detected by the imaging system used (contrast resolution, see above). This is of special importance in projection radiography. The kidney contour and the border of the iliopsoas muscle, for example, are appreciated well in abdominal films because these soft tissue density structures (of kidney or muscle) are surrounded by retroperitoneal fat of much lower density (Fig. 4.8). If such an interface is lost or its continuity is altered, a pathological process in this region must be suspected. In the kidney the phenomenon could point to a renal carcinoma that has broken through the renal capsule; in the case of the iliopsoas muscle, retroperitoneal fibrosis or a large psoas abscess could result in the loss of its contour. When taking a closer look at the analysis of chest radiographs we will make extensive use of this phenomenon (also called "silhouette sign"). Exposure geometry also influences the image appearance and can help us to assign a finding to a specific location. For the naked eye of an observer, far objects result in a small projection and close objects in a large projection on our retina. In projection radiography it is just the other way around: looking at a radiograph we actually see through the patient at a point source of radiation, the focus. Whatever is closer to the focus, i.e., farther away from the detector, projects larger on the detector. (It is like looking at your own shadow on a wall—the closer you get to the wall the smaller the shadow gets, and the closer you get to

Fig. 4.8 The collimated view of an abdominal radiograph shows the oblique course of the iliopsoas muscle as it extends from the spine into the pelvis, made visible by its interface with the surrounding fat (arrow). The renal contour is also obvious. Sitting on top of the kidney one can, with a little imagination, appreciate the adrenal gland. The patient's breast is superimposed over it. All these retroperitoneal structures are perceptible owing to the their interface with surrounding fat, which is of different radiopacity. The dark, irregular areas represent intestinal air.

I A Sharper Image, Please!

Detector

Object far from detector

I A Sharper Image, Please!

Detector

Object far from detector

Object close to detector

Fig. 4.9 The finite size of the focus causes unsharpness of the object margins that increases with growing distance to the detector.

Object close to detector

Fig. 4.9 The finite size of the focus causes unsharpness of the object margins that increases with growing distance to the detector.

■asuaD|| jo suonipuoD pue suua} 04 pafqns a6esn "paAJasaj s}i|6u nv aiuaiqi 90OZ © A60|0|pey ibdiuio in panels Bilbao 'ueuqseg

No, nobody has left. Two bears are sitting in full harmony in the middle of the tub facing each other (do not try this in your own bathtub)—their densities add up to a higher value than the bears imaged sideways. There is another fluid in the tub now. The bears displace this fluid, showing less density than the surrounding fluid. The fluid must thus have much higher radiation attenuation than whatever these bears are made of. It is iodine contrast medium.

¿qnj ai|} i|j|M jajjeiu aqj si l|}jb3 uojbl|m puy ¿paiejlsnjj pueAiajnjeujajd ya| luainjoauo sei| jo - qnj ai|} ui Bujjjjs aje sjeaq ||v 'pua ue oj Bujujod si Ajjed aqi j

There are two fluids in the tub now. Bear 2 must still be the one in the tub because its contour is still lost in the lower fluid. Its silhouette remains discernible in the upper fluid, though: The density of this fluid must be significantly lower than the gelatin's density. What wondrous fluid could that be? Of course, it is an oil bath for luxurious relaxation: The water remains at the bottom, the oil—or liquid fat—swims on top. Bear 1 still sits outside the tub in apathy. Bears 3 and 4 now both project laterally; bear 4 is now as bright as bear3 because they absorb the same amount of radiation.

¿qnj ai|} ui pauaddeq sei| }ei|/v\ 'uoijenjis aqj azA|bub pue peai|e 09 'A|mo|s sassajBojd BujuaAa aqi a

Bear 2 must be sitting in the tub since its contour or "silhouette" is lost—it displaces the fluid (water), which is of the same approximate density as gelatin. Bear 1 must sit outside of the tub since its shadow is added to the attenuation of the tub's water which it does not displace—its silhouette remains visible. Bears 3 and 4 have approached each other and discuss the further course of events. Both are close to the detector. Bear 3 stands at a 90° angle to bear 4. In this lateral projection it absorbs more radiation than bear 4 in the anterior-posterior position. The same is true for human beings: a lateral chest radiograph requires triple the dose of standard view! Remember this when you get a chest x-ray yourself or if you order one for a young patient.

•uojjjsod sj| paBueip sei| £ jeag ¿ip!i|M 'qnl 3L|1 u! s! JB3q auo :jsib| sajnujiu uai p

The tub is empty; which bear sits inside is impossible to say. Bears 3 and 4 are far away from each other; they look at each other and discuss the further course of events, but we cannot conclude this from the radiograph. ¿pejuod aAa aaeq Aaip oq ¿p pue £ sjeaq aje jaipo ipea oj asop moh ¿ui 3i|}eq }i saop }ei|/v\ ¿qnj aqj ui sjjs jeaq ip!i|M '3jns os jou ajb p pue £ sjeaq 'qjeq e aaeq oj jubm 2 pue l sjeaa 'ujooj -i|}eq aqj oju| pajjjaj 9abl| sjeaq AiuiunB aqi :aiueB aip jbjjv 3

Bears 2 and 4 stand close to the detector, bear 1 most distant from it. Balls 5 and 6 are of the same size and sit next to bears 1 and 2. Ball 7 is the smallest—its unsharp contour tells all: it was suspended with tape (see the faint shadow of it?) just a few centimeters from the focus.

¿jsaBjB| ai|} sj ||eq ipii|M ¿}sa||ews aqj si ||eq ip!i|M ¿1! wojj juejsjp L|D|L|M 'jopajap aqj oj asop puejs sjeaq i_|D|L|/\a "(¿—s) ||eq ,,pB3|„ Ae|d (t?-|.) sjeaq AiuiunB jnoj q 'adA( pue azjs awes aip jo aje sanBea||OD siq ||v 'BujjeoD jeBns e ipjM ui}B| -aB juaiBAinba-anssjj-yos jo apeiu jeaq AiuiunB v e Ol'p "Bjj aspjBxg noA 3>|B|/\| o) sjeag A inning

Contour Summation

uopdaDjad pue BmBEiui m Euaiuouaiid p

I Summation Effect

Fig. 4.11 You can see hyperlucent areas in both lung apices. On the patient's right side it is a real finding representing an abscess with thick walls that can be followed around the full perimeter of the lesion. The hyperlucent area on the left is bordered by the first and second ribs, the clavicle, and the spine. This is a typical summation effect— your brain tries to fool you.

Fig. 4.11 You can see hyperlucent areas in both lung apices. On the patient's right side it is a real finding representing an abscess with thick walls that can be followed around the full perimeter of the lesion. The hyperlucent area on the left is bordered by the first and second ribs, the clavicle, and the spine. This is a typical summation effect— your brain tries to fool you.

the light source the larger the shadow gets). We also have to consider that the focus of the x-rays has a physical size (in plain radiography between 0.1 and 2.0 mm), and that the size predicates the inherent unsharpness of the image. The larger the focus or the closer an object is to the focus, the less well defined are the apparent visible margins of structures on the corresponding radiograph (Fig. 4.9). You can go ahead and try out your new abilities by analyzing Figure 4.10b-f. Figure 4.10a shows a gummy bear and all the gummy bears shown have the same physical size.

to generate, present, view, and analyze radiological images (such as chest radiographs or mammograms) in a highly standardized fashion.

Sometimes during the analysis of a radiological image we discover lesions that are not real but are conceived by our visual system as such: parts of different anatomical structures are fused into an apparent "lesion," which is a "summation effect." This effect can make rib crossings

| "Satisfaction of Search" Effect

What Can Go Wrong in Perception?

The fact that a lesion exists in the body of a patient does not automatically mean that it is visible. And the fact that a lesion is visible does not mean that it is always perceived by the observer. The diagnostic process is not complete, nor is it worth a penny to the patient, until the diagnosis has been communicated in writing or verbally to the responsible clinician in a timely manner and can be acted upon. During the perception of a visible lesion, a number of interesting neurophysiological and cognitive effects come into play that one should be aware of. Like human faces, radiographic images have a "gestalt" that one can appreciate and remember after very little training. If we see the face of a good friend on the bus, we recognize them at once—we will perceive the new pimple on their face with a glance of the eye. A trained radiologist will also detect the nodule on a chest radiograph at once, in less than a second and thus without systematic search. The standardized way of generating and presenting the image supports this rapid detection. If a friend were lying on their back on the beach and you were passing behind them, you would find it much more difficult to recognize their face with certainty and even more so the pimple of course. It is the unusual orientation or "gestalt" of the otherwise well-known face that poses the problem. The same is true for other representations: Looking at fragments of an image data set (such as in multiple narrow windows of a digital image or by analyzing an image with the magnifying glass only) does not make the overall analysis of an image superfluous. Rather the combination of detailed observation and larger context of the entire image permit proper perception and interpretation. Experience and science make a strong case always

Brain Scan Interpretation

Fig. 4.12 a This CT scan of the head shows a dislocated ventricle drain that led to a hemorrhage in the brain's white matter. The responsible surgeon was present. The finding was discussed animatedly, communicated, and shown to other colleagues. b The additional infarction of the left cerebellar hemisphere (arrow) escaped attention in this setting and was diagnosed in a later review.

Fig. 4.12 a This CT scan of the head shows a dislocated ventricle drain that led to a hemorrhage in the brain's white matter. The responsible surgeon was present. The finding was discussed animatedly, communicated, and shown to other colleagues. b The additional infarction of the left cerebellar hemisphere (arrow) escaped attention in this setting and was diagnosed in a later review.

appear as "nodules." The lung apex is another area where the superimposition of first and second rib, the clavicle, and the spine can simulate a pseudolesion—a zone of "overinflation" in this case (Fig. 4.11). If we assume such a phenomenon, we have to trace the contours of the pseudolesion and dissect it into its individual components. But beware not to merely try to explain away all findings; do not automatically call all small apparent nodules in a smoker summation phenomena just because you know about this now—we do have other modalities to ensure that in fact there is no reason to worry. Comparison with previous radiographs, additional imaging, or follow-up examination after a few weeks may help work out what is going on.

Prior knowledge or findings can influence our subsequent analysis and distract us from relevant findings on the image. The discussion continues whether the patient history should be read before or after the initial analysis of the study. We suggest a preliminary analysis of the image without any knowledge about the patient's symptoms to get an unbiased first impression. Subsequently the patient history is read and the study is analyzed with this information in mind. The physician's focus on the patient's complaints is not always an advantage as it can divert from less straightforward findings and incidental diagnoses. The role of patient history and complaints with regard to the interpretation of findings is, of course, undisputed: relevant clinical points must be communicated to the radiologist or the diagnostic process goes awry. The final diagnosis then takes both perspectives into account and ideally consists of a list of differential diagnoses in the order of plausibility and probability. A particularly tricky phenomenon has already been mentioned in Chapter 3—the "satisfaction of search" effect (see p. 17). If the (often inexperienced) observer has managed to find a first relevant lesion, further interest in the image may deteriorate quickly: additional relevant information is overlooked or ignored (Fig. 4.12). So while it is great for the neophyte to make a sound first finding right away, the analysis of a study should be continued with great discipline and care.

0 0

Responses

  • Pasqualina
    What is a hyperlucent spot on the pelvic bone?
    4 years ago

Post a comment