 Okay so my last talk is on misdiagnosis and body CT and hopefully you all can see the slides. You know the thing we all worry about I know all the residents, fellows, faculty on the call basically is how do you do studies correctly? I think as radiologists our job is to get the right answer on every patient. Errors are always something we deal with when we try to minimize. I have given a talk on errors for many years and keep updating it and when I updated it most recently I kind of realized that also the amount of errors have increased sort of in this post-COVID era. I think everybody is facing limited staffing at the technologist and radiologist level. We modified lots of protocols for the couple years of COVID and I think people have kind of gotten a little bit laxadaisical perhaps you know for many years we didn't give oral contrast right because nobody wanted to take their mask off. Well you know now people often don't give oral contrast they need to and then also you know this is a great lecture series this morning and the whole weekend is a series of lectures but I think CME has really gone down a lot during this COVID era. I think general fatigue affects everybody and it's shown now that medical error at a minimum is the third leading cause of death in the United States behind heart disease and cancer and medical error really is underestimated because for example if you miss a lung nodule today that's a sonometer and the patient presents three years from now with a five sonometer mask with metastasis and the patient dies it's going to say lung cancer as the cause of death it's not going to say misdiagnosis when you could have cured the patient you didn't but then the patient presented when it was too late and so medical error may be number two particularly as the number of deaths from cancer and heart disease continues to decrease now that article also made the point that human error is inevitable none of us are ever going to be perfect the only perfect person is someone who doesn't read any films so you are going to make errors the question is what is it that we can do knowing how people make mistakes to really minimize or decrease the mistake now if you ask the question how frequently are errors in practice this is an article and this article's numbers have been have been shown to be correct by other articles the rate of interpretation error is between three and four percent across all studies however studies that contain abnormalities the error rate can reach 30 percent which is an amazingly high number now if you look at the majority of errors there were errors of under reading so the number one error and the error that's most critical is under reading because let's say you know Linda showed you a bunch of renal masses and let's say you say this mass maybe i'm not sure maybe it's benign rather than malignant well because you said there was a mass presence someone else is going to look at it and hopefully they'll reach the right diagnosis but if you read the if the report says kidneys are normal probably nobody is going to go back and relook at the scan so under reading is the most problematic you also will say what about speed we're all reading more scans than ever well there's no surprise that if you increase the rate of reading by 20 percent you increase the rate of error by three times people have also so there's no doubt about it that because we're reading more and reading faster error rates are going up they're not going down and there's a recent article published from ucsf that made the point that this is for neuro radiology this is in the er setting that made the point that errors were associated with higher volume shifts okay and that's it and and there they just showed very strictly pure error rates going up just based on volume and their conclusion was in order to decrease the error rate you need to decrease the volumes and the errors were clinically significant so their conclusion was instead of reading 60 cases a day they should read 40 however they can't get the work done so it's really reading 60 so now you know you're reading too much and your error rates are increasing now people have written articles that suggest reading twice have a second reader and perhaps having a resident and a fellow help you helps that a little bit but the reality is double reading doesn't work because you barely can get through the studies the first time let alone have two people read them you also know when you start looking at errors how frequent they are there's been a number of articles like this one published when you have outside films brought in and then reinterpreted where error rates as high as 40 percent or 41 percent have been reported and this was for head and neck imaging but this is the same numbers that are published in articles on basic body trauma so if you ask the question why don't we miss things on a ct often it's a poor search strategy you're looking for rula pancreatitis and there's a pe but you weren't looking for the pe there's poor understanding of pathology over cooling or under cooling the presence of bowel pathology making assumptions lindes showed you some nice examples of renal masses which looked when you looked at quickly they looked benign they were really malignant they were papillary rcc's well you're looking at an aortic dissection or a trauma patient you're reading quickly you're assuming that that's a benign lesion when it's not now other reasons we miss things on ct poor scan protocols lack of oral contrast delivery of ivy contrast or not giving ivy contrast or scanning in the wrong phases or just looking at axials and not mpr's some of the most common things i have seen there was an article published in jama surgery just a couple weeks ago this is a wonderful article you know i mentioned in the first slide that because of covet everyone's trying to push things no oral no ivy and you know in the er setting lots of articles were written from er docs why ivy's not necessary well this article was written by the surgeons and it's with radiology and it showed that unenhanced ct was 30 less accurate than contrast enhanced ct for evaluating abdominal pain in the er 30 percent okay that means you missed 30 percent of the findings because you did not give ivy contrast now you can argue ivy contrast potential risk but it's very low risk patients with renal function it's essentially has no risk a small percent of patients could have reaction but you have to give ivy contrast you can't assume a non-contrast study looks negative and is negative you're going to miss 30 percent of important findings okay so that becomes very critical protocol is everything if you do the study correctly you do dual phase when you need it you give ivy contrast you give over contrast that makes all the difference in the world now let's look at some other things okay full field of view do you need to have full field of view normally we do full field of view but in two areas cardiac ct and spine imaging we don't the reason we do it in the car the coronary area is because we want to look at the coronaries with the highest spatial resolution so you need to do those targeted images but you also and so you get a nice look at the right coronary but you also need to do the full field of view images because the same age group has lung disease and in this case lung cancer i remember when cardiac ct first came along and the cardiologist wanted to do it they said we shouldn't read the entire lung because they're now requesting it they're only requesting the coronaries well the answer is we know better you need to look at the entire study and so here we are so you can see this patient has lung cancer you need to reconstruct the full field of view and dictate that if you only do a coronary ct you have 70% of the lung but that indeed becomes important in the lumbar spine the same thing this article by lee made the point that a patient with back pain you do a dedicated spine study it's not uncommon to see extra spinal findings that are important most were benign but others required to work up including carcinoma lymphoma sarcoid triple a and 4% of patients had important findings and 80% of the time you would have missed these findings if you didn't do a full field of view image most neuro radiologists do only targeted views of the lumbar spine because they don't want to read the abdomen you need to look at the entire abdomen because you may see the findings that are causing the patient symptoms another practical thing do you look at the topogram or scout view on all your cases now in the old days we always looked at the topogram because when before packs you had film and the technologists would always shoot the topogram one without lines and one with lines in this case the patient was said to have retained barium they were running a post op fever when you look at the topogram you realize there's no barium but that's the ring and that's the ring left on a sponge okay so the patient had a retained sponge that was the reason for the patient's fever or this case again patient post op febrile this was just felt to be residual barium and some artifact but look at the topogram it's hard to believe that but they left that retractor in the patient's abdomen that's what was causing the pain here it just looks like barium maybe you begin to wonder it's too much but here it's very obvious you need to look at the topogram or this case which was an IV study for the section where's the IV you don't see anything on the study maybe see a little bit right here by the axilla when you look at the topogram you see what happened they extravasated right here and the contrast is going up in the subcut tissues along the inner arm and distally again very easy to see what's specifically going on now the question is the need to look at the topogram i showed you a couple examples where it was helpful but is that the rule there was an article written from hopkins two very good radiologists bill scott and bob gaylor who expert playing film radiographers read over 2000 topograms in a row and read them like they were reading a plane film and they found some finding of note in about a quarter of the cases in most of the cases those findings were seen on ct so no issue but in two percent they found findings that weren't seen on ct that may be because the topogram covered more area than the scan maybe it had the chest and the imaging was only the abdomen or it had some abdomen the imaging was only of the chest but in two percent of cases it found findings that were important and would have gone unrecognized without looking at the topogram berlin wrote an article that made the point that although two percent is small if you have 80 million 85 million ct's a year it means 1.7 million patients may have a finding that's critical but is missed now the reason he wrote that article was because in in his city and this was reported in a jr there was a case where the radiologist evaluated a skull ct in a in a child who fell read it as negative a few hours later the child had a seizure and died when you look back at the ct that there was no bleed seen even by experts but when you looked at the topogram there was an obvious skull fracture that every member of the jury could see but it was hard to see on the axials now when they asked the radiologist why he didn't look at the topogram the radiologist said we don't look at topograms okay that was about a three million dollar settlement in that case so berlin made the point that you need to look at the topograms even if you look at it just with a cursory look you need to look at the topograms and itch we made the point that perhaps on our structured reporting this should be aligned for topograms that i looked at the topogram and i saw this i saw that or i saw nothing so that becomes very very important now let's get down to a few specific things now remember this is um misdiagnosis i'm not talking about if i say bladder cancer i'm not talking about cases where the patient has hematuria or is staging bladder cancer what i want to know is how often is bladder cancer an incidental finding and missed bladder cancer is obviously a very common cancer in the u.s. and worldwide but my question is how often is it an incidental finding and if it's incidental are you missing it well how did this come about i looked at this legal case for the maryland society and this was a patient who came in the middle of night an older woman with acute abdomen and the radiologist on this non-contrast study read this as ischemic bowel and they were right the patient had ischemic bowel the patient went to surgery and survived okay the radiologist at that moment doing the study did not see that in the bladder there was a soft tissue mass your face forward three years later now you can see that the bladder cancer has more than doubled in size the patient presented with metastasis the patient died and they sued for misdiagnosis so the question to me was how often are we missing bladder cancer as an incidental finding now one of the things we recognize is bladder cancer particularly when small is very vascular as we scan more and more patients who are older think about all the orders we scanned we give patients 500 cc of water before the study so the bladder is extended what you have to remember is any enhancement in the bladder at all even five millimeters is going to be suspicious for bladder cancer so here's a patient with an aortic study and there's a five millimeter lesion at 12 o'clock in the bladder you can't blow that off that's a bladder cancer there it is on the sagittal view as well another case here it is by the uv junction a little over a sonometer and enhancing lesion again show nicely on the coronal views that's a bladder cancer or this case where there's a little bit larger lesion but also a bladder cancer we realize that we were missing bladder cancers and we just published a study which shows that bladder cancers are often missed unless the patient has hematuria so what you need to do you need to be very conscientious particularly in cases where you're doing a orders older patients good vascular enhancement bladders extended take a look at the bladder if there's anything enhancing in the bladder you have to worry about a cancer and recommend cystoscopy there's no magic but you will see you will pick up many bladder cancers that are commonly missed obviously if you had the late scans and the bladder was distended with contrast it's easy to see as well but again you saw the range of findings that were typically missed so you need to be very careful and we published this a couple years ago i haven't seen us miss a bladder cancer in a long time in hopkins because now everyone is very careful but unless you make an effort you're going to miss bladder cancers other things common sources of error pulmonary emboli now we all do pulmonary emboli and we're all pretty good at it i think i don't know i can't do a show of hands but a lot of people are using ai ai tends to even make us better but that's not what i'm talking about the rule out pe's i'm talking about the incidental pe's and what we found commonly when i was doing a lot of three d's of the pancreas when we only would scan the lower lung fields and the radiologist reading it first would look at the three by threes or five by fives it would say nothing about the lower lungs and i would pick up incidental pulmonary emboli incidental pe's are common in up to five percent of patients with cancer patients are asymptomatic or it's not felt to be symptomatic because they're tired they have all sorts of symptoms which everyone relates to their cancer you need to look very carefully here's just a good example of a pancreas of cancer there's a pe in the lower lung field it's interesting most of the pe's 90 percent are in the right lower lung so really look carefully at the right lower lung but it's very very important to look because you will pick up incidental pe's once you pick them up all the patients are treated with therapy now it's very important to remember that this is one of the areas which ai is playing a big role now and this has been a couple articles now charlie wight and maryland made the point that pe studies are often missed on abdominal ct for the same reasons i mentioned one of the articles just published this one and another one talking about using ai for pulmonary emboli it shows that one of the most common areas of pe detection is patients when you're not really looking for pe's where it's routine oncology patients or routine chest ct patients so it becomes very important our work provides more scientific ground for the concept of ai augmented radiologists instead of the the theory of radiologists replaced by ai but pe is one of the things you really want to look at this article in radiology just published a couple months back made the point high diagnostic accuracy and significantly shorten the time of diagnosis in patients in a setting with backlog of exams so if you're very busy let's say looking at oncology patients maybe a study done in the morning at nine o'clock isn't read till two o'clock but if you have the software running all the time at nine o five you will know there's a pe present and you will read that study quickly okay another problem gastric tumor detection i mentioned one of the challenges we have and i know you have is having the patients always get oral contrast and oral contrast can be just water but you need the stomach distention one of the issues is that the ear docs often say no oral contrast this is an article by mega bone abdominal imaging a number of years back making the point that cutting corners is happening at the expense of excellence in patient care you need to give oral contrast and perry pickard looking at his work in oncology patients showed that when you went from oral contrast to no oral contrast the number of misdiagnosis increased significantly so some examples this was a patient who was having weight loss and all sorts of symptoms but this was a chest CT if you look at the gastric fundus now in a chest CT we didn't give any water but i'm really worried about the gastric fundus is there's something going on here we brought the patient back we gave water look how well the stomach's distended the stomach is absolutely normal if you give water to every patient patients are always thirsty anyway the good news is you're going to eliminate the need for repeating this study for a day this patient thought he had gastric cancer till he realized he had nothing going on and look how nice the stomach looked when the stomach is distended you can pick up incidental findings like gastric polyps there you also can pick up out of no carcinoma patient with vague symptoms just a typical ear or patient there's an infiltrating tumor in the atrium which you would have missed if the patient wasn't given oral contrast or this patient what do you do here is this thickened is this normal well this patient was npo because they were going for e us but that same patient a few months ago has hundreds of benign polyps in the stomach here they're obvious here there's no way you could make any diagnosis linda spoke about the kidneys critical things phase of data acquisition image display format and rendering are all things that lead to errors we spoke before about non-contrast CT that's ideal for picking up renal calculi but in a patient with hematuria if you just say this study is normal you're going to miss a lot of cancers there's no perfect phase for the kidneys but multiple phases are commonly necessary i always worry in the ER setting when we say the stone study is negative i think the clinician here is the CT of the kidneys is negative but stone studies only tell you there's no stone but you're going to miss small tumors infection and vascular pathology this case was an ER doc who had hematuria told me the study was negative and it looks negative look at that right kidney looks fine but i convinced them to get IV contrast against his best wishes and look at that two and a half sonamin and renal cell carcinoma that's there and there on the delayed phase tumors that do not distort the outline will be missed on non-contrast studies again you need IV contrast another example here what do you do with this mass now one thing important here is i think linda mentioned and i showed you an example this little dots well that lesion is minus 71 that means there's an angiomyelopoma the entire lesion enhances but that little bright dot makes it a myelopoma but you can see the evaluation made it very simple to reach the right diagnosis also people at times particularly when the studies look normal may only focus on the axials you look quickly at the left kidney it looks pretty good but you know the cortical interface should be perfect it doesn't match over there but it's much more obvious on the coronal that the patient has a renal cell carcinoma coming off the kidney or this case i was checking a fellow who said this was normal upper pole of the kidney and maybe it is but in the coronal it's a carcinoma in the upper pole routinely looking at coronals and sagittals is critical with the kidney if you don't do that you're going to miss subtle tumors we also know there are other mimics of malignancy infection and inflammatory disease so be aware of that non-contrary sct this was a stone study some stranding around the kidney i don't see a stone but you need to give IV because then you see the patchy enhancement you see the patient's polynephritis so and you see it bilateral so again very very important i also will comment on misdiagnosis in the ureter the ureter we always think about a dilated ureter perhaps you need to see that in order to corporal theology but the reality is ureters can have small lesions within them which you can see right here you need to widen the window something like 550 over 50 rather than 400 over 10 and you can see there's something inside that ureter which is a one centimeter ureteral tumor but there's no hydronephrosis but that tumor can be resected and the patient will have no metastasis if you miss that the patient will come back with widespread metastasis i routinely look at mip imaging of the kidneys because you can see here that lesion shows very nicely on the mip imaging again there's no hydronephrosis but there is an obvious tumor in the patient's distal left ureter okay and we wrote an article about this it's worthwhile reading because you'll see lots of examples of where mistakes are made or in this case you look at the kidneys quickly they look okay but when you look at the mip imaging of the ureter normal right ureter abnormal left ureter look at the irregularity here that was a subtle infiltrating tumor again easy to miss there's no hydronephrosis you could miss that very easily on the axial CT but on the mip imaging it's much more obvious or in this case you follow the ureters down here's sort of a crescent i'm showing you there there's that circle or doughnut there's one ureter there's the other which means there's a filling effect here was a one centimeter tumor of the ureter again these are the things we commonly miss unless we're very careful we always think about the kidney with hematuria maybe we think about the bladder a little bit we don't really think about the ureter unless it's obstructed and that's a mistake i mentioned looking at different projections looking at coronals you need to look at the sagittal views routinely in the abdomen anything vascular is going to be missed in up to 20% of cases without sagittal views now obviously that's true with pancreatic cancer staging but i'm not talking about cancer if you don't look carefully at the mesenteric vessels of the sagittal views you're going to miss clot we also always look at clot proximal because that's where atherosclerotic disease is but with embolic phenomena the clot is typically more distal and so in this case you look and you say oh the ciliac looks good maybe a little narrowing sma looks good there's no plaque but as you follow it down the plaque is distal in the vessel okay there's the there's the thrombus in the patient's sma is very obvious when you do the mips when you do the reconstructions but it's so easy to miss unless you're very very careful you need to look at the entire ciliac and sma and ima everyone gets fooled particularly distally you don't you know you're kind of tracking the vessel down and you say it looks okay but you need to see the vessel in its entirety and the sagittal view is critical for doing that as well as many other things that are vascular related it's also important to look at sagittal views because of the bone the last thing i do when i look at imaging is looking at reconstructions in the sagittal with bone windows patients particularly older patients with back pain or abdominal pain it may be because of osteoporosis when you look at the axial views here you see a lot of plaque in the aorta dj d in the spine but when you look at the sagittal view you see the collapse of l1 one of the first of two ai projects that is funded that you get reimbursement on is for osteoporosis of the lumbar spine because of the fact that 85 percent of compression fractures are missed on routine abdominal ct because people don't routinely look at the sagittal views so you need to do that that's a very common source of error and a very important error i'll finish up with a couple comments there are a lot of things coming along to help us there are things that help us count ribs okay and look for rib fractures is basically curved coronal reconstructions but laid out nicely rib fractures are easy to miss but also to me they're often hard to count here you can see there's a first rib fracture nothing very tricky but here's the software which automatically counts the ribs for you counts the vertebral bodies lays out all of the ribs and they lays out all of the ribs which you can then rotate increases your accuracy by probably 30 percent for picking up rib fractures also it makes it much easier to count specifically what rib you're dealing with it make life very very easy we spoke a little bit about pancreatic cysts i'll just make a comment pancreatic cancers 40 percent of cancers two centimeters or less that are present and those lesions are typically the ones that are receptable or missed a dilated pancreatic duct with a abrupt cutoff means you have a tumor here the tumor is seen but is subtle but even if i don't see a tumor to me it's going to mean then there is a tumor present you need to evaluate further and again you could see it very nicely from the coronals and the cinematic as well the ability to look at transitions of duct become very very important we talked about neuroendocrine tumors 20 years ago we would say that we were 50 percent accurate with neuroendocrine tumors now in 95 plus percent and in fact we pick up many endocrine lesions that are not suspected which is a challenge one of the things now it's felt that an endocrine tumor under one cm surely should not be removed some people even say under two cm i mentioned protocols if you look at this case it looks like a nice venous scan no dilated pancreatic duct i don't see a tumor but there's a tumor there if i had the arterial phase there's the tumor again your accuracy on ct is going to be dependent on your protocols if you want to rule out a pancreatic mass it could be neuroendocrine you must do arterial and venous phase imaging look at that two plus sonometer mass you would have missed here it's obvious but it's not distorting anything it's not causing duct dilatation that's how easy it is to miss things again it's a challenge for us particularly in the er where you don't have all the phases you need and perhaps the texts are designing the protocols you need to be very careful here was just another example of a renal cell looks good distal pancreatic patient had a left nephrectomy here's the patient's pancreas but when you gave arterial phase there was a metastasis but look how easy it was to miss a three sonometer mass easy to see so protocols become everything i think you need to be certain that you have your protocols updated twice yearly i think it's important also that you train your technologists in doing the protocols i think case conference we had discussed protocols becomes very important now things people are doing i think ai is going to be big some people use checklists some people like checklists for reporting we have some checklists online if you're an apple user you can download for free from the apple store we've done checklists on pancreatic masses and adrenal masses as a way of walking you through the tumors and making better decisions so there's a lot more we can do ai i think is going to be very big articles now show ai is surely better than humans in picking up lung nodules it's more reproducible it's better on pe and things are going to can it's better on osteoporosis and it's going to keep getting better and better we've spent a bunch of time linda and i the past seven years working on early detection of pancreatic cancer and we believe that that's the way you're going to pick up incidental pancreatic cancers the computer is going to find the tumors and predict precisely what they are at an earlier stage now i think all of us make mistakes there's no doubt about it and no one's going to be perfect i think hopefully what this talk makes the point is by reviewing cases by reviewing your errors you're always going to make that first mistake perhaps but the key is not making it a second time and with that thank you for your attention