Diagnostic errors when interpreting abdominopelvic computed tomography: a pictorial review

Abstract

Diagnostic errors remain an inevitable occurrence during abdominopelvic CT (APCT) interpretation, despite advances in imaging technology. The main cause of error is failure to identify a lesion (i.e. perceptual error) and failure to recognize a finding's significance (i.e. interpretive or cognitive error). Awareness and understanding of the causes of errors can reduce their occurrence and may lead to a reduction in morbidity and mortality. This pictorial essay highlights various causes of error in interpreting APCT scans and briefly discusses possible solutions for minimizing these errors.

INTRODUCTION

As the importance of abdominopelvic CT (APCT) scans increases, radiologists are faced with the challenge of ensuring comprehensive and accurate imaging interpretations.¹ It has been shown that multiplanar reformatted (MPR) images in multidetector row CT can improve the diagnostic performance, resulting in a more accurate image interpretation compared with axial images alone.¹ However, our experience is that abdominal radiologists are still predisposed to diagnostic errors in the interpretation of routine APCT imaging, despite these advances in imaging technology.

Among the various diagnostic errors in radiology, the most serious problems are perceptual error and interpretive error.² These diagnostic errors can cause delays in appropriate treatment and lengthen hospital stays.² In daily practice, approximately 7.6% of abdominal CT interpretations have contained diagnostic errors.³

Although it is well known that knowledge of diagnostic errors can be meaningful for improving diagnostic performance,³ no previous systematic or illustrative reviews have focused specifically on errors related to APCT scan interpretation. Thus, the purpose of this pictorial essay was to illustrate the various causes of perceptual and interpretive errors during interpretation of APCT and to briefly discuss possible solutions for minimizing these errors.

PERCEPTUAL ERROR

Perceptual error is defined as a failure to recognize an abnormality during initial image interpretation and a subsequent recognition of this abnormality when the image is retrospectively reviewed (false-negative error).^2,4 This type of error can have damaging consequences and is associated with many psychophysiologic factors, including poor conspicuity of the lesion, satisfaction of search, rapid interpretation, reader fatigue and distractions.^3,4 The causes of perceptual errors can be categorized by our own modification of the Kim–Mansfield radiologic error classification system⁴ as follows: (1) overlooking other lesions owing to a conspicuous lesion, (2) missing a small lesion owing to another nearby lesion (superimposed lesion), (3) lesion oversight owing to location in a commonly overlooked anatomic site (blind spot) and (4) failure to detect lesions as a result of improper use of the CT window level.

Overlooking other lesions owing to a conspicuous lesion

The satisfaction-of-search error can contribute to lesions being overlooked.⁵ This error may be a result of the radiologist tendency to stop a search once an abnormality is identified.² Small or large bowel injuries are often missed on multidetector row CT examinations, particularly in patients with multiple injuries after trauma (Figure 1). In addition, synchronous primary cancers or multiple primary cancers are rare, having been reported in only 1–3% of malignancies.⁶ It can be easy to miss cancer of another organ in the absence of an organized approach (Figure 2). In cases of bowel perforation, coexisting malignancy could be overlooked because of a surrounding abscess or severe pericolic inflammation (Figure 3). It is also possible that pre-existing lesions will lead to unsatisfactory search for new lesions when evaluating imaging in such patients.³ Therefore, it is essential that observers remain vigilant, even after diagnosing several lesions, until an entire analysis has been thoroughly performed.

Figure 1.

Missed transsection of the descending colon due to a conspicuous jejunal perforation in a 56-year-old male with trauma: (a) the axial CT image shows focal jejunal wall disruption (arrowhead) with adjacent air bubbles due to jejunal perforation. At the time of surgery, transsection of the descending colon was found, which was missed on initial CT interpretation. (b) On careful retrospective evaluation, an abrupt cut-off of the descending colon (arrows) was seen in the oblique sagittal reformatted image.

Figure 2.

A missed gastric cancer due to conspicuous pancreatic cancer in a 69-year-old female with anorexia: (a) the axial CT image showing an ill-defined, low-attenuation mass (arrows) on the pancreatic tail due to pancreatic cancer. (b) However, wall thickening of the stomach angle (open arrows) due to gastric cancer was unrecognized during the initial CT.

Figure 3.

A missed rectal cancer due to a conspicuous ileal perforation in a 60-year-old female with 1 week of periumbilical pain: (a, b) axial CT images show a defect (open arrow) in the posterior ileal wall with an abscess (asterisk) due to ileal perforation. (c) The coronal reformatted image shows irregular wall thickening of the proximal rectum (arrows) due to rectal cancer, which was missed on the initial CT interpretation.

Missing a small lesion owing to another adjacent lesion (superimposed lesion)

This error occurs when a small lesion is obscured by another adjacent lesion, resulting in non-detection. For example, the diagnosis of caecal cancer can be difficult in patients with coexisting acute appendicitis (Figure 4). The incidence of obstructing caecal cancer as a cause of acute appendicitis is only 0.8% in elderly patients; however, obstructing caecal cancer could mimic simple acute appendicitis, which can cause delays in appropriate treatment.⁷ To minimize this error, radiologists employ a high index of suspicion for subtle abnormal findings; to be successful, MPR images should be reviewed meticulously.¹ In addition, reviewing thinner slices is helpful when a questionable abnormality arises.

Figure 4.

A missed caecal cancer due to superimposed perforated appendicitis in a 61-year-old male with right lower quadrant pain: (a) the initial axial CT image shows a distended appendix with appendiceal wall thickening (arrowhead) and a periappendiceal abscess (asterisk) due to perforated appendicitis. Thickening of the caecum (arrows) was initially thought to be a reactive change. (b) On axial CT imaging 1 week later, caecal wall thickening (arrows) remained, although the abscess resolved with percutaneous drainage. The drainage catheter (open arrowhead) can be noted. On colonoscopic biopsies, caecal cancer was confirmed.

Oversight of a lesion located in a commonly overlooked anatomic site (blind spot)

Blind spots refer to anatomic sites within the body that are commonly overlooked.³ On abdominal CT, areas such as the bone including the paravertebral soft tissue (Figure 5), lymph nodes, vessels, skin, and subcutaneous fat (Figure 6) can potentially become blind spots.³ Failure to identify abnormalities in these blind spots can be caused by cognitive factors and systemic factors.² Awareness of these problematic spots and careful evaluation in these areas are important.

Figure 5.

A missed metastatic nodule located in a blind spot in a 59-year-old male with a history of rectal cancer: (a) the axial CT image shows an ovoid right paravertebral nodule (arrow) that was missed on initial interpretation. (b) The axial CT image obtained at 1-month follow-up: the nodule (arrow) had enlarged, suggesting a metastatic process. The metastatic rectal cancer in the right lobe of the liver can be noted.

Figure 6.

A missed metastatic nodule located in a blind spot in a 46-year-old male with a history of lung cancer: (a) the axial CT image shows a tiny round nodule (arrow) in the subcutaneous fat of the left posterior abdomen. The nodule was unrecognized initially. The metastatic lymph nodes in the para-aortic area and the mass in the left perirenal space can be noted. (b) The axial CT image obtained at the 1-month follow-up shows that the missed subcutaneous nodule (arrow) had increased in size. The nodule was thought to be lung cancer metastasis.

Failure to detect lesions owing to improper use of the CT window settings

Approximately 11% of cases had a significant diagnostic finding that substantially affected the final diagnosis that was not detected in the soft tissue window but was visible on review using the liver and bone window settings.⁸ Thus, appropriate use of the CT window settings (Hounsfield unit) is key to diagnostic efficacy.³ In general, a narrow window width is useful for evaluation of soft tissue structures, while a wide window width is more accurate for bones.⁹ Increasing the contrast by using a narrow window setting will facilitate the detection of hepatic lesion and renal cell carcinoma (Figure 7). To ensure proper diagnoses, the radiologist should evaluate the APCT images not only with the soft tissue window settings (width/level: 350/50 HU), but also with liver (narrow, width/level: 150/100 HU) and bone (wide, width/level: 3000/500 HU) window settings.

Figure 7.

Renal cell carcinoma made more visible by adjusting the CT window level in a 63-year-old female: (a) the coronal reformatted renal corticomedullary phase CT in the soft tissue window (width/level: 350/50 HU) shows a subtle enhancing mass (open arrow) at the mid-pole of the right kidney. (b) Changing the window setting (width/level: 150/100 HU) in the same image shows an obvious small renal cell carcinoma (open arrow).

INTERPRETIVE ERROR

Despite the proper identification of lesions in the abdomen and pelvis, the interpretation of these findings can be incorrect. Interpretive error is defined as a failure to recognize the ramifications or the significance of a finding, leading to an incorrect final diagnosis.^2,4 This error arises from lack of knowledge, cognitive bias, misleading clinical information and/or overreliance on previous examinations.⁴ Mistakes in previous reports can be perpetuated as an additional reason for diagnostic error.^3,4 These interpretive errors can be categorized as follows: (1) misinterpretation due to improper patient information, (2) mistaking a normal finding for a lesion, (3) diagnostic difficulty due to atypical manifestation of a lesion and (4) incorrect diagnosis by mistaking the origination of a lesion.

Misinterpretation due to improper patient information

Aside from imaging analysis, an important part of a radiologist work involves investigating patient clinical information and reviewing prior studies before formulating diagnoses on any imaging studies. If the patient's relevant clinical information is not provided and/or adequate communication with the clinician cannot be achieved, the risks of misinterpreting a lesion increase (Figures 8 and 9). Consequently, improper patient information can result in poor clinical outcomes. Communication with the clinician about the patient clinical history and/or investigating the electronic medical record, if available, can directly reduce the occurrence of lesion misinterpretation. Thus, acquiring appropriate clinical information is important to reduce adverse events.

Figure 8.

Misinterpreted intraperitoneal urinary bladder perforation due to incomplete patient information in a 45-year-old male: (a) the axial CT image shows free air (asterisk) in the anterior peritoneal space. In the initial CT interpretation, the site of bowel perforation was not identified, but bowel perforation was suspected. (b) After learning the patient history of trauma to the lower abdomen, urinary bladder perforation was diagnosed on subsequent analysis, indicated by the irregularly margined wall (open arrows) of the bladder dome with adjacent extraluminal air bubbles (arrows) which might have been caused by Foley catheter insertion.

Figure 9.

A pelvic mass misinterpreted as a normal uterus owing to missing patient information in a 56-year-old female: (a) the axial CT image shows an ovoid pelvic mass (arrows) with smooth margins that was thought to be the uterus during the initial interpretation. (b) Axial CT images obtained at the 3-month follow-up show that the heterogeneous mass (arrows) had increased in size and had an irregular margin. Initially, the history of hysterectomy with bilateral salpingo-oophorectomy due to ovarian cancer was not provided. Ultimately, this mass was presumed to be a recurrence.

Mistaking a normal finding for a lesion

This error, also known as a false-positive error, refers to mistaking a normal finding as abnormal.¹⁰ For example, collapsed bowel (Figure 10) and appendiceal dilatation can be misinterpreted as malignancy and acute appendicitis, respectively. A bowel loop or vascular structure can be misinterpreted as a metastatic mass or an enlarged lymph node.³ By carefully analyzing the serial images along the bowel or vascular course in a picture archiving and communication system or workstation, radiologists may be able to distinguish between normal bowel/vascular structure and a solid mass.³ Reviewing and comparing previous imaging studies for normal bowel/vascular structure and using additional imaging studies such as appendiceal ultrasound could also be helpful.

Figure 10.

Normal bowel contraction misinterpreted as sigmoid colon cancer in a 64-year-old male with a history of lung cancer: (a, b) axial (a) and coronal (b) reformatted CT images show eccentric wall thickening of the sigmoid colon (arrows) with suspicious pericolic fat infiltration (arrowheads). On initial CT interpretation, sigmoid colon cancer was suspected. However, upon colonoscopy, sigmoid colon abnormalities were not seen. Thus, the finding was thought to be normal bowel contraction.

Diagnostic difficulty due to atypical manifestation of a lesion

The presence of atypical features in these cases can lead to an incorrect diagnosis or can be confused with other lesions. For example, atypical manifestations of pancreatic adenocarcinoma, such as a mass without upstream ductal dilatation (Figure 11), are often challenging to diagnose correctly. In another example, perforated gallbladder can be the cause of misdiagnosis of gallbladder carcinoma. It can be challenging to differentiate mass-forming (replacing the gallbladder) gallbladder carcinoma with liver invasion from gallbladder perforation with liver abscess (Figure 12). Familiarity with the various radiologic appearances of diseases is essential for correct diagnosis, differentiation and appropriate management.

Figure 11.

Pancreatic adenocarcinoma misdiagnosed as either a neuroendocrine tumour or a solid pseudopapillary tumour in a 51-year-old female with left flank pain: the axial CT image shows an ill-defined, ovoid, peripheral, enhancing mass (arrows) with a central cystic portion in the pancreatic tail. Dilatation of the upstream pancreatic duct is not seen. The initial interpretation was a neuroendocrine tumour or a solid pseudopapillary tumour. However, pancreatic adenocarcinoma was confirmed surgically.

Figure 12.

Gallbladder carcinoma misinterpreted as gallbladder perforation in an 83-year-old female with right upper quadrant pain: (a) the coronal reformatted CT image shows focal discontinuity of gallbladder wall (open arrow) with a rim-enhancing low-attenuation (20–30 HU) mass (arrows) at the adjacent liver. Initially, gallbladder perforation with a liver abscess was suspected. (b) On ultrasonography, an isoechoic to hypoechoic mass (arrowheads) is seen in the liver adjacent to the gallbladder (asterisk), for which biopsy was performed. Adenocarcinoma arising from the gallbladder was confirmed.

Mistaking the origination of a lesion

Even if a lesion is identified, its origin might be unclear. For example, radiologists can have difficulty in identifying the origin of a large pelvic mass abutting multiple organs because of the wide spectrum of benign and malignant pathologies that can have a similar appearance (Figure 13). Exophytic tumours arising from the liver can mimic gastrointestinal stromal tumours arising from the stomach, which could lead to misdiagnosis (Figure 14). To minimize these errors, identifying the mass feeding or draining vessels on CT imaging can be helpful. The relevant clinical history along with a comprehensive analysis of the imaging findings and precise anatomical localization of the lesion allow one to narrow the differential diagnosis.

Figure 13.

A gastrointestinal stromal tumour misinterpreted as a subserosal uterine myoma in a 70-year-old female with lower abdominal pain: (a, b) axial (a) and coronal (b) reformatted CT images show an enhancing mass with a lobulated contour (arrows) located posterior to the ileum (arrowheads) and superior to the uterus (open arrow). This mass was initially misinterpreted as a uterine subserosal myoma. However, gastrointestinal stromal tumour originating from the distal ileum was confirmed surgically.

Figure 14.

Hepatocellular carcinoma misinterpreted as gastrointestinal stromal tumour in a 50-year-old male with uncontrolled gastric ulcer bleeding: (a, b) axial (a) and coronal (b) reformatted hepatic arterial phase CT images show a well-defined, ovoid mass (a and b, arrows) abutting the stomach and the left hepatic lobe. The mass was initially interpreted as a gastrointestinal stromal tumour arising from the stomach. However, hepatocellular carcinoma in the left hepatic lobe was confirmed surgically. (c) On careful retrospective evaluation with maximal intensity projection imaging, the left hepatic artery (open arrows) that supplies this mass can be seen.

CONCLUSION

The perceptual and interpretive errors reviewed above are challenges radiologists commonly confront when interpreting APCT imaging. To reduce perceptual errors, careful scrutiny of multiple lesions and blind spots, proper use of MPR images and proper use of the CT window settings are essential. These precautionary measures are helpful for improving an observer approach. To reduce interpretation error, one needs to consider the possibility of an anomaly or an atypical manifestation in order to recommend the appropriate additional studies. Acquiring the patient clinical information is also essential prior to interpreting the APCT. These efforts can reduce the chances of errors in the interpretation of APCT.