Abstract
Background
Due to their crucial role in many diagnostic processes, a considerable number of papers in non-radiological medical journals contain images. We noticed that radiologists are seldom mentioned as co-authors.
Purpose
To investigate how often radiologists are involved as co-authors and to measure the influence of their involvement on the choice and quality of the illustrations and captions.
Material and Methods
We queried PubMed for papers in English with the search string “case report” in the title, examined them for the presence of radiological images, and excluded those published in radiological journals. We developed a scoring system, containing objective and subjective qualification criteria. In addition, we checked if a radiologist was involved as co-author, or mentioned in the Acknowledgments section. We performed a statistical analysis to check if the involvement of a radiologist had a significant effect on the overall quality of the case report.
Results
In 21% (45/218) of the papers, a radiologist was mentioned as co-author and in 3% (7/218) a radiologist was mentioned in the acknowledgments. In 76% (166/218), radiologists were neither involved as co-authors, nor mentioned in acknowledgments. We found statistically significant quality differences between these three groups (P < 0.001).
Conclusion
This study indicated that the quality of case reports can be improved when radiologists are involved in the preparation and publication of papers containing imaging studies.
Introduction
“Render honor to whom honor is due.” These words appear as a self-evident and straightforward principle. Unfortunately, this is not always the case when it comes to authorship of scientific papers. Authorship of medical images and captions are a good example. A considerable number of papers in medical journals contain images, due to their crucial role in many diagnostic processes. A radiologist or nuclear medicine specialist is seldom involved in the preparation of such a text (1). In what follows, whenever we mention “radiologist,” we refer to nuclear medicine physicians as well.
Although radiologists play a major role in the diagnostic work-up, in particular in the complex cases presented in most case reports, they often appear not to be involved in the authoring process. Apparently, clinical specialists still tend to see radiologists as “image providers,” not as clinical colleagues, despite the fact that radiologists are crucial in choosing the right method for this indication, the use of contrast agents, how to produce the best diagnostic image, and, of course, the interpretation of the images, an intellectual act not in any way different from the clinical contribution of other physicians.
There is an increasing demand for radiologists to become acknowledged as fully fledged medical specialists and to play a role in the diagnosis, management, and follow-up of patients. Unfortunately, patients are grossly unfamiliar with our profession (2,3). As artificial intelligence and deep learning algorithms (4–10) reshape the imaging world, it becomes even more essential to highlight the important role radiologists play in the diagnostic and therapeutic management of our patients. Some articles in the medical and lay press have stated that Western medicine will no longer need radiologists, thanks to deep machine learning on one hand and to outsourcing of medical image interpretation to cheaper workforces abroad on the other (11,12). Consequently, we, radiologists, need to become visible again and be recognized as clinical specialists. Co-authorship of case reports can contribute to this objective (13).
Over the course of many years as radiologists in an academic setting, we seldom noticed that radiologists were involved in the composition of case reports in non-radiological journals, even though these frequently contained medical images. Quite often, we were struck by the low quality of some images, as well as by many incomplete, erroneous, if not irrelevant captions. As far as we could see, this observation and its potential effect on the overall quality of the paper had not yet been investigated. We therefore decided to investigate how often radiologists were involved in the composition of non-radiological case as co-authors in non-imaging case reports containing radiological images. As a second objective, we measured the influence of the involvement of a radiologist on the choice and presentation of the illustrations, and on the content of the captions.
Material and Methods
On a single day, 31 December 2016, we queried PubMed for “case report” in the title. We chronologically registered the first 400 hits. We excluded papers in radiological journals and selected for further study only those in English on human medicine, of which a pdf file could be obtained via our university library. Included papers were screened for the presence of images from conventional radiography (CR), ultrasound (US), CT scan (CT), magnetic resonance imaging (MRI) or nuclear medicine (NM) studies.
To our knowledge, there were no universally accepted criteria for the assessment of the quality of the images and the captions in such papers. Therefore, we developed our own scoring system, in which we included both objective and subjective criteria. Examples of “objective criteria” were imaging modality, examined region, projection or plane, and MRI sequence. Examples of “subjective criteria” were the quality of the image and of its description (Table 1). To reduce as much as possible an arbitrary look and reader bias, scoring of radiological images was performed in consensus by a senior staff radiologist (AS) and a senior radiology resident (EL), and of the nuclear medicine images by a staff radiologist (AS) and a senior staff member from the department of Nuclear Medicine (SC).
Quality scoring parameters of radiological images.
We created a scoring sheet that could be adjusted to the particularities of the different types of images. There were small differences in the overall denominators for the different imaging modalities (Table 1). This approach allowed the freedom necessary due to the fact that more parameters can be scored where it concerns CT or MR images than images from US and CR.
Objective criteria for the captions were, among others:
Was the type of imaging modality (CR, US, CT, MRI) mentioned? Was the examined region (thorax, pituitary gland, shoulder…) correctly identified? Depending on the imaging modality, was the projection indicated (CR: PA, AP…) or the reconstruction/scan plane (CT and MRI: axial, sagittal, coronal…)? In the case of MRI scans, were the type of contrast and the sequence (T1, T2, DWI, SWI…) correctly labeled? If CT images were presented in different kernels or windows, were they correctly mentioned? Did the caption include information about whether or not a contrast agent was administered and, if so, did it specify type, concentration, and volume of the agent as well as the way of administration?
To rate the subjective criterion “quality of image description,” the following parameters were taken into consideration:
Were the image findings correctly described? Were correct radiological terms mentioned (e.g. “density” in case of CT, “intensity” in case of MRI…) and the precise anatomical location of the findings? Were patient data anonymized? Were arrows and annotations present, if relevant? Was image resolution sufficient? Were images correctly cropped? Were images relevant for the subject the authors wanted to illustrate?
For NM images the criteria were: mention of imaging modality; examined region; fusion technique (PET-CT, PET-MRI…); tracer (18F-FDG, GLP-1…); quality of the description of image findings; and image quality (Table 2).
Quality scoring parameters of nuclear medicine images.
We further registered the medical specialty of the first, second, and last authors, the name of the journal, the title of the paper, the date of publication, the number of pictures per modality, the presence or absence of a radiologist among the authors, the total number of authors, the ranking of this radiologist among the authors, and the presence of a radiologist in the acknowledgments.
Statistical analysis
We performed the statistical analysis of our results using R, version 3.2.3 (Wooden Christmas Tree). The quality score differences between case reports with a radiologist as co-author, with a radiologist in the acknowledgments, and without a radiologist were performed using the Kruskal–Wallis test. The Wilcoxon rank sum test with a Bonferroni correction for multiple testing was used for post hoc testing. The same stepwise analysis was performed to evaluate quality score differences between case reports without a radiologist as co-author, without MRI (Group 1); without a radiologist as co-author but with MRI (Group 2); with a radiologist as co-author, without MRI (Group 3); and with a radiologist as co-author and with MRI (Group 4). To evaluate the potential association between case reports with or without MRI and the presence or absence of a radiologist as co-author, Fisher’s exact test was used. Significance was set at an α-level of ≤0.05.
Results
Of 400 papers selected using PubMed, we excluded 25 because the full text in English was not available to us. Eleven were rejected because they were published in a radiological journal. The remaining 364 papers (91%) were screened for radiological images. We finally retained 218 for inclusion.
In 21% (45/218) of the papers, a radiologist was mentioned as co-author. In 3% (7/218) a radiologist was mentioned in the acknowledgments. In 76% (166/218), radiologists were neither involved as co-authors, nor mentioned in acknowledgments. We found statistically significant quality differences between these three groups (P < 0.001). Median quality scores in papers with and without a radiologist as co-author, and with a radiologist in the acknowledgments, were 74%, 50%, and 50%, respectively (Fig. 1). Quality scores were statistically significantly different between case reports with and without a radiologist as co-author (P < 0.001), as well as between case reports with a radiologist as co-author and a radiologist in the acknowledgments (P < 0.05).
Box plot of the quality scoring of case reports without involvement of a radiologist, with a radiologist in the acknowledgments and with a radiologist as co-author. No radiologist as co-author: median = 0.500 (range = 0.4070–0.6127); radiologist in acknowledgments: median = 0.500 (range = 0.3167–0.6054); radiologist as co-author: median = 0.7368 (range = 0.5833–0.8889).
We found that in papers at the lower end of the quality score spectrum, imaging specialists were mainly absent, while more were involved in papers at the higher end of the quality score.
Our next analysis took the imaging modality into account, assessing whether the presence of a radiologist was also important for non-MRI imaging modalities. We observed the same consistent quality trend as before (overall comparison: P < 0.001) (Fig. 2). Whenever a radiologist was involved as co-author, the quality scores were significantly higher (Group 1 vs. Group 3: P = 0.001; Group 1 vs. Group 4: P < 0.001; Group 2 vs. Group 3: P < 0.03; Group 2 vs. Group 4: P < 0.001); case reports without a radiologist, irrespective of the question whether they contained MRI images or not, did not show any significant quality difference (Group 1 vs. Group 2: P = 0.53). The same was observed for case reports with a radiologist: no significant difference in quality score was found between case reports with or without MRI images (Group 3 vs. Group 4: P = 0.63). Case reports with MRI and with a radiologist involved obtained the highest quality scores (Fig. 2). Furthermore, an association between MRI and the presence of a radiologist as a co-author: 29% of case reports containing MR images had a radiologist as co-author was shown, while only 13% of case reports without MR images (P < 0.001).
Box plot of the quality scoring of case reports with or without MR-images, with or without a radiologist involved as co-author. Group 1: no radiologist, no MRI: median = 0.5000 (range = 0.3750–0.5833); Group 2: no radiologist, MRI: median = 0.5157 (range = 0.4390–0.6364); Group 3: radiologist, no MRI: median = 0.6250 (range = 0.5649–0.7714); Group 4: radiologist, MRI: median = 0.7917 (0.6364–0.9231).
The report with the maximum score with a description of imaging findings of a very high level is shown in Fig. 3.
Image from case report number 42, with a high overall quality score. In this case report, a radiologist was involved as co-author. The caption correctly described the imaging modality, examined region, scan plane, sequence, and contrast administration. It also had an impeccable quality of image description and image quality with the use of correct radiological terms, precise anatomical location of the findings, presence of arrows/annotations, and a high image resolution.
Some figure legends were of unacceptably poor quality according to the scoring system (Fig. 4). In some cases, the imaging modality was not even mentioned, while in others, the imaging findings were not described at all. We found case reports containing information allowing patient identification. There were instances in which the image quality itself was low, where it concerned poor spatial orientation, poor image resolution, suboptimal contrast, absent or poor cropping, and many other quality issues.
Image from case report number 177. Figure of a strikingly low quality according to the criteria of the study and with a very short, incomplete caption. Moreover, there the images are distorted. In this case report, no radiologist was involved as co-author, nor was one mentioned in the acknowledgments.
Discussion
In our study, in only 21% of case reports containing radiological images a radiologist was mentioned as a co-author, a disappointingly low number. We do agree that the simple use of radiological images in a paper in itself does not justify authorship. The situation is different when a radiologist is involved in the preparation of the paper. Such involvement may be fundamental, such as taking part in planning the case report and performing the literature search, or auxiliary, when it comes to selecting appropriate images and writing dedicated captions. These tasks are to be considered equivalent to writing the text. But there is more.
Our scores of overall quality of the images and captions – 74% with a radiologist as co-author versus 50% without – demonstrate that a radiologist as a co-author in a publication significantly improves its quality. The striking difference between good and bad examples of image descriptions, and even the selection and reproduction of images, suggest that the contribution of a radiologist to such papers is highly desirable.
The median quality score in case reports mentioning a radiologist in the acknowledgments section of the papers was low. This may be due to the low sample size (n = 7). Another explanation could be those radiologists were not actively involved in the preparation of the paper, but were simply acknowledged because the images came from their department.
One might wonder whether the contribution of a radiologist is really necessary when only basic imaging, such as that of CR, is being used. We are fully conscious that, up to a certain level, such images can be reliably interpreted by a dedicated non-radiologist. When it comes to more complex imaging modalities however, such as CT, MRI, PET, SPECT, angiography or mammography, the input of an imaging specialist provides a very significant added value. The case reports with the best scores in our study were those which included MRI and benefited from the involvement of a radiologist. The case reports containing MR images were also more likely to have a radiologist as co-author.
The guidelines for authors of most journals include criteria concerning the images (format, minimum resolution, etc.). By and large, such guidelines do not mention the specific role of the authors. The requirement that a radiologist be involved in the selection and the description of images could improve the quality of the images and figure legends and thus the overall quality of the paper. Case reports in non-radiological journals are most frequently reviewed by clinical specialists. Including radiologists as reviewers could be another way to improve the quality.
This study did have a few limitations. Establishing a large number of criteria (Tables 1 and 2) came as close as possible to an objective assessment, but any quality scoring will be subjective up to a certain degree. A second limitation was the size of the sample. Only 218 case reports were scored, out of an initial number of 400. Finally, this study did not correlate the quality of the findings with the impact factor of the journals. We rejected this criterion because journals with a high impact factor rarely publish case reports.
In conclusion, in this analysis of case reports containing medical images published in non-imaging journals, only a minority included a radiologist as co-author. Our findings indicate that the quality of case reports may improve significantly if imaging specialists are involved in the preparation and publication of case reports.
Footnotes
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.