Editor’s Corner: February 2024

Current Practice Patterns, Challenges, and Need for Education in Performing and Reporting Advanced Pelvic US and MRI to Investigate Endometriosis: A Survey by the Canadian Association of Radiologists Endometriosis Working Group

Motivation

Endometriosis is a common chronic condition impacting premenopausal women, ¹ which results in significant morbidity. Ultrasound and Magnetic Resonance Imaging (MRI) play an important role in the diagnosis and follow-up of endometriosis. The CAR endometriosis working group conducted a national survey to identify current practice patterns pertaining to endometriosis imaging and help shape future imaging practice guidelines.

Methodology

A survey was developed based on expert opinion and made available to all CAR members. The survey was targeted to radiologists involved in advanced pelvic ultrasound and MRI imaging for the assessment of endometriosis. Descriptive statistics were subsequently collected and summarized.

Results

Eighty-nine radiologists responded to the survey, most from Ontario (23%) and Quebec (24%). A minority (38%) of respondents performed advanced pelvic US for endometriosis, after undergoing several different training pathways. Most respondents were interested in obtaining additional training in this subject (71%).

Most respondents were involved with MRI for endometriosis (70%), and the majority were interested in additional training (63%). The MRI protocols used at the different institutions were heterogeneous, with 55% utilizing an antispasmodic agent.

Insight

Future works needs to focus on raising awareness of the value of imaging for guiding the management of endometriosis and positively affecting patient outcomes. More focus should also be placed on ensuring that radiology learners are being adequately trained in this subject. The results of the survey can also be used to develop standardized reporting templates and Canadian-specific guidelines.

Accuracy of Information and References Using ChatGPT-3 for Retrieval of Clinical Radiological Information

Motivation

To investigate the accuracy of responses provided by the artificial intelligence chatbot ChatGPT-3 to clinical radiology questions. ² The aim was to also assess the response pertaining to the references provided for these answers.

Methodology

ChatGPT-3 was tested with 88 questions from all radiology subspecialties. The categories of questions included: imaging findings of a condition (n = 15), imaging findings of 2 differential diagnoses (n = 29), modality-related (n = 11), indications and contraindications (n = 7), prognosis (n = 5), as well as mixed content including anatomical questions and reference values (n = 21). Response accuracy was assessed using a 5-point Likert scale.

Questions were followed up by a prompt to provide a reference. Reference accuracy was then assessed, including if the article was indexed in PubMed.

Results

Reasonable accuracy was obtained from ChatCPT-3, with 67% of questions answered correctly, 17% largely correct, 7% half correct, 4.5% mostly incorrect, and 4.5% entirely incorrect. However, of 343 references provided, only 124 (36%) were real references, while the remaining ones appear to have been generated by the chatbot. In addition, some of the real references provided had little or no relevant background information needed to answer the specific question.

Insight

This pilot study raises awareness of the potential value, as well as the potential pitfalls associated with the utilization of ChatGPT-3 in answering and providing references for clinical radiology questions. However, caution is needed for the time being if it is to be used in the clinical context.

Prevalence of ‘Fat-Poor’ Adrenal Adenomas at Chemical-Shift MRI

Motivation

CT and Magnetic Resonance imaging is used to make the diagnosis of an adrenal adenoma, ³ primarily based on the ability to detect microscopic fat. However, a subset of these benign lesions are ‘lipid-poor’. The aim was to identify the proportion of these ‘lipid-poor’ adrenal adenomas on 2D chemical-shift MRI from a large sample of indeterminant adrenal masses.

Methodology

A single-centre prospective study was performed. All patients referred for imaging of indeterminant adrenal masses were scheduled for an unenhanced adrenal MRI. The reference standard for an adrenal adenoma was considered as one of the following: 1 year size stability with no biochemical evidence of pheochromocytoma, 5-year clinical follow-up with no signs of an adrenal disorder, or presence of microscopic fat with absence of biochemical evidence of pheochromocytoma. Two blinded radiologists independently measured the 2D chemical shift signal intensity (SI) index on MRI (SI index >16.5% confirmed the presence of microscopic fat). The radiologists further assessed the unenhanced CT attenuation in cases where CT was available.

Results

One hundred four patients with 127 indeterminant adrenal masses were assessed. Of the adrenal masses, 119 (94%) were adrenal adenoma. Of these, 117 (98%) met the criteria based on the SI index. Only 2 (2%) of the adenomas were considered ‘lipid-poor’ by MRI in this cohort. In the cases where CT was available, 17 of 50 (34%) of adenomas were considered ‘lipid-poor’ with attenuation >10 HU.

Insight

Based on a relatively large sample size, the frequency of ‘lipid-poor’ adrenal adenomas appears to be low when utilizing 2D chemical shift MRI and calculating the signal intensity index. Unenhanced CT, on the other hand, did not perform as well in depicting microscopic fat within adrenal adenomas. The results suggest that chemical-shift MRI is superior to unenhanced CT for determining the presence/absence of microscopic fat in adrenal masses.

MRI Markers of Degenerative Disc Disease in Young Patients with Multiple Sclerosis

Motivation

Multiple sclerosis (MS) is a common cause of disability in young adults. ⁴ MRI is widely used in the diagnostic work-up, as well as in identify potential mimics and confounders of MS, including degenerative disc disease (DDD) and associated myelopathy. The current study assessed the presence and extent of DDD of the cervical spine in young patients (<35 years of age) with MS.

Methodology

A single-centre retrospective study was performed. Cervical spine MRI studies were reviewed by multiple readers and interobserver agreement was assessed. The presence, distribution, and severity of DDD of the cervical spine was independently assessed by each reader. The location of cord signal abnormalities was assessed in relation to the level of the DDD, irrespective of etiology (myelopathy or MS plaque).

Results

Eighty patients (51 females, mean age 26) with MS were included. At least mild DDD was identified in 91% of patients. Substantial interobserver agreement was seen pertaining to the DDD level and severity, and very good for cord signal change. Cord signal abnormalities were observed in 56%-63% of cases. Of these, cord signal abnormalities occurred exclusively at DDD levels in only 10%-15% of cases.

Insight

A large proportion of this young MS patient cohort had at least mild DDD, which is more than expected given their age. Additional work is required to determine the underlying pathophysiology, including any link to altered biomechanics. In addition, cord lesions were noted to occur independently of DDD.

Characterizing the Variety of Call Structures Across Canadian Diagnostic Radiology Postgraduate Medical Education Programs

Motivation

There are 16 Diagnostic Radiology (DR) residency training programs in Canada. ⁵ All programs require residents independently cover after-hours emergency and inpatient imaging interpretation (call) as part of their training. The goal is to determine the call structure and models utilized by the Canadian DR residency programs. The aim is to tabulate this data into an online database for future reference of the DR residency programs.

Methodology

A questionnaire was sent to resident representatives from each of the Canadian DR residency programs. Questions addressed: institution, call structure, call responsibilities, call support, call preparation, as well as pre-call and post-call processes.

Results

Twelve of the 16 programs (75%) have a call structure that can require residents to work 24 consecutive hours or more. This includes time spent performing regular clinical duties and/or on-call work. In fact, three programs require their residents to review their call cases prior to ending their shift. Only one program provides direct in-house attending support for all resident call shifts.

Insight

There is marked heterogeneity in the DR residency on-call structures across Canada. This is partly related to differing program sizes, departmental structure, and catchment areas. The high prevalence of programs requiring residents to work continuously for 24 hours or more may contribute to resident burnout.