Artificial intelligence in radiology: Are Saudi residents ready, prepared, and knowledgeable?

Mawya A. Khafaji; Mohammed A. Safhi; Roia H. Albadawi; Salma O. Al-Amoudi; Salah S. Shehata; Fadi Toonsi

doi:10.15537/smj.2022.43.1.20210337

Article Figures & Data

Tables

View popup

Table 1

- Exposure assessment to AI in radiology corresponding to gender, training level and familiar with big data.

Questions/category	n (%)	P-value
		Gender	Training level	Big data
*Which radiological subspecialties do you foresee will be more influenced by AI in the next 5-10 years? (choose up to 3)*
Breast	99 (64.3)
Molecular/nuclear imaging	56 (36.4)
Neuroradiology	54 (35.1)
Thoracic	54 (35.1)
Emergency	32 (20.8)
Musculoskeletal	24 (15.6)
Cardiovascular	22 (14.3)
General	22 (14.3)
Gastrointestinal/abdominal	20 (13)
Interventional	17 (11)
Oncologic imaging	16 (10.4)
Head and neck	11 (7.1)
Urogenital	3 (1.9)
Pediatric	2 (1.3)
*Which techniques do you foresee will be the most important fields of AI applications in the next 5-10 years? (choose up to 3)*
Mammography	91 (59.1)
PET/nuclear	72 (46.8)
CT	69 (44.8)
Radiography	61 (39.6)
MRI	46 (29.9)
DXA	37 (24)
Angiography/fluoroscopy	18 (11.7)
Hybrid imaging	9 (5.8)
Ultrasound	8 (5.2)
Experimental imaging (animal models)	6 (3.9)
Optical imaging	4 (2.6)
*Which of the following AI applications do you think are more relevant as aids to the radiological profession? (choose up to 3)*
Detection in asymptomatic subjects (screening)	82 (53.2)
Detection of incidental findings	74 (48.1)
Image post-processing	73 (47.4)
Imaging protocol optimization	54 (35.1)
Support to structured reporting	44 (28.6)
Lesion characterization/diagnosis in symptomatic subjects	43 (27.9)
Staging/restaging in oncology	43 (27.9)
Quantitative measure of imaging biomarkers	31 (20.1)
Prognosis	12 (7.8)
*Do you foresee an impact of AI on the professional life of radiologists in terms of the number of job positions in the next 5-10 years?*
No	67 (43.5)
Yes, job positions will be reduced	64 (41.6)	0.742	0.919^‡	0.869
Yes, job positions will increase	23 (14.9)
*Do you foresee an impact of AI on the professional life of radiologist in terms of total reporting workload in the next 5-10 years?*
No	29 (18.8)
Yes, it will increase	43 (27.9)	0.44	0.192	0.905
Yes, it will be reduced	82 (53.2)
*In the next 5-10 years, the use of AI-based applications will make radiologists’ duties*
More technical	28 (18.2)
More clinical	38 (24.7)
Unchanged	9 (5.8)	0.566^‡	0.269^‡	0.244^‡
More technical and clinical	79 (51.3)	0.566^‡	0.269^‡	0.244^‡
*Do you think that in the next 5-10 years, the use of AI-based applications will help to reduce the need for subspecializing?*
No, radiologists will be more focused on radiology subspecialties	102 (66.2)
Yes, radiologists will be less focused on radiology subspecialties	16 (10.4)	0.685	0.033	0.065
The rate of dedication to subspecialties will remain unchanged	36 (23.4)
*In the next 5-10 years, who will take the legal responsibility of AI-system output?*
Radiologists	105 (68.2)
Other physicians (namely, clinicians asking for the imaging study)	9 (5.8)
Developers of AI applications	79 (51.3)
Insurance companies	35 (22.7)
*In the next 5-10 years, will patients accept a report from AI applications without supervision and approval by a physician?*
Yes	17 (11)
No	79 (51.3)	0.381	0.489^‡	0.847
Difficult to estimate at present	58 (37.7)
*What will be the role of radiologists in the development/validation of AI applications to medical imaging? (choose at most 3)*
Supervise all stages needed to develop an AI based application	97 (63)	0.320	0.169	0.687
Help in task definition	67 (43.5)	0.255	0.663	0.381
Develop AI-based applications	59 (38.3)	0.175	0.070	0.742
Provide labelled images	49 (31.8)	0.114	0.268	0.762
None	7 (4.5)	1.00^‡	0.452^‡	1.00^‡

PET: positron emitted tomography, CT: computed tomography, MRI: magnetic resonance imaging, DXA: dual-energy x-ray absorptiometry, AI: artificial intelligence, ^‡Fisher’s exact test

View popup

Table 2

- Artificial intelligence applications in radiology corresponding to gender, training level and familiar with big data.

Questions/answers	n (%)	P-value
		Gender	Training level	Big data
*Should radiologists be educated on (choose at most 3):*
Clinical use of AI applications	117 (76)
Advantages and limitations of AI applications	115 (74.7)
Technical methods, such as machine/deep learning algorithm	63 (40.9)
How to get into the driver seat in using AI	62 (40.3)
How to survive the AI revolution	29 (18.8)
How to avoid the use of AI applications	11 (7.1)
*What are the advantages of using AI? (choose at most 2)*
AI can speed up processes in health care	122 (79.2)	0.257	0.468	0.748
AI can help reduce medical errors	73 (47.4)	0.298	0.734	0.784
AI has no emotional exhaustion nor physical limitation	43 (27.9)	1.00	0.190	0.618
AI can deliver vast amounts of clinically relevant high-quality data in real time	27 (17.5)	1.00	0.440^‡	0.582
AI has no space-time constraint	16 (10.4)	0.860	0.118^‡	1.00
*What are you concerned about regarding the application of AI in medicine?*
It cannot be used to provide opinions in unpredicted situations due to inadequate information	59 (38.3)
It is not flexible enough to be applied to every patient	53 (34.4)
It is difficult to apply to controversial subjects	21 (13.6)	0.193^‡	0.198^‡	0.968^‡
The low ability to sympathize and consider the emotional well-being of the patient	11 (7.1)
It was developed by a less experienced medical clinician	10 (6.5)
*Are you involved in any research project based on AI-based application development*
Yes, testing	5 (3.2)
Yes, developing	10 (6.5)	0.485^‡	0.774^‡	0.027^‡
No, but planning to be involved	44 (28.6)	0.485^‡	0.774^‡	0.027^‡
No	95 (61.7)
*Would you be willing to help in learning about ML and training a ML algorithm so that it can imitate some of the tasks you perform as a radiologist?*
Yes	120 (77.9)	0.096	0.049	0.026
No	34 (22.1)	0.096	0.049	0.026

AI: artificial intelligence, ML: machine learning, ^‡Fisher’s exact test

View popup

Table 3

- Evaluation of AI effect on radiology and medicine.

Question	Strongly disagree	Disagree	Neutral	Agree	Strongly agree
	n (%)
Artificial intelligence will augment capability of radiologists and make radiologists more efficient	5 (3.2)	10 (6.5)	40 (26)	61 (39.6)	38 (24.7)
Radiologists should embrace artificial intelligence, and work with the IT industry for its application	0 (0)	4 (2.6)	34 (22.1)	63 (40.9)	53 (34.4)
You expect a significant acceleration of your work from new technologies (AI)	0 (0)	6 (3.9)	33 (21.4)	73 (47.4)	42 (27.3)
If artificial intelligence achieves high diagnostic accuracy, it should be used to evaluate radiological images alone	31 (20.1)	50 (32.5)	43 (27.9)	22 (14.3)	8 (5.2)
Artificial intelligence should be used as a support for evaluating radiological images	2 (1.3)	2 (1.3)	15 (9.7)	84 (54.5)	51 (33.1)

AI: artificial intelligence, IT: information technology

View popup

Table 4

- Perception of radiologist on AI in radiology.

Question	N/A	Disagree entirely	Rather disagree	Rather agree	Agree entirely
	n (%)
A potential application for AI in radiology (automated detection of pathologies in imaging examinations)	26 (16.9)	2 (1.3)	10 (6.5)	79 (51.3)	37 (24)
Artificial intelligence will improve medicine in general	14 (9.1)	3 (1.9)	11 (7.1)	77 (50)	49 (31.8)
These developments frighten me	23 (14.9)	41 (26.6)	41 (26.6)	38 (24.7)	11 (7.1)
These developments make radiology more exciting to me	18 (11.7)	13 (8.4)	12 (7.8)	66 (42.9)	45 (29.2)
Artificial intelligence should be part of residency training	17 (11)	9 (5.8)	11 (7.1)	67 (43.5)	50 (32.5)