312 B. Amaoui et al.: Radioprotection 2025, 60( 4), 310 – 317
Table 2. Data for acquisition and reconstruction of images using AI according to the two groups( expressed in %)
The use of AI in the acquisition and reconstruction of radiotherapy images Manual Partially automated Fully automated Uncertain
G1 G2 G1 G2 G1 G2 G1 G2 P-value
Image intervention and reconstruction |
20 |
15.4 |
70 |
69.2 |
10 |
7.7 |
0 |
7.7 |
1.000 |
Image fusion |
30 |
15.4 |
50 |
61.5 |
20 |
15.4 |
0 |
7.7 |
0.910 |
Target volume contouring |
50 |
53.8 |
20 |
23.1 |
20 |
23.1 |
10 |
0 |
0.919 |
OAR contouring |
40 |
76.9 |
40 |
23.1 |
10 |
0 |
10 |
0 |
0.163 |
Setting up a plan |
40 |
46.2 |
50 |
53.8 |
10 |
0 |
0 |
0 |
0.656 |
Plan optimisation |
40 |
15.4 |
50 |
61.5 |
10 |
0 |
0 |
23.1 |
0.171 |
Plan evaluation |
50 |
38.5 |
50 |
38.5 |
0 |
0 |
0 |
23.1 |
0.316 |
Quality assurance |
20 |
30.8 |
30 |
23.1 |
20 |
0 |
30 |
46.2 |
0.456 |
Image analysis and matching |
30 |
53.8 |
50 |
30.8 |
20 |
7.7 |
0 |
7.7 |
0.505 |
3.1 Socio-professional characteristics of the study population
The socio-professional characteristics of the population who took part in our survey are summarised in Table 1. Of the participants, 26.1 % were male and 73.9 % female. In terms of professional experience, 65.2 % had less than 5 years’ experience, 4.3 % between 5 and 10 years, 17.4 % between 10 and 20 years, and 13 % more than 20 years. In terms of sector of activity, 34.8 % worked in the public sector, 8.7 % in the private sector and 56.5 % in teaching hospitals.
3.2 Areas of RT mastered by participants
Concerning the areas of radiotherapy mastered, all the onco-radiotherapists( G1) reported proficiency in CT simulation and CT fusion / contouring, 70 % were proficient in dosimetry and treatment, and only 30 % were proficient in quality assurance. As for the Onco-Radiotherapy Residents( G2), all of them stated that they were proficient in CT simulation, 84.6 % were proficient in CT Fusion / Contouring, 46.2 % were proficient in dosimetry and treatment, and only 7.7 % were proficient in quality assurance.
3.2 Participants’ knowledge of AI
The results do not show a significant association between the two groups and their knowledge of AI( p = 0.075). 60 % of the G1s stated that they had moderate knowledge of AI, whereas 72 % of the G2s stated that they had limited knowledge of AI. In addition, 66.7 % of G1s and 33.3 % of G2s stated that they had obtained their knowledge of AI through self-directed learning, whereas 33.3 % of G1s and 50 % of G2s had acquired this knowledge through continuing education, with no significant difference between the two groups( p = 0.567). Furthermore, all G1s and G2s expressed interest in learning more about AI. In this context, 70.0 % of G1s and 61.5 % of G2s would like to know more about the theoretical foundations of AI, all of the two groups would like to know more about the clinical applications of AI, 70.0 % of G1s and 61.5 % of G2s would like to know more about the ethical implications of AI, with no significant difference between the two groups( p = 1.000).
60 % of G1s and 53.8 % of G2s prefer face-to-face workshops as a method of acquiring AI technologies safely and effectively, although 40 % of G1s and 30.8 % of G2s prefer online self-directed resources on an online platform. It should also be noted that 50 % of G1s and 61.5 % of G2s participants indicated insufficient training to use AI technologies safely and effectively.
3.4 The use of AI in the acquisition and reconstruction of radiotherapy images
According to the Table 2, 70.0 % of G1s and 69.2 % of G2s said that the image acquisition and reconstruction function is partially automated at their hospitals, 50.0 % of G1s and 61. 5 % of G2s said that the image fusion function is also partially automated, and around 50 % of both groups said that the contouring of target volumes is performed manually, and 40 % of G1s and 77 % of G2s say that the contouring of organs at risk is done also manually. Approximately 50 % of both groups stated that the options for setting up optimisation and plan evaluation, as well as image analysis and matching, are partially automated.
3.5 The impact of AI on the productivity of radiotherapy practices in image acquisition and reconstruction
According to Table 3 the majority of G1s supported increased productivity in image intervention and reconstruction functions( 70 %), image fusion( 70 %), target volume contouring( 60 %), organ at risk contouring( 70 %), plan implementation( 80 %), plan optimisation and evaluation( 70 %), quality assurance( 60 %), and image analysis and matching( 90 %). The majority of G2s were also in favour of increased productivity in the following functions: Image Intervention and Reconstruction( 77 %), Image Fusion( 79.2 %), Target Volume Contouring( 61. 5 %), Contouring of organs at risk( 69.2 %), Plan implementation( 61.5 %), Plan optimisation( 84.4 %), Plan evaluation( 69.2 %) and Image analysis and matching( 61.5 %).