ABSTRACT
INTRODUCTION: In Australia, 64% of pharmacists are women but continue to be under-represented. Generative artificial intelligence (AI) is potentially transformative but also has the potential for errors, misrepresentations, and bias. Generative AI text-to-image production using DALL-E 3 (OpenAI) is readily accessible and user-friendly but may reinforce gender and ethnicity biases. METHODS: In March 2024, DALL-E 3 was utilized to generate individual and group images of Australian pharmacists. Collectively, 40 images were produced with DALL-E 3 for evaluation of which 30 were individual characters and the remaining 10 images were comprised of multiple characters (N = 155). All images were independently analysed by two reviewers for apparent gender, age, ethnicity, skin tone, and body habitus. Discrepancies in responses were resolved by third-observer consensus. RESULTS: Collectively for DALL-E 3, 69.7% of pharmacists were depicted as men, 29.7% as women, 93.5% as a light skin tone, 6.5% as mid skin tone, and 0% as dark skin tone. The gender distribution was a statistically significant variation from that of actual Australian pharmacists (P < .001). Among the images of individual pharmacists, DALL-E 3 generated 100% as men and 100% were light skin tone. CONCLUSIONS: This evaluation reveals the gender and ethnicity bias associated with generative AI text-to-image generation using DALL-E 3 among Australian pharmacists. Generated images have a disproportionately high representation of white men as pharmacists which is not representative of the diversity of pharmacists in Australia today.
ABSTRACT
Generative artificial intelligence (AI) algorithms for both text-to-text and text-to-image applications have seen rapid and widespread adoption in the general and medical communities. While limitations of generative AI have been widely reported, there remain valuable applications in patient and professional communities. Here, the limitations and biases of both text-to-text and text-to-image generative AI are explored using purported applications in medical imaging as case examples. A direct comparison of the capabilities of four common text-to-image generative AI algorithms is reported and recommendations for the most appropriate use, DALL-E 3, justified. The risks use and biases are outlined, and appropriate use guidelines framed for use of generative AI in nuclear medicine. Generative AI text-to-text and text-to-image generation includes inherent biases, particularly gender and ethnicity, that could misrepresent nuclear medicine. The assimilation of generative AI tools into medical education, image interpretation, patient education, health promotion and marketing in nuclear medicine risks propagating errors and amplification of biases. Mitigation strategies should reside inside appropriate use criteria and minimum standards for quality and professionalism for the application of generative AI in nuclear medicine.
ABSTRACT
The emergence of ChatGPT has challenged academic integrity in teaching institutions, including those providing nuclear medicine training. Although previous evaluations of ChatGPT have suggested a limited scope for academic writing, the March 2023 release of generative pretrained transformer (GPT)-4 promises enhanced capabilities that require evaluation. Methods: Examinations (final and calculation) and written assignments for nuclear medicine subjects were tested using GPT-3.5 and GPT-4. GPT-3.5 and GPT-4 responses were evaluated by Turnitin software for artificial intelligence scores, marked against standardized rubrics, and compared with the mean performance of student cohorts. Results: ChatGPT powered by GPT-3.5 performed poorly in calculation examinations (31.4%), compared with GPT-4 (59.1%). GPT-3.5 failed each of 3 written tasks (39.9%), whereas GPT-4 passed each task (56.3%). Conclusion: Although GPT-3.5 poses a minimal risk to academic integrity, its usefulness as a cheating tool can be significantly enhanced by GPT-4 but remains prone to hallucination and fabrication.
Subject(s)
Nuclear Medicine , Humans , Artificial Intelligence , Radionuclide Imaging , Students , SoftwareABSTRACT
ChatGPT chatbot powered by GPT 3.5 was released in late November 2022 but has been rapidly assimilated into educational and clinical environments. Method: Insight into ChatGPT capabilities was undertaken in an interview-style approach with the chatbot itself. Results: ChatGPT powered by GPT 3.5 exudes confidence in its capabilities in supporting and enhancing student learning in nuclear medicine and in supporting clinical practice. ChatGPT is also self-aware of limitations and flaws in capabilities and the risks these pose to academic integrity. Conclusion: Further objective evaluation of ChatGPT capabilities in authentic learning and clinical scenarios is required.
Subject(s)
Learning , Nuclear Medicine , Humans , Software , StudentsABSTRACT
Academic integrity has been challenged by artificial intelligence algorithms in teaching institutions, including those providing nuclear medicine training. The GPT 3.5-powered ChatGPT chatbot released in late November 2022 has emerged as an immediate threat to academic and scientific writing. Methods: Both examinations and written assignments for nuclear medicine courses were tested using ChatGPT. Included was a mix of core theory subjects offered in the second and third years of the nuclear medicine science course. Long-answer-style questions (8 subjects) and calculation-style questions (2 subjects) were included for examinations. ChatGPT was also used to produce responses to authentic writing tasks (6 subjects). ChatGPT responses were evaluated by Turnitin plagiarism-detection software for similarity and artificial intelligence scores, scored against standardized rubrics, and compared with the mean performance of student cohorts. Results: ChatGPT powered by GPT 3.5 performed poorly in the 2 calculation examinations (overall, 31.7% compared with 67.3% for students), with particularly poor performance in complex-style questions. ChatGPT failed each of 6 written tasks (overall, 38.9% compared with 67.2% for students), with worsening performance corresponding to increasing writing and research expectations in the third year. In the 8 examinations, ChatGPT performed better than students for general or early subjects but poorly for advanced and specific subjects (overall, 51% compared with 57.4% for students). Conclusion: Although ChatGPT poses a risk to academic integrity, its usefulness as a cheating tool can be constrained by higher-order taxonomies. Unfortunately, the constraints to higher-order learning and skill development also undermine potential applications of ChatGPT for enhancing learning. There are several potential applications of ChatGPT for teaching nuclear medicine students.
Subject(s)
Nuclear Medicine , Humans , Artificial Intelligence , Radionuclide Imaging , Students , LearningABSTRACT
Academic integrity in both higher education and scientific writing has been challenged by developments in artificial intelligence. The limitations associated with algorithms have been largely overcome by the recently released ChatGPT; a chatbot powered by GPT-3.5 capable of producing accurate and human-like responses to questions in real-time. Despite the potential benefits, ChatGPT confronts significant limitations to its usefulness in nuclear medicine and radiology. Most notably, ChatGPT is prone to errors and fabrication of information which poses a risk to professionalism, ethics and integrity. These limitations simultaneously undermine the value of ChatGPT to the user by not producing outcomes at the expected standard. Nonetheless, there are a number of exciting applications of ChatGPT in nuclear medicine across education, clinical and research sectors. Assimilation of ChatGPT into practice requires redefining of norms, and re-engineering of information expectations.
Subject(s)
Artificial Intelligence , Nuclear Medicine , HumansABSTRACT
INTRODUCTION: Pre-clinical molecular imaging, particularly with mice, is an essential part of drug and radiopharmaceutical development. There remain ethical challenges to reduce, refine and replace animal imaging where possible. METHOD: A number of approaches have been adopted to reduce the use of mice including using algorithmic approaches to animal modelling. Digital twins have been used to create a virtual model of mice, however, exploring the potential of deep learning approaches to digital twin development may enhance capabilities and application in research. RESULTS: Generative adversarial networks produce generated images that sufficiently resemble reality that they could be adapted to create digital twins. Specific genetic mouse models have greater homogeneity making them more receptive to modelling and suitable specifically for digital twin simulation. CONCLUSION: There are numerous benefits of digital twins in pre-clinical imaging including improved outcomes, fewer animal studies, shorter development timelines and lower costs.
Subject(s)
Artificial Intelligence , Molecular Imaging , Animals , Mice , Computer Simulation , RadiopharmaceuticalsABSTRACT
Ventilation and perfusion (V/Q) lung scintigraphy has been used in the assessment of patients with suspected pulmonary embolism for more than 50 y. Advances in imaging technology make SPECT and SPECT/CT feasible. This article will examine the application and technical considerations associated with performing 3-dimensional V/Q SPECT and the contribution of a coacquired CT scan. The literature tends to be mixed and contradictory in terms of appropriate investigation algorithms for pulmonary embolism. V/Q SPECT and SPECT/CT offer significant advantages over planar V/Q, with or without the advantages of Technegas ventilation, and if available should be the preferred option in the evaluation of patients with suspected pulmonary embolism.
Subject(s)
Pulmonary Embolism , Tomography, Emission-Computed, Single-Photon , Humans , Tomography, Emission-Computed, Single-Photon/methods , Tomography, X-Ray Computed/methods , Lung , Single Photon Emission Computed Tomography Computed Tomography , Ventilation-Perfusion RatioABSTRACT
The scope of practice of the medical radiation practitioner demands knowledge and understanding of the indications, contraindications, warnings, precautions, proper use, drug interactions and adverse reactions of a variety of medications. The risk of patient deterioration or acute emergent event, particularly following contrast administration, makes the command of crash cart medications particularly important. This article explores the pharmacological principles of medications most likely to be required in a medical emergency in the medical radiation department and in particular by the computed tomography (CT) technologist. The article also outlines early warning signs to assist in identifying the emergent or deteriorating patient. The learning outlined is designed to equip medical radiation practitioners with the capacity to identify and respond to a medical emergency typical of the medical radiation department, and to respond to that situation with the appropriate use of emergency medications where appropriate. The ability of medical radiation practitioners to recognise and respond to (including the use of medicines) the deteriorating patient or circumstances of a medically urgent nature are key capabilities required to meet minimum standards for Medical Radiation Practice Board of Australia registration and National Safety and Quality Health Service standards.
Subject(s)
Emergency Service, Hospital , Tomography, X-Ray Computed , Humans , AustraliaABSTRACT
Developments in artificial intelligence, particularly convolutional neural networks and deep learning, have the potential for problem solving that has previously confounded human intelligence. Accurate prediction of radiation dosimetry pre-treatment with scope to adjust dosing for optimal target and non-target tissue doses is consistent with striving for improved the outcomes of precision medicine. The combination of artificial intelligence and production of digital twins could provide an avenue for an individualised therapy doses and enhanced outcomes in theranostics. While there are barriers to overcome, the maturity of individual technologies (i.e. radiation dosimetry, artificial intelligence, theranostics and digital twins) places these approaches within reach.
Subject(s)
Artificial Intelligence , Neural Networks, Computer , Humans , Precision Medicine , RadiometryABSTRACT
Conventional radiomics in nuclear medicine involve hand-crafted and computer-assisted regions of interest. Recent developments in artificial intelligence (AI) have seen the emergence of AI-augmented segmentation and extraction of lower order traditional radiomic features. Deep learning (DL) affords the opportunity to extract abstract radiomic features directly from input tensors (images) without the need for segmentation. These fourth-order, high dimensional radiomics produce deep radiomics and are well suited to the data density associated with the molecular nature of hybrid imaging. Molecular radiomics and deep molecular radiomics provide insights beyond images and quantitation typical of semantic reporting. While the application of molecular radiomics using hand-crafted and computer-generated features is integrated into decision-making in nuclear medicine, the acceptance of deep molecular radiomics is less universal. This manuscript aims to provide an understanding of the language and principles associated with radiomics and deep radiomics in nuclear medicine.
Subject(s)
Artificial Intelligence , Diagnostic Imaging , Radionuclide ImagingABSTRACT
A higher degree of emotional intelligence among health professionals has been shown to result in better patient care and improved wellbeing of the health professional. For nuclear medicine, emotional competence of staff and emotional proficiency of institutions, are important expectations. Nonetheless, there is a paucity of material outlining purposeful honing of emotional intelligence, or the tools for such development, across the literature. While the hidden curriculum provides powerful and authentic educational opportunities, incidental or accidental (organic) capability development does not benefit overall professionalism. Deliberate curricula can be achieved through a scaffold of emotional training and immersion programs that allow the nuclear medicine student or practitioner to recognize and foster emotionally safe environments. This requires careful planning to drive the emotional intelligence pipeline. Central to this is an understanding of learning taxonomies. There remain substantial gaps between the most and least emotionally insightful that could be addressed by rich immersive activities targeting emotional proficiency among students and the graduate workforce.
ABSTRACT
INTRODUCTION: While artificial intelligence (AI) and recent developments in deep learning (DL) have sparked interest in medical imaging, there has been little commentary on the impact of AI on imaging technologists. The aim of this survey was to understand the attitudes, applications and concerns among nuclear medicine and radiography professionals in Australia with regard to the rapidly emerging applications of AI. METHODS: An anonymous online survey with invitation to participate was circulated to nuclear medicine and radiography members of the Rural Alliance in Nuclear Scintigraphy and the Australian Society of Medical Imaging and Radiation Therapy. The survey invitations were sent to members via email and as a push via social media with the survey open for 10 weeks. All information collected was anonymised and there is no disclosure of personal information as it was de-identified from commencement. RESULTS: Among the 102 respondents, there was a high level of acceptance of lower order tasks (e.g. patient registration, triaging and dispensing) and less acceptance of high order task automation (e.g. surgery and interpretation). There was a low priority perception for the role of AI in higher order tasks (e.g. diagnosis, interpretation and decision making) and high priority for those applications that automate complex tasks (e.g. quantitation, segmentation, reconstruction) or improve image quality (e.g. dose / noise reduction and pseudo CT for attenuation correction). Medico-legal, ethical, diversity and privacy issues posed moderate or high concern while there appeared to be no concern regarding AI being clinically useful and improving efficiency. Mild concerns included redundancy, training bias, transparency and validity. CONCLUSION: Australian nuclear medicine technologists and radiographers recognise important applications of AI for assisting with repetitive tasks, performing less complex tasks and enhancing the quality of outputs in medical imaging. There are concerns relating to ethical aspects of algorithm development and implementation.
Subject(s)
Artificial Intelligence , Deep Learning , Australia , Humans , Radiography , Radionuclide ImagingABSTRACT
The aim of this study is to assess the influence of semiquantitative PET-derived parameters as well as hematological parameters in overall survival in HNSCC patients using neural network analysis. Retrospective analysis was performed on 106 previously untreated HNSCC patients. Several PET-derived parameters (SUVmax, SUVmean, TotalSUV, MTV, TLG, TLRmax, TLRmean, TLRTLG, and HI) for primary tumor and lymph node with highest activity were assessed. Additionally, hematological parameters (LEU, LEU%, NEU, NEU%, MON, MON%, PLT, PLT%, NRL, and LMR) were also assessed. Patients were divided according to the diagnosis into the good and bad group. The data were evaluated using an artificial neural network (Neural Analyzer version 2.9.5) and conventional statistic. Statistically significant differences in PET-derived parameters in 5-year survival rate between group of patients with worse prognosis and good prognosis were shown in primary tumor SUVmax (10.0 vs. 7.7; p = 0.040), SUVmean (5.4 vs. 4.4; p = 0.047), MTV (23.2 vs. 14.5; p = 0.010), and TLG (155.0 vs. 87.5; p = 0.05), and mean liver TLG (27.8 vs. 30.4; p = 0.031), TLRmax (3.8 vs. 2.6; p = 0.019), TLRmean (2.8 vs. 1.9; p = 0.018), and in TLRTLG (5.6 vs. 2.3; p = 0.042). From hematological parameters, only LMR showed significant differences (2.5 vs. 3.2; p = 0.009). Final neural network showed that for ages above 60, primary tumors SUVmax, TotalSUV, MTV, TLG, TLRmax, and TLRmean over (9.7, 2255, 20.6, 145, 3.6, 2.6, respectively) are associated with worse survival. Our study shows that the neural network could serve as a supplement to PET-derived parameters and is helpful in finding prognostic parameters for overall survival in HNSCC.
ABSTRACT
Social and health care equity and justice should be prioritized by the mantra of medicine, first do no harm. Despite highly motivated national and global health strategies, there remains significant health care inequity. Intrinsic and extrinsic factors, including a number of biases, are key drivers of ongoing health inequity including equity of access and opportunity for nuclear medicine and radiology services. There is a substantial gap in the global practice of nuclear medicine in particular, but also radiology, between developed health economies and those considered developing or undeveloped. At a local level, even in developed health economies, there can be a significant disparity between health services, including medical imaging, between communities based on socioeconomic, cultural or geographic differences. Artificial intelligence (AI) has the potential to either widen the health inequity divide or substantially reduce it. Distributed generally, AI technology could be used to overcome geographic boundaries to health care, thus bringing general and specialist care into underserved communities. However, should AI technology be limited to localities already enjoying ample healthcare access and direct access to health infrastructure, like radiology and nuclear medicine, it could then accentuate the gap. There are a number of challenges across the AI pipeline that need careful attention to ensure beneficence over maleficence. Fully realized, AI augmented health care could be crafted as an integral part of the broader strategy convergence on local, national and global health equity. The applications of AI in nuclear medicine and radiology could emerge as a powerful tool in social and health equity.
Subject(s)
Artificial Intelligence , Radiology , Diagnostic Imaging , HumansABSTRACT
The First Nations peoples in the United States, Canada, Australia, and around the world are substantially disadvantaged by colonialization, including health inequity. For nuclear medicine, the cultural competence of the staff and cultural proficiency of the institution are important minimum expectations. This minimum can be achieved through a scaffold of Indigenous cultural training and immersion programs that allow the nuclear medicine department to be a culturally safe environment for Indigenous patients. Development of such programs requires careful planning and inclusivity of Indigenous people as the key stakeholders but, done appropriately, can positively drive the Indigenous equity pipeline. Central to this undertaking is an understanding of Indigenous ways of learning and the nexus of these ways of learning and learning taxonomies. There remain substantial gaps between the most culturally insightful and the least culturally insightful (individuals and institutions)-gaps that can be addressed, in part, by rich immersive professional development activities in nuclear medicine targeting cultural proficiency and creating culturally safe clinical environments. The opportunity lies before us to provide leadership in nation building and in yindyamarra winhanganha: living respectfully while creating a world worth living in.
Subject(s)
Cultural Competency , Australia , Canada , Humans , United StatesABSTRACT
Targeted molecular imaging with PET uses chemical ligands that are peptides specifically targeting a receptor of interest. Prostate-specific membrane antigen (PSMA) is substantially upregulated in prostate cancer but is also expressed in the neovascular tissue of several malignancies, including renal cell carcinoma (RCC). Radiolabeled peptide targets for PSMA may be helpful in detecting metastatic RCC lesions. We present a case of incidental detection of RCC metastatic disease with PSMA-targeted PET, and we explore potential use for deliberate evaluation of RCC with PSMA-targeted tracers.
Subject(s)
Carcinoma, Renal Cell , Kidney Neoplasms , Prostatic Neoplasms , Carcinoma, Renal Cell/diagnostic imaging , Carcinoma, Renal Cell/metabolism , Carcinoma, Renal Cell/pathology , Humans , Kidney Neoplasms/diagnostic imaging , Kidney Neoplasms/pathology , Ligands , Lysine/chemistry , Male , Positron Emission Tomography Computed Tomography/methods , Prostatic Neoplasms/diagnostic imaging , Prostatic Neoplasms/pathology , Urea/chemistryABSTRACT
Background: While normal ranges for 99mTc thyroid percentage uptake vary, the seemingly intuitive evaluation of thyroid function does not reflect the complexity of thyroid pathology and biochemical status. The emergence of artificial intelligence (AI) in nuclear medicine has driven problem solving associated with logic and reasoning that warrant re-examination of established benchmarks in thyroid functional assessment. Methods: There were 123 patients retrospectively analysed in the study sample comparing scintigraphic findings to grounded truth established through biochemistry status. Conventional statistical approaches were used in conjunction with an artificial neural network (ANN) to determine predictors of thyroid function from data features. A convolutional neural network (CNN) was also used to extract features from the input tensor (images). Results: Analysis was confounded by sub-clinical hyperthyroidism, primary hypothyroidism, sub-clinical hypothyroidism and T3 toxicosis. Binary accuracy for identifying hyperthyroidism was highest for thyroid uptake classification using a threshold of 4.5% (82.6%), followed by pooled physician 6interpretation with the aid of uptake values (82.3%). Visual evaluation without quantitative values reduced accuracy to 61.0% for pooled physician determinations and 61.4% classifying on the basis of thyroid gland intensity relative to salivary glands. The machine learning (ML) algorithm produced 84.6% accuracy, however, this included biochemistry features not available to the semantic analysis. The deep learning (DL) algorithm had an accuracy of 80.5% based on image inputs alone. Conclusion: Thyroid scintigraphy is useful in identifying hyperthyroid patients suitable for radioiodine therapy when using an appropriately validated cut-off for the patient population (4.5% in this population). ML ANN algorithms can be developed to improve accuracy as second readers systems when biochemistry results are available. DL CNN algorithms can be developed to improve accuracy in the absence of biochemistry results. ML and DL do not displace the role of the physician in thyroid scintigraphy but could be used as second reader systems to minimize errors and increase confidence.