ABSTRACT
Purpose: Use of mechanical imaging (MI) as complementary to digital mammography (DM), or in simultaneous digital breast tomosynthesis (DBT) and MI - DBTMI, has demonstrated the potential to increase the specificity of breast cancer screening and reduce unnecessary biopsies compared with DM. The aim of this study is to investigate the increase in the radiation dose due to the presence of an MI sensor during simultaneous image acquisition when automatic exposure control is used. Approach: A radiation dose study was conducted on clinically available breast imaging systems with and without an MI sensor present. Our estimations were based on three approaches. In the first approach, exposure values were compared in paired clinical DBT and DBTMI acquisitions in 97 women. In the second approach polymethyl methacrylate (PMMA) phantoms of various thicknesses were used, and the average glandular dose (AGD) values were compared. Finally, a rectangular PMMA phantom with a 45 mm thickness was used, and the AGD values were estimated based on air kerma measurements with an electronic dosemeter. Results: The relative increase in exposure estimated from digital imaging and communications in medicine headers when using an MI sensor in clinical DBTMI was 11.9 % ± 10.4 . For the phantom measurements of various thicknesses of PMMA, the relative increases in the AGD for DM and DBT measurements were, on average, 10.7 % ± 3.1 and 11.4 % ± 3.0 , respectively. The relative increase in the AGD using the electronic dosemeter was 11.2 % ± < 0.001 in DM and 12.2 % ± < 0.001 in DBT. The average difference in dose between the methods was 11.5 % ± 3.3 . Conclusions: Our measurements suggest that the use of simultaneous breast radiography and MI increases the AGD by an average of 11.5 % ± 3.3 . The increase in dose is within the acceptable values for mammography screening recommended by European guidelines.
ABSTRACT
BACKGROUND: Phase-contrast computed tomography (CT) using high-brilliance, synchrotron-generated x-rays enable three-dimensional (3D) visualization of microanatomical structures within biological specimens, offering exceptionally high-contrast images of soft tissues. Traditional methods for phase-contrast CT; however, necessitate a gap between the subject and the x-ray camera, compromising spatial resolution due to penumbral blurring. Our newly developed technique, Superimposed Wavefront Imaging of Diffraction-enhanced x-rays (SWIDeX), leverages a Laue-case Si angle analyzer affixed to a scintillator to convert x-rays to visible light, capturing second-order differential phase contrast images and effectively eliminating the distance to the x-ray camera. This innovation achieves superior spatial resolution over conventional methods. PURPOSE: In this paper, the imaging principle and CT reconstruction algorithm based on SWIDeX are presented in detail and compared with conventional analyzer-based imaging (ABI). It also shows the physical setup of SWIDeX that provides the resolution preserving second-order differential images for reconstruction. We compare the spatial resolution and the sensitivity of SWIDeX to conventional ABI. METHODS: To demonstrate high-spatial resolution achievable by SWIDeX, the internal structures of four human tissues-ductal carcinoma in situ, normal stomach, normal pancreas, and intraductal papillary mucinous neoplasm of the pancreas-were visualized using an imaging system configured at the Photon Factory's BL14B beamline under the High Energy Accelerator Research Organization (KEK). Each tissue was thinly sliced after imaging, stained with hematoxylin and eosin (H&E) for conventional microscope-based pathology. RESULTS: A comparison of SWIDeX-CT and pathological images visually demonstrates the effectiveness of SWIDeX-CT for biological tissue imaging. SWIDeX could generate clearer 3D images than existing analyzer-based phase-contrast methods and accurately delineate tissue structures, as validated against histopathological images. CONCLUSIONS: SWIDeX can visualize important 3D structures in biological soft tissue with high spatial resolution and can be an important tool for providing information between the disparate scales of clinical and pathological imaging.
ABSTRACT
INTRODUCTION: Radiological imaging has played an important role in diagnostic medicine for over a century, though it is known to contribute to dermatological conditions, cataracts, and cancer. The associated risk of harm has led to the introduction of protective regulations around the world. Present-day NHS clinicians are increasingly requesting and relying on diagnostic imaging. Knowledge surrounding the radiation doses of common radiological investigations and the associated risks is imperative, and on a global level has been found to be inadequate. Consequently, there is a need for the formal inclusion of teaching within training programmes. AIMS/OBJECTIVES: This prospective audit aims to establish the knowledge of radiation doses and risks of common radiological investigations of both medical students and referrers within four NHS Health Boards based in the North of Scotland. It also seeks to establish prior teaching and the preference for further educational interventions. AUDIT STANDARD: Referrers should have adequate knowledge of radiation doses and the risks associated with common radiological investigations. AUDIT TARGET: The standard should be achieved by 90% of referrers. METHODS: A 19-question online survey was devised to include subjective and objective questions on ionising radiation awareness, education preference, and respondent demographics, based on RCR (Royal College of Radiologists) audit criteria and previous studies. Data collection was conducted between the 22/02/23 to the 22/03/2023 and the questionnaire was distributed to senior medical students and radiological referrers of different grades within NHS Grampian, NHS Highland, NHS Shetland, and NHS Orkney. A descriptive analysis of the data was undertaken using Microsoft Excel Version 16.71. RESULTS: Two hundred eight questionnaires were completed. 22.11% (n = 46) of the sample population had received no prior teaching on the topic of ionising radiation. Over half of the respondents (51.92%, n = 108) rated the importance of radiation risks as either important or extremely important, with 69.71% (n = 145) of participants rating their perceived knowledge as limited or average. Most correctly identified that a CT scan (n = 203), PET-CT scan (n = 199) and a chest x-ray (n = 196) exposed patients to ionising radiation. A small proportion of the participants incorrectly thought that an MRI scan (n = 21) and an ultrasound scan (n = 2) involved ionising radiation. The results obtained failed to meet the RCR audit target, which states that 90% of doctors should be aware of common radiological doses. It was observed that only 17.79% (n = 37) of survey respondents scored over 50% in the knowledge assessment, with the median knowledge score of the whole cohort being 2.5 out of 9 (27.78%). Respondents who had prior teaching on the topic performed better those who had no prior teaching, with average scores of 3.19 (35.44%) and 2.04 (22.67%) respectively. Senior clinicians performed better when compared to junior clinicians and medical students. CONCLUSION & FUTURE RECOMMENDATIONS: This audit found that the knowledge of radiation risks within the North of Scotland in the selected sample population was insufficient across all levels of the clinical team. Further, continuous education around the topic and future audit opportunities may help to optimise knowledge and training.
Subject(s)
Health Knowledge, Attitudes, Practice , Radiation Dosage , Students, Medical , Humans , Scotland , Prospective Studies , Female , Referral and Consultation , Male , Surveys and Questionnaires , Adult , Medical AuditABSTRACT
Purpose: Breast cancer is the leading cause of cancer mortality among women. Radiotherapy can reduce recurrence and prolong survival of patients accepting breast-conserving surgery (BCS). This study aims to compare acute skin reactions in patients receiving hypofractionated versus conventional radiotherapy at a single institution and to summarize the relevant influencing factors. Methods: This study analyzed 152 patients who underwent either hypofractionated or conventional whole-breast irradiation (WBI) after BCS. Acute skin toxicity was assessed according to the Radiation Therapy Oncology Group (RTOG) criteria. Predictive factors for acute skin toxicity were identified using multivariate analysis and visualized using a forest spot. Results: Grade 0 reactions occurred in 75.34% vs 70.89%, grade 1 in 16.44% vs 15.19%, grade 2 in 8.22% vs 12.66%, and grade 3 in 0% vs 1.27% of patients receiving hypofractionated and conventional WBI, respectively. There was no statistically significant difference in acute skin reaction in patients treated with hypofractionated radiation compared with conventional radiation (P = 0.62). Multivariate analysis revealed that metastatic lymph nodes (P = 0.021), whole-breast planning target volume (PTV-WB) (P < 0.001), and tumor bed planning target volume (PTV-TB) (P = 0.002) were significantly correlated with higher rates of acute skin toxicity. Conclusion: Hypofractionated WBI demonstrated similar acute skin adverse reactions compared to conventional WBI. These findings indicate that hypofractionated radiotherapy offers comparable tolerance, equivalent curative effect, convenience, and economic benefits, supporting its clinical promotion.
ABSTRACT
BACKGROUND: Barrier films or dressings were reported to be effective in preventing radiation dermatitis (RD) in breast cancer patients, but their comparative efficacy is unknown. METHODS: A systematic literature search was performed on Embase, MEDLINE and Cochrane CENTRAL Registry of Clinical Trials from inception to October 20, 2023. Randomised controlled trials (RCTs) comparing barrier films or dressings to the standard of care (SOC) or other interventions were included. We estimated summary odds ratios and mean differences using network meta-analysis with random effects. This study was registered with PROSPERO (ID: CRD42023475021). RESULTS: Fourteen RCTs met inclusion criteria. Six interventions were analysed: 3M™ Moisturizing Double Barrier Cream (MDBC), 3M™ No Sting Barrier Film (BF), Hydrofilm® (HF), Mepitel® Film (MF), Silver Leaf Nylon Dressing and StrataXRT®. HF, MF and StrataXRT® reduced the incidence of moist desquamation compared to SOC (HF: OR = 0.08; p = 0.02; MF: OR = 0.31 p < 0.01; StrataXRT®: OR = 0.22, p = 0.04). The ranking of agents from most to least effective in preventing moist desquamation according to P-scores was HF (92.5%), MF (78.5%), StrataXRT® (70.1%), BF (46.4%), Silver Leaf Nylon Dressing (24.9%), MDBC (22.9%) and SOC (14.7%). Only four RCTs on HF and MF included patient-reported outcome (PRO) assessments that allowed pooling for analysis. HF and MF were more effective in reducing pain, itchiness and burning sensation compared to SOC (p < 0.01 for all symptoms). CONCLUSION: HF and MF were effective in preventing RD in breast cancer. Future RCTs should compare these interventions to effective cream preparations, such as topical corticosteroids.
ABSTRACT
PURPOSE: To report a single-institutional experience with hypofractionated radiation therapy alone for human papillomavirus (HPV)-positive oropharyngeal cancer. METHODS AND MATERIALS: A total of 101 consecutive patients were treated by radiation therapy alone using a regimen of 66 Gy in 30 fractions (60 patients) or 70 Gy in 33 fractions (41 patients) for newly diagnosed p16-positive squamous cell carcinoma of the oropharynx. Sixty-seven patients (67%) were never smokers. RESULTS: The 3-year actuarial rates of overall survival, local-regional control, and progression-free survival were 94%, 93%, and 89%, respectively. Among never-smokers, the 3-year rates of overall survival and local-regional control were 98% and 100%, respectively. The grade 3+ acute toxicity rate was 21%, with the most commonly observed side effects related to mucositis. CONCLUSION: Hypofractionated radiation alone resulted in excellent outcomes for patients with HPV-positive oropharyngeal cancer. A prospective clinical trial investigating this modality in the setting of de-escalation is currently underway.
ABSTRACT
Although stereotactic body radiotherapy (SBRT) is a curative treatment option for stage I non-small cell lung cancer (NSCLC), limited data are available regarding chest wall (CW) toxicities during an extended follow-up of over 10 years. We report an unusual case of a bone tumor-like CW mass lesion with pathological rib fractures observed 13 years after SBRT for peripheral lung cancer. Despite the initial suspicion of radiation-induced sarcoma, a subsequent incisional biopsy revealed no evidence of malignancy, and a definitive diagnosis of osteonecrosis was made. Thus, long-term observation of over 10 years is required to identify late chronic complications following SBRT.
ABSTRACT
Pulse radiolysis was employed to investigate fundamental radiation chemical reactions, which are essential in the radiation protection of DNA. Two positively charged peptides (PCPs), histidine-tyrosine-histidine (His-Tyr-His) and lysine-tyrosine-lysine (Lys-Tyr-Lys), as well as the amino acids that constitute them, were involved. The reaction rate constants for tyrosine (Tyr), histidine (His), lysine (Lys), His-Tyr-His, and Lys-Tyr-Lys with OH radicals (â¢OH) were (1.6 ± 0.3) × 1010, (9.0 ± 0.9) × 109, (1.4 ± 0.3) × 109, (1.8 ± 0.1) × 1010, and (1.0 ± 0.1) × 1010 M-1s-1, respectively, indicating that formation of peptide bond can affect the reaction of amino acids with â¢OH. Observed transient absorption spectra indicated a shielding effect of the His or Lys residues at both ends of the PCPs on the centrally located Tyr. The measurement of chemical repair capabilities using deoxyguanosine monophosphate (dGMP) as a model for DNA demonstrated that the reaction rate constants of Tyr, His-Tyr-His, and Lys-Tyr-Lys with dGMP radicals were (2.2 ± 0.5) × 108, (2.3 ± 0.1) × 108, and (3.3 ± 0.4) × 108 M-1s-1, respectively, implying that the presence of a positive charge may enhance the chemical repair process.
ABSTRACT
BACKGROUND: We report on the characterization and introduction of a novel prognostic score for patients undergoing stereotactic body radiotherapy (SBRT) for the treatment of single and multiple pulmonary metastases (PMs) derived from head and neck cancer (HNC). METHODS: In this retrospective study, we examined selected factors associated with progression-free survival (PFS) and overall survival (OS) among 59 patients with HNC treated with SBRT for a total of 118 PMs, between 2009 and 2023. Factors related to survival were included in the prognostic scoring system. RESULTS: Prognostic factors including histology, age, number of metastases, and performance status at first SBRT were weighted differently depending on the strength of correlation to PFS and OS. Total prognostic scores (HAMP) ranged from 13 to 24 points, with a cut-off total score of ≤18 scoring points for patients in a high-risk (HR) subcohort, and of ≥19 scoring points for patients in a low-risk group (LR). Median PFS (23.8 vs. 5.5 months, p < 0.001) and OS (61.3 vs. 16.4 months, p < 0.001) were significantly longer in the low-risk group compared to the high-risk group. CONCLUSION: The HAMP score might be a convenient tool to facilitate individualized treatment decisions and appropriate follow-up. The accuracy and reliability of the score requires further evaluation in prospective studies.
ABSTRACT
The mechanisms that work alongside BRAFV600E oncogene in melanoma development, in addition to ultraviolet (UV) radiation (UVR), are of great interest. Analysis of human melanoma tumors [data from The Cancer Genome Atlas (TCGA)] revealed that 50% or more of the samples expressed no or low amounts of serine/threonine protein kinase STK11 (also known as LKB1) protein. Here, we report that, in a mouse model, concomitant neonatal BrafV600E activation and Lkb1 tumor suppressor ablation in melanocytes led to full melanoma development. A single postnatal dose of UVB radiation had no effect on melanoma onset in Lkb1-depleted mice compared with BrafV600E-irradiated mice, but increased tumor multiplicity. In concordance with these findings and previous reports, Lkb1-null irradiated mice exhibited deficient DNA damage repair (DDR). Histologically, tumors lacking Lkb1 were enriched in neural-like tumor morphology. Genetic profiling and gene set enrichment analyses of tumor sample mutated genes indicated that loss of Lkb1 promoted the selection of altered genes associated with neural differentiation processes. Thus, these results suggest that the loss of Lkb1 cooperates with BrafV600E and UVR, impairing the DDR and increasing melanoma multiplicity and neural-like dedifferentiation.
ABSTRACT
BACKGROUND: Treatment delivery safety and accuracy are essential to control the disease and protect healthy tissues in radiation therapy. For usual treatment, a phantom-based patient specific quality assurance (PSQA) is performed to verify the delivery prior to the treatment. The emergence of adaptive radiation therapy (ART) adds new complexities to PSQA. In fact, organ at risks and target volume re-contouring as well as plan re-optimization and treatment delivery are performed with the patient immobilized on the treatment couch, making phantom-based pretreatment PSQA impractical. In this case, phantomless PSQA tools based on multileaf collimator (MLC) leaf open times (LOTs) verifications provide alternative approaches for the Radixact® treatment units. However, their validity is compromised by the lack of independent and reliable methods for calculating the LOT performed by the MLC during deliveries. PURPOSE: To provide independent and reliable methods of LOT calculation for the Radixact® treatment units. METHODS: Two methods for calculating the LOTs performed by the MLC during deliveries have been implemented. The first method uses the signal recorded by the build-in detector and the second method uses the signal recorded by optical sensors mounted on the MLC. To calibrate the methods to the ground truth, in-phantom ionization chamber LOT measurements have been conducted on a Radixact® treatment unit. The methods were validated by comparing LOT calculations with in-phantom ionization chamber LOT measurements performed on two Radixact® treatment units. RESULTS: The study shows a good agreement between the two LOT calculation methods and the in-phantom ionization chamber measurements. There are no notable differences between the two methods and the same results were observed on the different treatment units. CONCLUSIONS: The two implemented methods have the potential to be part of a PSQA solution for ART in tomotherapy.
ABSTRACT
Introduction: Organ dose distribution calculation in radiotherapy and knowledge about its side effects in cancer etiology is the most concern for medical physicists. Calculation of organ dose distribution for breast cancer treatment plans with Monte Carlo (MC) simulation is the main goal of this study. Materials and Methods: Elekta Precise linear accelerator (LINAC) photon mode was simulated and verified using the GEANT4 application for tomographic emission. Eight different radiotherapy treatment plans on RANDO's phantom left breast were produced with the ISOgray treatment planning system (TPS). The simulated plans verified photon dose distribution in clinical tumor volume (CTV) with TPS dose volume histogram (DVH) and gamma index tools. To verify photon dose distribution in out-of-field organs, the point dose measurement results were compared with the same point doses in the MC simulation. Eventually, the DVHs for out-of-field organs that were extracted from the TPS and MC simulation were compared. Results: Based on the implementation of gamma index tools with 2%/2 mm criteria, the simulated LINAC output demonstrated high agreement with the experimental measurements. Plan simulation for in-field and out-of-field organs had an acceptable agreement with TPS and experimental measurement, respectively. There was a difference between DVHs extracted from the TPS and MC simulation for out-of-field organs in low-dose parts. This difference is due to the inability of the TPS to calculate dose distribution in out-of-field organs. Conclusion and Discussion: Based on the results, it was concluded that the treatment plans with the MC simulation have a high accuracy for the calculation of out-of-field dose distribution and could play a significant role in evaluating the important role of dose distribution for second primary cancer estimation.
ABSTRACT
This dataset is comprised of a library of atomistic structure files and corresponding X-ray diffraction (XRD) profiles and vibrational density of states (VDoS) profiles for bulk single crystal silicon (Si), gold (Au), magnesium (Mg), and iron (Fe) with and without disorder introduced into the atomic structure and with and without mechanical loading. Included with the atomistic structure files are descriptor files that measure the stress state, phase fractions, and dislocation content of the microstructures. All data was generated via molecular dynamics or molecular statics simulations using the Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) code. This dataset can inform the understanding of how local or global changes to a materials microstructure can alter their spectroscopic and diffraction behavior across a variety of initial structure types (cubic diamond, face-centered cubic (FCC), hexagonal close-packed (HCP), and body-centered cubic (BCC) for Si, Au, Mg, and Fe, respectively) and overlapping changes to the microstructure (i.e., both disorder insertion and mechanical loading).
ABSTRACT
Background and purpose: Integrating simultaneous interstitial hyperthermia in high-dose-rate brachytherapy treatments (HDR-BT) is expected to lead to enhanced therapeutic effect. However, there is currently no device available for such an integration. In this study, we presented and validated the thermobrachytherapy (TBT) preclinical prototype system that is able to seamlessly integrate into the HDR-BT workflow. Materials and methods: The TBT system consisted of an advanced radiofrequency power delivery and control system, dual-function interstitial applicators, and integrated connection and impedance matching system. The efficiency and minimum heating ability of the system was calculated performing calorimetric experiments. The effective-heating-length and heating pattern was evaluated using single-applicator split phantom experiments. The heating independence between applicators, the ability of the system to adaptable and predictable temperature steering was evaluated using multi-applicator split phantom experiments. Results: The system satisfied interstitial hyperthermia requirements. It demonstrated 50 % efficiency and ability to reach 6 °C temperature increase in 6 min. Effective-heating-length of the applicator was 43.7 mm, following the initial design. Heating pattern interference between applicators was lower than recommended. The system showed its ability to generate diverse heating patterns by adjusting the phase and amplitude settings of each electrode, aligning well with simulations (minimum agreement of 88 %). Conclusions: The TBT preclinical prototype system complied with IHT requirements, and agreed well with design criteria and simulations, hence performing as expected. The preclinical prototype TBT system can now be scaled to an in-vivo validation prototype, including an adaptable impedance matching solution, appropriate number of channels, and ensuring biocompatibility and regulatory compliance.
ABSTRACT
The treatment of thoracic tumors with ionizing radiation can cause radiation-induced lung injury (RILI), which includes radiation pneumonitis and radiation-induced pulmonary fibrosis. Preventing RILI is crucial for controlling tumor growth and improving quality of life. However, the serious adverse effects of traditional RILI treatment methods remain a major obstacle, necessitating the development of novel treatment options that are both safe and effective. This review summarizes the molecular mechanisms of RILI and explores novel treatment options, including natural compounds, gene therapy, nanomaterials, and mesenchymal stem cells. These recent experimental approaches show potential as effective prevention and treatment options for RILI in clinical practice.
ABSTRACT
Visiting electives provide an opportunity for medical students to engage with radiation oncology (RO) programs, likely influencing residency match outcomes. However, some student's out-of-pocket costs may be prohibitive, and in attempts to offset the financial burden of visiting electives, particularly for students underrepresented in medicine (URiM), some institutions offer scholarships. Here, we characterized the current domestic landscape of funded RO electives. Visiting electives were identified through the FREIDA and VSLO databases in April 2024. Funded elective availability and departmental characteristics were identified via internet search by two independent reviewers. Fisher's exact test was used to determine whether there was a difference in the distribution of scholarships across the US due to the small sample size. Ninety-two visiting electives were identified, with 40 programs offering URiM elective scholarships (43.5%). Twelve (30%) were funded specifically by RO departments, and 28 (70%) were part of broader institutional URiM scholarship initiatives. The median stipend provided was $2000 (IQR $500), range $1000-$5000. Analysis of scholarships by US census division and metro area revealed unequal distribution. Electives in New England, Mountain, and East North Central divisions had higher funding proportion compared to electives in the East South Central, West South Central, and Middle Atlantic divisions. Only 1/9 electives in New York City were funded compared with 4/6 in Los Angeles. Departments with funded electives had more faculty physicians and medical residents. In our review of the 2024 landscape, over 40% of RO electives offer financial support. However, we identified geographical disparities in the distribution of scholarships, highlighting the need for interventions to address unequal access to a wide array of training programs. Our study represents a valuable resource for students interested in RO and highlights the continued need to positively contribute to increasing diversity in the field. Future work exploring the impact of funded electives is needed.
ABSTRACT
BACKGROUND: In recent years, evidence has accumulated that a second method of conserving the breast from cancer with re-irradiation as part of treatment may be feasible and safe. Many oncologists are skeptical of breast re-irradiation due to concerns about late complications, so access to quantitative data on the prevalence of breast re-irradiation complications is very important. In this meta-analysis, we determine the prevalence of complications in normal tissue after breast re-irradiation. MATERIALS AND METHODS: A search was done to recognize qualified studies using EMBASE, MEDLINE, PUBMED, Google Scholar, and Cochrane Collaboration Library electronic databases from 2000 to 2023. In total, ten primary studies were applied in this meta-analysis to estimate the prevalence of complications of disorders, skin fibrosis, and chest pain. Heterogeneity was investigated using the I2 index and the meta-regression to evaluate variables suspected of causing heterogeneity. Statistical analysis and synthesis were performed using Stata 17. RESULTS: The average dose received by patients who underwent radiation therapy in two stages was 100.32 Gy, and in these patients, the prevalence of skin fibrosis and disorders was 47% (95% CI 71-22%; I2 = 96.76%, P < 0.001) and the prevalence of chest pain was 35% (95% CI 68-8%; I2 = 98.13%, P < 0.001). CONCLUSIONS: There is little clinical information about the incidence of complications in breast re-irradiation therapy. This meta-analysis presents the prevalence of complications after breast re-irradiation to help radiation oncologists and physicists make better decisions.
ABSTRACT
Aim: This article aims to identify risk factors for severe radiation-induced oral mucositis (RIOM) in head and neck cancer (HNC) patients. In addition, we intend to establish a predictive model in patients undergoing intensity-modulated radiotherapy. Patients & methods: In this retrospective study, several HNC patients (n = 179) treated at Zhejiang Provincial People's Hospital from January 2019 to June 2023 were considered. The recruited subjects were divided into modeling and validation groups. The experimental data on clinical characteristics and treatment were collected and analyzed to identify predictive factors for severe RIOM based on the logistic regression approach. Results: The results indicated that severe RIOM occurred in 55.3% of patients. Accordingly, significant predictors included smoking history, diabetes, concurrent chemotherapy, cumulative radiation dose and weight loss of ≥5% in relative to admission weight. A nomogram based on these factors was validated, showing excellent predictive accuracy. Conclusion: In summary, the predictive model could effectively identify high-risk patients for severe RIOM, enabling the design of targeted interventions and improving patient management during radiotherapy.
[Box: see text].
ABSTRACT
PURPOSE: In the rapidly expanding field of artificial intelligence (AI) there is a wealth of literature detailing the myriad applications of AI, particularly in the realm of deep learning. However, a review that elucidates the technical principles of deep learning as relevant to radiation oncology in an easily understandable manner is still notably lacking. This paper aims to fill this gap by providing a comprehensive guide to the principles of deep learning that is specifically tailored toward radiation oncology. METHODS: In light of the extensive variety of AI methodologies, this review selectively concentrates on the specific domain of deep learning. It emphasizes the principal categories of deep learning models and delineates the methodologies for training these models effectively. RESULTS: This review initially delineates the distinctions between AI and deep learning as well as between supervised and unsupervised learning. Subsequently, it elucidates the fundamental principles of major deep learning models, encompassing multilayer perceptrons (MLPs), convolutional neural networks (CNNs), recurrent neural networks (RNNs), transformers, generative adversarial networks (GANs), diffusion-based generative models, and reinforcement learning. For each category, it presents representative networks alongside their specific applications in radiation oncology. Moreover, the review outlines critical factors essential for training deep learning models, such as data preprocessing, loss functions, optimizers, and other pivotal training parameters including learning rate and batch size. CONCLUSION: This review provides a comprehensive overview of deep learning principles tailored toward radiation oncology. It aims to enhance the understanding of AI-based research and software applications, thereby bridging the gap between complex technological concepts and clinical practice in radiation oncology.