Chen, Quan

Background/Objectives: Lung cancer is a devastating disease with the highest mortality rate among cancer types. Over 60% of non-small cell lung cancer (NSCLC) patients, accounting for 87% of lung cancer diagnoses, require radiation therapy. Rapid treatment initiation significantly increases the patient’s survival rate and reduces the mortality rate. Accurate tumor segmentation is a critical step in diagnosing and treating NSCLC. Manual segmentation is time- and labor-consuming and causes delays in treatment initiation. Although many lung nodule detection methods, including deep learning-based models, have been proposed. Most of these methods still have a long-standing problem of high false positives (FPs). Methods: Here, we developed an electronic health record (EHR)-guided lung tumor auto-segmentation called EXACT-Net (EHR-enhanced eXACtitude in Tumor segmentation), where the extracted information from EHRs using a pre-trained large language model (LLM) was used to remove the FPs and keep the TP nodules only. Results: The auto-segmentation model was trained on NSCLC patients’ computed tomography (CT), and the pre-trained LLM was used with the zero-shot learning approach. Our approach resulted in a 250% boost in successful nodule detection using the data from ten NSCLC patients treated in our institution. Conclusions: We demonstrated that combining vision-language information in EXACT-Net multi-modal AI framework greatly enhances the performance of vision only models, paving the road to multimodal AI framework for medical image processing.

Abstract Innovation in medical imaging artificial intelligence (AI)/machine learning (ML) demands extensive data collection, algorithmic advancements, and rigorous performance assessments encompassing aspects such as generalizability, uncertainty, bias, fairness, trustworthiness, and interpretability. Achieving widespread integration of AI/ML algorithms into diverse clinical tasks will demand a steadfast commitment to overcoming issues in model design, development, and performance assessment. The complexities of AI/ML clinical translation present substantial challenges, requiring engagement with relevant stakeholders, assessment of cost-effectiveness for user and patient benefit, timely dissemination of information relevant to robust functioning throughout the AI/ML lifecycle, consideration of regulatory compliance, and feedback loops for real-world performance evidence. This commentary addresses several hurdles for the development and adoption of AI/ML technologies in medical imaging. Comprehensive attention to these underlying and often subtle factors is critical not only for tackling the challenges but also for exploring novel opportunities for the advancement of AI in radiology.

Abstract The adoption of artificial intelligence (AI) tools in medicine poses challenges to existing clinical workflows. This commentary discusses the necessity of context-specific quality assurance (QA), emphasizing the need for robust QA measures with quality control (QC) procedures that encompass (1) acceptance testing (AT) before clinical use, (2) continuous QC monitoring, and (3) adequate user training. The discussion also covers essential components of AT and QA, illustrated with real-world examples. We also highlight what we see as the shared responsibility of manufacturers or vendors, regulators, healthcare systems, medical physicists, and clinicians to enact appropriate testing and oversight to ensure a safe and equitable transformation of medicine through AI.

AbstractBackgroundInaccurate manual organ delineation is one of the high‐risk failure modes in radiation treatment. Numerous automated contour quality assurance (QA) systems have been developed to assess contour acceptability; however, manual inspection of flagged cases is a time‐consuming and challenging process, and can lead to users overlooking the exact error location.PurposeOur aim is to develop and validate a contour QA system that can effectively detect and visualize subregional contour errors, both qualitatively and quantitatively.Methods/MaterialsA novel contour subregion error detection (CSED) system was developed using subregional surface distance discrepancies between manual and deep learning auto‐segmentation (DLAS) contours. A validation study was conducted using a head and neck public dataset containing 339 cases and evaluated according to knowledge‐based pass criteria derived from a clinical training dataset of 60 cases. A blind qualitative evaluation was conducted, comparing the results from the CSED system with manual labels. Subsequently, the CSED‐flagged cases were re‐examined by a radiation oncologist.ResultsThe CSED system could visualize the diverse types of subregional contour errors qualitatively and quantitatively. In the validation dataset, the CSED system resulted in true positive rates (TPR) of 0.814, 0.800, and 0.771; false positive rates (FPR) of 0.310, 0.267, and 0.298; and accuracies of 0.735, 0.759, and 0.730, for brainstem and left and right parotid contours, respectively. The CSED‐assisted manual review caught 13 brainstem, 19 left parotid, and 21 right parotid contour errors missed by conventional human review. The TPR/FPR/accuracy of the CSED‐assisted manual review improved to 0.836/0.253/0.784, 0.831/0.171/0.830, and 0.808/0.193/0.807 for each structure, respectively. Further, the time savings achieved through CSED‐assisted review improved by 75%, with the time for review taking 24.81 ± 12.84, 26.75 ± 10.41, and 28.71 ± 13.72 s for each structure, respectively.ConclusionsThe CSED system enables qualitative and quantitative detection, localization, and visualization of manual segmentation subregional errors utilizing DLAS contours as references. The use of this system has been shown to help reduce the risk of high‐risk failure modes resulting from inaccurate organ segmentation.

AbstractBackgroundDeep learning auto‐segmentation (DLAS) models have been adopted in the clinic; however, they suffer from performance deterioration owing to the clinical practice variability. Some commercial DLAS software provide an incremental retraining function that enables users to train a custom model using their institutional data to account for clinical practice variability.PurposeThis study was performed to evaluate and implement the commercial DLAS software with the incremental retraining function for definitive treatment of patients with prostate cancer in a multi‐user environment.MethodsCT‐based target organs and organs‐at‐risk (OAR) delineation of 215 prostate cancer patients were utilized. The performance of three commercial DLAS software built‐in models was validated with 20 patients. A retrained custom model was developed using 100 patients and evaluated on the remaining data (n = 115). Dice similarity coefficient (DSC), Hausdorff distance (HD), mean surface distance (MSD), and surface DSC (SDSC) were utilized for quantitative evaluation. A multi‐rater qualitative evaluation was blindly performed with a five‐level scale. Visual inspection was performed in consensus and non‐consensus unacceptable cases to identify the failure modes.ResultsThree commercial DLAS vendor built‐in models achieved sub‐optimal performance in 20 patients. The retrained custom model had a mean DSC of 0.82 for prostate, 0.48 for seminal vesicles (SV), and 0.92 for rectum, respectively. This represents a significant improvement over the built‐in model with DSC of 0.73, 0.37, and 0.81 for the corresponding structures. Compared to the acceptance rate of 96.5% and consensus unacceptable rate (i.e., both reviewers rated as unacceptable) of 3.5% achieved by manual contours, the custom model achieved a 91.3% acceptance rate and 8.7% consensus unacceptable rate. The failure modes of retrained custom model were attributed to the following: cystogram (n = 2), hip prosthesis (n = 2), low dose rate brachytherapy seeds (n = 2), air in endorectal balloon(n = 1), non‐iodinated spacer (n = 2), and giant bladder(n = 1).ConclusionThe commercial DLAS software with the incremental retraining function was validated and clinically adopted for prostate patients in a multi‐user environment. AI‐based auto‐delineation of the prostate and OARs is shown to achieve improved physician acceptance, overall clinical utility, and accuracy.

AbstractSince the publication of AAPM Task Group (TG) 148 on quality assurance (QA) for helical tomotherapy, there have been many new developments on the tomotherapy platform involving treatment delivery, on‐board imaging options, motion management, and treatment planning systems (TPSs). In response to a need for guidance on quality control (QC) and QA for these technologies, the AAPM Therapy Physics Committee commissioned TG 306 to review these changes and make recommendations related to these technology updates. The specific objectives of this TG were (1) to update, as needed, recommendations on tolerance limits, frequencies and QC/QA testing methodology in TG 148, (2) address the commissioning and necessary QA checks, as a supplement to Medical Physics Practice Guidelines (MPPG) with respect to tomotherapy TPS and (3) to provide risk‐based recommendations on the new technology implemented clinically and treatment delivery workflow. Detailed recommendations on QA tests and their tolerance levels are provided for dynamic jaws, binary multileaf collimators, and Synchrony motion management. A subset of TPS commissioning and QA checks in MPPG 5.a. applicable to tomotherapy are recommended. In addition, failure mode and effects analysis has been conducted among TG members to obtain multi‐institutional analysis on tomotherapy‐related failure modes and their effect ranking.

Purpose: To evaluate the accuracy of deep-learning-based auto-segmentation of the superior constrictor, middle constrictor, inferior constrictor, and larynx in comparison with a traditional multi-atlas-based method. Methods and Materials: One hundred and five computed tomography image datasets from 83 head and neck cancer patients were retrospectively collected and the superior constrictor, middle constrictor, inferior constrictor, and larynx were analyzed for deep-learning versus multi-atlas-based segmentation. Eighty-three computed tomography images (40 diagnostic computed tomography and 43 planning computed tomography) were used for training the convolutional neural network, and for atlas-based model training. The remaining 22 computed tomography datasets were used for validation of the atlas-based auto-segmentation versus deep-learning-based auto-segmentation contours, both of which were compared with the corresponding manual contours. Quantitative measures included Dice similarity coefficient, recall, precision, Hausdorff distance, 95th percentile of Hausdorff distance, and mean surface distance. Dosimetric differences between the auto-generated contours and manual contours were evaluated. Subjective evaluation was obtained from 3 clinical observers to blindly score the autosegmented structures based on the percentage of slices that require manual modification. Results: The deep-learning-based auto-segmentation versus atlas-based auto-segmentation results were compared for the superior constrictor, middle constrictor, inferior constrictor, and larynx. The mean Dice similarity coefficient values for the 4 structures were 0.67, 0.60, 0.65, and 0.84 for deep-learning-based auto-segmentation, whereas atlas-based auto-segmentation has Dice similarity coefficient results at 0.45, 0.36, 0.50, and 0.70, respectively. The mean 95th percentile of Hausdorff distance (cm) for the 4 structures were 0.41, 0.57, 0.59, and 0.54 for deep-learning-based auto-segmentation, but 0.78, 0.95, 0.96, and 1.23 for atlas-based auto-segmentation results, respectively. Similar mean dose differences were obtained from the 2 sets of autosegmented contours compared to manual contours. The dose–volume discrepancies and the average modification rates were higher with the atlas-based auto-segmentation contours. Conclusion: Swallowing-related structures are more accurately generated with DL-based versus atlas-based segmentation when compared with manual contours.

AbstractPurposeImage guidance is used to improve the accuracy of radiation therapy delivery but results in increased dose to patients. This is of particular concern in children who need be treated per Pediatric Image Gently Protocols due to long‐term risks from radiation exposure. The purpose of this study is to design a deep neural network architecture and loss function for improving soft‐tissue contrast and preserving small anatomical features in ultra‐low‐dose cone‐beam CTs (CBCT) of head and neck cancer (HNC) imaging.MethodsA 2D compound U‐Net architecture (modified U‐Net++) with different depths was proposed to enhance the network capability of capturing small‐volume structures. A mask weighted loss function (Mask‐Loss) was applied to enhance soft‐tissue contrast. Fifty‐five paired CBCT and CT images of HNC patients were retrospectively collected for network training and testing. The output enhanced CBCT images from the present study were evaluated with quantitative metrics including mean absolute error (MAE), signal‐to‐noise ratio (SNR), and structural similarity (SSIM), and compared with those from the previously proposed network architectures (U‐Net and wide U‐Net) using MAE loss functions. A visual assessment of ten selected structures in the enhanced CBCT images of each patient was performed to evaluate image quality improvement, blindly scored by an experienced radiation oncologist specialized in HN cancer.ResultsAll the enhanced CBCT images showed reduced artifactual distortion and image noise. U‐Net++ outperformed the U‐Net and wide U‐Net in terms of MAE, contrast near structure boundaries, and small structures. The proposed Mask‐Loss improved image contrast and accuracy of the soft‐tissue regions. The enhanced CBCT images predicted by U‐Net++ and Mask‐Loss demonstrated improvement compared to the U‐Net in terms of average MAE (52.41 vs 42.85 HU), SNR (14.14 vs 15.07 dB), and SSIM (0.84 vs 0.87), respectively (, in all paired t‐tests). The visual assessment showed that the proposed U‐Net++ and Mask‐Loss significantly improved original CBCTs (), compared to the U‐Net and MAE loss.ConclusionsThe proposed network architecture and loss function effectively improved image quality in soft‐tissue contrast, organ boundary, and small structure preservation for ultra‐low‐dose CBCT following Image Gently Protocol. This method has potential to provide sufficient anatomical representation on the enhanced CBCT images for accurate treatment delivery and potentially fast online‐adaptive re‐planning for HN cancer patients.

PurposeAccurate radiation dosimetry in radiobiological experiments is crucial for preclinical research in advancement of cancer treatment. Vendors of cell irradiators often perform calibration for end‐users. However, calibration accuracy remains unclear due to missing detailed information on calibration equipment and procedures. In this study, we report our findings of a vender miscalibration of the radiation output and our investigation on the root cause of the discrepancy.MethodsIndependent calibration verification for a commercial preclinical orthovoltage irradiator was conducted. Initially, in the absence of ionization chambers calibrated at kV energy, radiochromic films (EBT3) was first calibrated at MV energy. Energy correction factors from literature were used to create an in‐house kV dosimetry system. The miscalibration identified with the in‐house kV EBT3 dosimetry was later confirmed by ADCL calibrated ionization chambers (Exradin A1SL and PTW 30013) at kV energy. Ionization chambers were suspended in‐air following TG‐61 recommendation for output calibration. To investigate the root cause of the miscalibration, additional measurements were performed with ionization chambers placed on the shelf. A validated Monte Carlo simulation code was also used to investigate the impact of placing the ionization chamber on the shelf instead of suspending it in air during the vendor‐performed calibration process.ResultsUp to a 6% dosimetry error was observed when comparing the vendor calibrated output of the preclinical irradiator with our independent calibration check. Further investigation showed incorrect setups in the vendor’s calibration procedure which may result in dose errors up to 11% from the backscatter of the shelf board during calibration, and up to 5% from omitting temperature and pressure corrections to ionization chamber readings.ConclusionOur study revealed large dose calibration errors caused by incorrect setup and the omission of temperature/pressure correction in the vendor’s calibration procedure. The findings also highlighted the importance of performing an independent check of the dose calibration for preclinical kV irradiators. More absolute dosimetry training is needed for both vendors and end users for establishing accurate absolute dosimetry.