Artificial intelligence in cancer imaging: Clinical challenges and applications.

Judgement, as one of the core tenets of medicine, relies upon the integration of multilayered data with nuanced decision making. Cancer offers a unique context for medical decisions given not only its variegated forms with evolution of disease but also the need to take into account the individual condition of patients, their ability to receive treatment, and their responses to treatment. Challenges remain in the accurate detection, characterization, and monitoring of cancers despite improved technologies. Radiographic assessment of disease most commonly relies upon visual evaluations, the interpretations of which may be augmented by advanced computational analyses. In particular, artificial intelligence (AI) promises to make great strides in the qualitative interpretation of cancer imaging by expert clinicians, including volumetric delineation of tumors over time, extrapolation of the tumor genotype and biological course from its radiographic phenotype, prediction of clinical outcome, and assessment of the impact of disease and treatment on adjacent organs. AI may automate processes in the initial interpretation of images and shift the clinical workflow of radiographic detection, management decisions on whether or not to administer an intervention, and subsequent observation to a yet to be envisioned paradigm. Here, the authors review the current state of AI as applied to medical imaging of cancer and describe advances in 4 tumor types (lung, brain, breast, and prostate) to illustrate how common clinical problems are being addressed. Although most studies evaluating AI applications in oncology to date have not been vigorously validated for reproducibility and generalizability, the results do highlight increasingly concerted efforts in pushing AI technology to clinical use and to impact future directions in cancer care.

Efficacy of Intralesional Candida Antigen Versus Measles, Mumps, and Rubella Vaccine Versus Topical Podophyllin in Treatment of Resistant Genital Warts.

BACKGROUND: No single treatment is ideal for genital warts with high rate of resistance using conventional modalities as topical podophyllin; however, several intralesional immunotherapies are being tested nowadays, with variable results. In this study, we compared the safety and efficacy of treating resistant and recurrent genital warts by 2 intralesional immunotherapies [Candida antigen and measles, mumps, and rubella (MMR) vaccine] and compared them with topical podophyllin. PATIENTS/METHODS: A total of 45 patients with resistant or recurrent genital warts were enrolled in this study. Size and number of warts were detected in each patient, patients were divided into 3 groups. Group A injected with intralesional Candida antigen. Group B with intralesional MMR vaccine. Group C were treated with topical 25% podophyllin. Patients received a session every 2 weeks for 3 treatment sessions. RESULTS: With regard to the reduction in size and number of all warts, the best response was obtained in Candida antigen group where 46.7% showed complete clearance and 40% showed partial response followed by MMR group and the last was the podophyllin group, with no significant difference between them. Complete clearance of mother warts was noticed in 86.7% of Candida group, 53.3% in MMR group, and last 40% in podophyllin group, with a significantly better response in the Candida group (P = .027). CONCLUSION: Both intralesional Candida antigen and MMR vaccine are simple, safe, and effective treatment options with comparable results and better response than topical podophyllin.

Evaluation of the anterior chamber angle by anterior segment optical coherence tomography after implantable phakic contact lens implantation in myopic eyes.

PURPOSE: To evaluate the changes in the angle of the AC and lens vault after IPCL implantation by AS-OCT in myopic patients. METHODS: This was a prospective observational study involving 30 myopic eyes implanted with IPCL. AS-OCT was used to evaluate lens vault and AC angle parameters including anterior chamber angle, angle opening distance and trabecular-iris space area (TISA) at 1, 3 and 6 months postoperatively. RESULTS: All 3 AC angle parameters were significantly reduced at the 1st postoperative month compared to preoperative values, but remained stable thereafter with no significant change at the 3rd or 6th postoperative months. The lens vault showed no significant change over the entire follow-up period. CONCLUSION: IPCL implantation is a safe method for correction of myopia with stable AC angle narrowing over the course of 6 months postoperatively as monitored using AS-OCT.

Erectile dysfunction in patients with first-episode psychosis.

Sexual dysfunction is more prevalent in psychotic patients than in the nonpsychotic population. The objective of this study was to identify correlations between serum prolactin levels, testosterone levels and erectile dysfunction in patients with first-episode psychosis (n = 40) compared to age-matched healthy controls (n = 40). All subjects underwent clinical evaluation, international index of erectile function (IIEF5) score assessment and measurement of serum prolactin and total testosterone levels. In first-episode psychotic patients, the IIEF-5 score and total testosterone levels were significantly lower, while serum prolactin levels were higher. We concluded that men with first-episode psychosis are at an increased risk for development of erectile dysfunction, and increased duration of untreated psychosis leads to a higher incidence of erectile dysfunction and hyperprolactinemia.

Deep learning for lung cancer prognostication: A retrospective multi-cohort radiomics study.

BACKGROUND: Non-small-cell lung cancer (NSCLC) patients often demonstrate varying clinical courses and outcomes, even within the same tumor stage. This study explores deep learning applications in medical imaging allowing for the automated quantification of radiographic characteristics and potentially improving patient stratification. METHODS AND FINDINGS: We performed an integrative analysis on 7 independent datasets across 5 institutions totaling 1,194 NSCLC patients (age median = 68.3 years [range 32.5-93.3], survival median = 1.7 years [range 0.0-11.7]). Using external validation in computed tomography (CT) data, we identified prognostic signatures using a 3D convolutional neural network (CNN) for patients treated with radiotherapy (n = 771, age median = 68.0 years [range 32.5-93.3], survival median = 1.3 years [range 0.0-11.7]). We then employed a transfer learning approach to achieve the same for surgery patients (n = 391, age median = 69.1 years [range 37.2-88.0], survival median = 3.1 years [range 0.0-8.8]). We found that the CNN predictions were significantly associated with 2-year overall survival from the start of respective treatment for radiotherapy (area under the receiver operating characteristic curve [AUC] = 0.70 [95% CI 0.63-0.78], p < 0.001) and surgery (AUC = 0.71 [95% CI 0.60-0.82], p < 0.001) patients. The CNN was also able to significantly stratify patients into low and high mortality risk groups in both the radiotherapy (p < 0.001) and surgery (p = 0.03) datasets. Additionally, the CNN was found to significantly outperform random forest models built on clinical parameters-including age, sex, and tumor node metastasis stage-as well as demonstrate high robustness against test-retest (intraclass correlation coefficient = 0.91) and inter-reader (Spearman's rank-order correlation = 0.88) variations. To gain a better understanding of the characteristics captured by the CNN, we identified regions with the most contribution towards predictions and highlighted the importance of tumor-surrounding tissue in patient stratification. We also present preliminary findings on the biological basis of the captured phenotypes as being linked to cell cycle and transcriptional processes. Limitations include the retrospective nature of this study as well as the opaque black box nature of deep learning networks. CONCLUSIONS: Our results provide evidence that deep learning networks may be used for mortality risk stratification based on standard-of-care CT images from NSCLC patients. This evidence motivates future research into better deciphering the clinical and biological basis of deep learning networks as well as validation in prospective data.

Cognitive functions, electroencephalographic and diffusion tensor imaging changes in children with active idiopathic epilepsy.

INTRODUCTION: Neurocognitive impairment represents one of the most common comorbidities occurring in children with idiopathic epilepsy. Diagnosis of the idiopathic form of epilepsy requires the absence of any macrostructural abnormality in the conventional MRI. Though changes can be seen at the microstructural level imaged using advanced techniques such as the Diffusion Tensor Imaging (DTI). AIM OF THE WORK: The aim of this work is to study the correlation between the microstructural white matter DTI findings, the electroencephalographic changes and the cognitive dysfunction in children with active idiopathic epilepsy. METHODS: A comparative cross-sectional study, included 60 children with epilepsy based on the Stanford-Binet 5th Edition Scores was conducted. Patients were equally assigned to normal cognitive function or cognitive dysfunction groups. The history of the epileptic condition was gathered via personal interviews. All patients underwent brain Electroencephalography (EEG) and DTI, which was analyzed using FSL. RESULTS: The Fractional Anisotropy (FA) was significantly higher whereas the Mean Diffusivity (MD) was significantly lower in the normal cognitive function group than in the cognitive dysfunction group. This altered microstructure was related to the degree of the cognitive performance of the studied children with epilepsy. The microstructural alterations of the neural fibers in children with epilepsy and cognitive dysfunction were significantly related to the younger age of onset of epilepsy, the poor control of the clinical seizures, and the use of multiple antiepileptic medications. CONCLUSION: Children with epilepsy and normal cognitive functions differ in white matter integrity, measured using DTI, compared with children with cognitive dysfunction. These changes have important cognitive consequences.

Use of a 3D-Printed Abdominal Compression Device to Facilitate CT Fluoroscopy-Guided Percutaneous Interventions.

OBJECTIVE: The purpose of this article is to describe a handheld external compression device used to facilitate CT fluoroscopy-guided percutaneous interventions in the abdomen. CONCLUSION: The device was designed with computer-aided design software to modify an existing gastrointestinal fluoroscopy compression device and was constructed by 3D printing. This abdominal compression device facilitates access to interventional targets, and its use minimizes radiation exposure of radiologists. Twenty-one procedures, including biopsies, drainage procedures, and an ablation, were performed with the device. Radiation dosimetry data were collected during two procedures.

Effect of treatment of chronic hepatitis c virus patients with direct-acting anti-retroviral drugs on semen and hormonal parameters.

Objective: Hepatitis C virus (HCV) infection is known to influence the seminal and hormonal parameters of infected men. This study was performed to assess the effects of HCV clearance using direct-acting antiviral (DAA) agents on semen and hormonal parameters. Methods: A total of 50 patients with chronic HCV were enrolled, and conventional semen analysis was performed according to World Health Organization guidelines. Basal levels of total testosterone, free testosterone (FT), follicle-stimulating hormone (FSH), luteinizing hormone (LH), estradiol (E2), prolactin, and sex hormone-binding globulin (SHBG) were assessed before and 3 months after treatment with DAAs. Results: Following DAA treatment, statistically significant increases were observed in sperm motility and the proportion of grade A sperm. Additionally, the percentage of abnormal forms was significantly decreased after treatment (p=0.000). However, no significant differences were observed in semen volume, concentration, or total sperm count. Sex hormone analysis of patients after DAA treatment revealed significant increases in FT, LH, and FSH levels, along with significant decreases in SHBG, prolactin, and E2 levels. Conclusion: Following HCV clearance, we noted an improvement in sperm motility and an increase in the percentage of sperm with normal morphology. Treatment with DAAs was also associated with increased levels of FT and LH, along with decreased levels of SHBG, prolactin, and E2.

Foundation model for cancer imaging biomarkers.

Foundation models in deep learning are characterized by a single large-scale model trained on vast amounts of data serving as the foundation for various downstream tasks. Foundation models are generally trained using self-supervised learning and excel in reducing the demand for training samples in downstream applications. This is especially important in medicine, where large labelled datasets are often scarce. Here, we developed a foundation model for cancer imaging biomarker discovery by training a convolutional encoder through self-supervised learning using a comprehensive dataset of 11,467 radiographic lesions. The foundation model was evaluated in distinct and clinically relevant applications of cancer imaging-based biomarkers. We found that it facilitated better and more efficient learning of imaging biomarkers and yielded task-specific models that significantly outperformed conventional supervised and other state-of-the-art pretrained implementations on downstream tasks, especially when training dataset sizes were very limited. Furthermore, the foundation model was more stable to input variations and showed strong associations with underlying biology. Our results demonstrate the tremendous potential of foundation models in discovering new imaging biomarkers that may extend to other clinical use cases and can accelerate the widespread translation of imaging biomarkers into clinical settings.

End-to-end reproducible AI pipelines in radiology using the cloud.

Artificial intelligence (AI) algorithms hold the potential to revolutionize radiology. However, a significant portion of the published literature lacks transparency and reproducibility, which hampers sustained progress toward clinical translation. Although several reporting guidelines have been proposed, identifying practical means to address these issues remains challenging. Here, we show the potential of cloud-based infrastructure for implementing and sharing transparent and reproducible AI-based radiology pipelines. We demonstrate end-to-end reproducibility from retrieving cloud-hosted data, through data pre-processing, deep learning inference, and post-processing, to the analysis and reporting of the final results. We successfully implement two distinct use cases, starting from recent literature on AI-based biomarkers for cancer imaging. Using cloud-hosted data and computing, we confirm the findings of these studies and extend the validation to previously unseen data for one of the use cases. Furthermore, we provide the community with transparent and easy-to-extend examples of pipelines impactful for the broader oncology field. Our approach demonstrates the potential of cloud resources for implementing, sharing, and using reproducible and transparent AI pipelines, which can accelerate the translation into clinical solutions.

Antioxidant, antiapoptotic, and antifibrotic abilities of L-Arginine ameliorate the testicular dysfunction in diabetic rats.

Testicular dysfunction and infertility are serious complications of diabetes mellitus (DM). L-Arginine (L-Arg) is a semi essential amino acid with various biological and metabolic functions. The molecular mechanisms of L-Arg on testicular dysfunction caused by DM remain elusive. This study aimed to assess the potential protective effect of L-Arg in diabetic testis and its possible mechanisms. 24 adult male Wistar albino rats were randomly divided into four groups: CON, L-Arg that received 1 g/kg body weight of L-Arg orally for 4 weeks, DM that fed a high fat diet followed by an injection of 30 mg/kg streptozotocin intraperitoneally, and L-Arg-treated DM that were diabetic and administered L-Arg. DM decreased relative testicular weight, reduced serum testosterone, and impaired semen parameters. Reduced total antioxidant capacity (TAC), superoxide dismutase (SOD), and glutathione peroxidase (GSH-Px), in addition to increased transforming growth factor B1 (TGF-ß1) and nitric oxide (NO) levels, were found in the testicular tissue. This was associated with severe degenerative changes in the seminiferous tubules and interstitial cells of Leydig, reduction of Johnsen's score, significantly increased expression of both inducible nitric oxide synthase (iNOS) and caspase-3, and reduced zonula occludens (ZO)- 1 expression. Ultrastructurally, disrupted intercellular junctions and degeneration of interstitial cells of Leydig were observed. In contrast, treatment of diabetic animals with L-Arg increased TAC, SOD and GSH-Px, decreased TGF-ß1 and NO levels, downregulated iNOS and caspase-3 expression, upregulated ZO-1 expression, and maintained the integrity of the Sertoli cell junctions. Hence, L-Arg restored the normal testicular structure and function via its antioxidant, antiapoptotic, and antifibrotic effects.

Characterization of Brucella abortus Mutant A19mut2, a Potential DIVA Vaccine Candidate with a Modification on Lipopolysaccharide.

BACKGROUND: Brucella abortus is the main causative agent for bovine brucellosis. B. abortus A19 is a widely used vaccine strain to protect cows from Brucella infection in China. However, A19 has a similar lipopolysaccharide (LPS) antigen to that of the field virulent Brucella strain, whose immunization interferes with the serodiagnosis of vaccinated and infected animals. [Aim] To develop a novel Brucella DIVA vaccine candidate. STUDY DESIGN AND METHODS: The B. abortus mutant A19mut2 with the formyltransferase gene wbkC is replaced by an acetyltransferase gene wbdR from E. coli O157 using the bacterial homologous recombination technique, generating a modified O-polysaccharide that cannot induce antibodies in mice against wild-type Brucella LPS. The biological phenotypes of the A19mut2 were assessed using a growth curve analysis, agglutination tests, Western blotting, and stress resistance assays. Histopathological changes and bacterial colonization in the spleens of vaccinated mice were investigated to assess the residual virulence and protection of the A19mut2. Humoral and cellular immunity was evaluated by measuring the levels of IgG, IgG subtypes, and the release of cytokines IFN-γ and IL10 in the splenocytes of the vaccinated mice. ELISA coated with wild-type LPS can distinguish mouse antibodies induced by A19 and A19mut2 immunization. RESULTS: The A19mut2 showed a decreased residual virulence in mice, compared to the A19 strain, but induced significant humoral and cellular immune responses, as the A19 immunization did. The protection efficacy of A19mut2 immunization against B. abortus S2308 NalR infection was similar to that of A19 immunization. CONCLUSION: The A19mut2 has potential as a novel DIVA vaccine candidate in the future.

Foundation Models for Quantitative Biomarker Discovery in Cancer Imaging.

Foundation models represent a recent paradigm shift in deep learning, where a single large-scale model trained on vast amounts of data can serve as the foundation for various downstream tasks. Foundation models are generally trained using self-supervised learning and excel in reducing the demand for training samples in downstream applications. This is especially important in medicine, where large labeled datasets are often scarce. Here, we developed a foundation model for imaging biomarker discovery by training a convolutional encoder through self-supervised learning using a comprehensive dataset of 11,467 radiographic lesions. The foundation model was evaluated in distinct and clinically relevant applications of imaging-based biomarkers. We found that they facilitated better and more efficient learning of imaging biomarkers and yielded task-specific models that significantly outperformed their conventional supervised counterparts on downstream tasks. The performance gain was most prominent when training dataset sizes were very limited. Furthermore, foundation models were more stable to input and inter-reader variations and showed stronger associations with underlying biology. Our results demonstrate the tremendous potential of foundation models in discovering novel imaging biomarkers that may extend to other clinical use cases and can accelerate the widespread translation of imaging biomarkers into clinical settings.

Contribution of interleukin 27 serum level to pathogenesis and prognosis in children with immune thrombocytopenia.

ABSTRACT: Concepts surrounding the mechanisms of thrombocytopenia in ITP have shifted from the traditional view of autoantibody mediated platelet destruction to more complex mechanisms in which impaired platelet production, T-cell-mediated effects, and disturbed cytokine profiles play a role. Interleukin 27 (IL-27) plays pleiotropic roles in immunomodulation and autoimmune diseases.We aimed to determine the level of IL-27 in patients with ITP and its relationship to patient and disease characteristics as well as disease chronicity and response to treatment.Sixty childrens with primary immune thrombocytopenia were consequetively enrolled in this study as well as 20 age and sex matched healthy controls.ITP patients had significantly higher levels of IL-27 than controls (770.6 and 373.8 pg/ml, respectively). Patients with acute ITP had the highest levels of IL-27 among patient groups, while patients in remission had the lowest IL-27 levels (860.1and 622.9 pg/ml, respectively). Patients who received IVIG and combined steroids plus IVIG had significantly higher IL-27 levels than others. Patients who received Eltrombopag had significantly lower IL-27 levels than others.IL-27 seems to play a role in pathogenesis of childhood ITP. IL-27 can be used as a predictor for disease occurrence as well as responsiveness to treatment.

Effect of Light Conditions and Accommodation on Implantable Phakic Contact Lens Vault by Anterior Segment Optical Coherence Tomography.

Purpose: To investigate the dynamic pupil and vault changes in eyes with implantable phakic contact lens (IPCL) under photopic and scotopic settings, as well as during accommodation using the anterior segment optical coherence tomography (AS-OCT). Methods: A prospective observational study included consecutive 36 eyes of myopic patients who underwent IPCL V2.0 implantation. Under photopic and scotopic light settings, as well as during accommodation, all patients were scanned using CASIA OCT (CASIA2; TOMEY, Nagoya, Japan). The pupil size, the vault (distance between the back surface of the IPCL and the anterior lens capsule), ACD-lens (distance between the posterior corneal surface and the anterior lens surface), IPCL-lens (distance between the posterior corneal surface and the anterior IPCL surface), and lens thickness (LT) were the study parameters. Results: The vault was significantly lower under photopic conditions (p-value<0.001). The pupil size was significantly smaller in photopic conditions (p-value<0.001). LT (p-value=0.975) and ACD-lens (p-value=0.917) were not significantly different between scotopic and photopic conditions, while the ACD-IPCL was significantly larger during photopic conditions (p-value=0.013). There were significant changes in all parameters between accommodative and non-accommodative conditions. Conclusion: The IPCL vault decreased significantly under photopic light conditions and accommodation.

Novel protective effect of diosmin against cisplatin-induced prostate and seminal vesicle damage: Role of oxidative stress and apoptosis.

Multiple organ toxicity has been associated with cisplatin (CIS) treatment, limiting its clinical use. The human prostate and seminal vesicles are accessory sex organs with androgen-dependent morphogenesis, growth, and secretion. The present study aimed to investigate, for the first time, the toxic effect of CIS on normal prostate and seminal vesicles in the presence and absence of diosmin (DS). The animals were randomized into 4 groups as follows: control (received vehicle), CIS group (7.5 mg/kg, i.p. on 5th and 12th day), DS group (100 mg/kg, p.o. for 15 days), and DS+CIS group. Histopathological and biochemical analyses were conducted to elucidate the goal of this study. CIS administration significantly induced prostate and seminal vesicle toxicity as evidenced by alteration of serum testosterone, LH, FSH, PSA, steroidogenic HSD17B6 as well as seminal analysis markers. Remarkably, marked histopathological changes in thin and ultrathin structures were observed. Besides, CIS significantly boosted oxidative stress as evidenced by the up-regulation of MDA and down-regulation of TAC. CIS significantly induced tissue apoptosis concomitant with suppression of cellular proliferation and stem cell expression as indicated by up-regulation of activated caspase-3 and Bax expression along with down-regulation of Bcl-2, Ki67, and CD44 expression. Interestingly, DS fixed all disturbances in the prostate and seminal vesicles induced by CIS. Together, CIS could cause prostate and seminal vesicle toxicity by affecting hormonal, steroidogenic, oxidative stress, apoptosis, and proliferation processes, and this effect was reversed by DS administration.

Deep Learning-based Detection of Intravenous Contrast Enhancement on CT Scans.

Identifying the presence of intravenous contrast material on CT scans is an important component of data curation for medical imaging-based artificial intelligence model development and deployment. Use of intravenous contrast material is often poorly documented in imaging metadata, necessitating impractical manual annotation by clinician experts. Authors developed a convolutional neural network (CNN)-based deep learning platform to identify intravenous contrast enhancement on CT scans. For model development and validation, authors used six independent datasets of head and neck (HN) and chest CT scans, totaling 133 480 axial two-dimensional sections from 1979 scans, which were manually annotated by clinical experts. Five CNN models were trained first on HN scans for contrast enhancement detection. Model performances were evaluated at the patient level on a holdout set and external test set. Models were then fine-tuned on chest CT data and externally validated. This study found that Digital Imaging and Communications in Medicine metadata tags for intravenous contrast material were missing or erroneous for 1496 scans (75.6%). An EfficientNetB4-based model showed the best performance, with areas under the curve (AUCs) of 0.996 and 1.0 in HN holdout (n = 216) and external (n = 595) sets, respectively, and AUCs of 1.0 and 0.980 in the chest holdout (n = 53) and external (n = 402) sets, respectively. This automated, scan-to-prediction platform is highly accurate at CT contrast enhancement detection and may be helpful for artificial intelligence model development and clinical application. Keywords: CT, Head and Neck, Supervised Learning, Transfer Learning, Convolutional Neural Network (CNN), Machine Learning Algorithms, Contrast Material Supplemental material is available for this article. © RSNA, 2022.

Isolation, Molecular Characterization, and Antibiotic Resistance of Avian Pathogenic Escherichia coli in Eastern China.

Avian pathogenic Escherichia coli (APEC) causes colibacillosis in avians, resulting in considerable losses in the poultry industry. APEC showed zoonotic potential initially related to the fact that APEC serves as the reservoir of virulence genes and antibiotic resistance genes for other E. coli. Thus, we determine the serotypes, phylogenetic groups, virulence genes distribution, and antibiotic resistance profiles of APEC isolates in eastern China. A total of 230 APEC were isolated from diseased chicken and duck with typical colibacillosis symptoms. Serotyping identified that O78 (44.78%) was the predominant serotype. The majority of APEC isolates were classified into B2 (29.57%), A (26.96%), D (20.00%), and B1 (18.26%), respectively. Among the 15 virulence genes, a high prevalence of ibeB (99.57%), fimC (91.74%), mat (91.30%), ompA (83.04%), and iss (80.43%) genes was observed. Except for low resistance rates for imipenem (1.7%) and polymyxin B (0.4%), most of the APEC isolates were resistant to erythromycin (98.7%), enrofloxacin (96.1%), tetracycline (95.2%), doxycycline (93.9%), lincomycin (90.0%), and streptomycin (90.0%). Moreover, all APEC exhibit multi-drug resistance. This study indicated that APEC isolates harbor a variety of virulence genes and showed multi-antibiotic resistance profiles, providing proof for understanding the epidemiological background and zoonotic potential of APEC in poultry farms.

Clinical validation of deep learning algorithms for radiotherapy targeting of non-small-cell lung cancer: an observational study.

BACKGROUND: Artificial intelligence (AI) and deep learning have shown great potential in streamlining clinical tasks. However, most studies remain confined to in silico validation in small internal cohorts, without external validation or data on real-world clinical utility. We developed a strategy for the clinical validation of deep learning models for segmenting primary non-small-cell lung cancer (NSCLC) tumours and involved lymph nodes in CT images, which is a time-intensive step in radiation treatment planning, with large variability among experts. METHODS: In this observational study, CT images and segmentations were collected from eight internal and external sources from the USA, the Netherlands, Canada, and China, with patients from the Maastro and Harvard-RT1 datasets used for model discovery (segmented by a single expert). Validation consisted of interobserver and intraobserver benchmarking, primary validation, functional validation, and end-user testing on the following datasets: multi-delineation, Harvard-RT1, Harvard-RT2, RTOG-0617, NSCLC-radiogenomics, Lung-PET-CT-Dx, RIDER, and thorax phantom. Primary validation consisted of stepwise testing on increasingly external datasets using measures of overlap including volumetric dice (VD) and surface dice (SD). Functional validation explored dosimetric effect, model failure modes, test-retest stability, and accuracy. End-user testing with eight experts assessed automated segmentations in a simulated clinical setting. FINDINGS: We included 2208 patients imaged between 2001 and 2015, with 787 patients used for model discovery and 1421 for model validation, including 28 patients for end-user testing. Models showed an improvement over the interobserver benchmark (multi-delineation dataset; VD 0·91 [IQR 0·83-0·92], p=0·0062; SD 0·86 [0·71-0·91], p=0·0005), and were within the intraobserver benchmark. For primary validation, AI performance on internal Harvard-RT1 data (segmented by the same expert who segmented the discovery data) was VD 0·83 (IQR 0·76-0·88) and SD 0·79 (0·68-0·88), within the interobserver benchmark. Performance on internal Harvard-RT2 data segmented by other experts was VD 0·70 (0·56-0·80) and SD 0·50 (0·34-0·71). Performance on RTOG-0617 clinical trial data was VD 0·71 (0·60-0·81) and SD 0·47 (0·35-0·59), with similar results on diagnostic radiology datasets NSCLC-radiogenomics and Lung-PET-CT-Dx. Despite these geometric overlap results, models yielded target volumes with equivalent radiation dose coverage to those of experts. We also found non-significant differences between de novo expert and AI-assisted segmentations. AI assistance led to a 65% reduction in segmentation time (5·4 min; p<0·0001) and a 32% reduction in interobserver variability (SD; p=0·013). INTERPRETATION: We present a clinical validation strategy for AI models. We found that in silico geometric segmentation metrics might not correlate with clinical utility of the models. Experts' segmentation style and preference might affect model performance. FUNDING: US National Institutes of Health and EU European Research Council.