Design and Development of a Medical Image Databank for Assisting Studies in Radiomics.

CompreHensive Digital ArchiVe of Cancer Imaging - Radiation Oncology (CHAVI-RO) is a multi-tier WEB-based medical image databank. It supports archiving de-identified radiological and clinical datasets in a relational database. A semantic relational database model is designed to accommodate imaging and treatment data of cancer patients. It aims to provide key datasets to investigate and model the use of radiological imaging data in response to radiation. This domain of research area addresses the modeling and analysis of complete treatment data of oncology patient. A DICOM viewer is integrated for reviewing the uploaded de-identified DICOM dataset. In a prototype system we carried out a pilot study with cancer data of four diseased sites, namely breast, head and neck, brain, and lung cancers. The representative dataset is used to estimate the data size of the patient. A role-based access control module is integrated with the image databank to restrict the user access limit. We also perform different types of load tests to analyze and quantify the performance of the CHAVI databank.

Clinicopathological characteristics, prognostic factors and treatment outcomes of seminomatous germ cell tumours from a tertiary cancer centre in eastern India.

Background: . Seminomatous germ cell tumour (SGCT) is a rare but curable malignancy of young adults. The literature on management and outcome of SGCT is scarce from India. We report the demography and treatment outcome of SGCT at our centre. Methods: . We did a retrospective analysis of patients with SGCT treated from March 2011 to December 2018. Patients were staged appropriately with imaging, and pre- and postoperative tumour markers. High inguinal orchiectomy was performed in all with a testicular primary and received subsequent stage-adjusted adjuvant treatment. Patients were monitored for metabolic syndrome during follow-up after completion of treatment. Results: . We treated 85 patients with a median age of 37 (range 20-68) years. The primary site of the tumour was the testis in 80 (94%) and mediastinum in 5 (6%) patients. Cryptorchidism was present in 20 (25%) patients and testicular violation was present in 11 (14%) patients. Stage of the disease was I in 61, II in 13 and III in 6 patients. Adjuvant treatment in stage I disease was single-agent carbo-platin (area under the curve ×7) in 38 (62%), surveillance in 20 (33%) and radiotherapy in 3 (5%) patients. Five patients in the surveillance group relapsed. The 7-year mean (SD) relapse-free survival and overall survival were 83.1% (8%) and 98.7% (1.3%), respectively. Thirty-one patients (n = 52, 60%) had features of metabolic syndrome. Conclusions: . SGCTs have a high cure rate. Long-term follow-up is essential for monitoring toxic effects. Early diagnosis, avoidance of testicular violation and multidisciplinary management are the key features for better long-term outcome in SGCT.

Research Goal-Driven Data Model and Harmonization for De-Identifying Patient Data in Radiomics.

There are various efforts in de-identifying patient's radiation oncology data for their uses in the advancement of research in medicine. Though the task of de-identification needs to be defined in the context of research goals and objectives, existing systems lack the flexibility of modeling data and normalization of names of attributes for accomplishing them. In this work, we describe a de-identification process of radiation and clinical oncology data, which is guided by a data model and a schema of dynamically capturing domain ontology and normalization of terminologies, defined in tune with the research goals in this area. The radiological images are obtained in DICOM format. It consists of diagnostic, radiation therapy (RT) treatment planning, RT verification, and RT response images. During the DICOM de-identification, a few crucial pieces of information are taken about the dataset. The proposed model is generic in organizing information modeling in sync with the de-identification of a patient's clinical information. The treatment and clinical data are provided in the comma-separated values (CSV) format, which follows a predefined data structure. The de-identified data is harmonized throughout the entire process. We have presented four specific case studies on four different types of cancers, namely glioblastoma multiforme, head-neck, breast, and lung. We also present experimental validation on a few patients' data in these four areas. A few aspects are taken care of during de-identification, such as preservation of longitudinal date changes (LDC), incremental de-identification, referential data integrity between the clinical and image data, de-identified data harmonization, and transformation of the data to an underlined database schema.

A Web-based Dose-volume Histogram Dashboard for Library-based Individualized Dose-constraints and Clinical Plan Evaluation.

Traditional methods of treatment planning and plan evaluation involve the use of generic dose-constraints. We aimed to build a web-based application to generate individualized dose-constraints and plan evaluation against a library of prior approved plan dose-volume histograms (DVH).A prototype was built for intensity modulated radiation therapy (IMRT) plans for prostate cancer. Using exported DVH files from the Varian and Accuray treatment planning systems, a library of plan DVHs was built by data extraction. Given structure volumes of a patient to be planned, a web based application was built to derive individual dose-constraints of the planning target volume (PTV) and organs-at-risk (OAR) based on achieved doses in a library of prior approved plans with similar anatomical volumes, selected using an interactive dashboard. A second web application was built to compare the achieved DVHs of the newly created plan against a library of plans of similar patients.These web application prototypes are a proof of principle that simple freely available tools can be built for library based planning and review.

Redefining adequate margins in oral squamous cell carcinoma: outcomes from close and positive margins.

PURPOSE: Adequacy of surgical margins impacts outcomes in oral cancer. We sought to determine whether close and positive margins have different outcomes in patients with oral cancer. METHODS: Retrospective data from 612 patients with oral carcinoma were analyzed for the effect of margin status on locoregional recurrence-free survival (LRFS), disease-free survival (DFS) and overall survival (OS). RESULTS: A total of 90 cases (14.7%) had close margins and 26 patients (4.2%) had positive margins. Recurrences were documented in 173 patients (28%), of which 137 (22% of the study sample) were locoregional, and 164 patients (27%) had died. Among patients with close or positive margins, a cutoff of 1 mm optimally separated LRFS (adjusted p = 0.0190) and OS curves (adjusted p = 0.0168) whereas a cutoff of 2 mm was sufficient to significantly separate DFS curves (adjusted p = 0.0281). CONCLUSIONS: Patients with oral carcinoma with positive margins (< 1 mm) had poorer outcomes compared to those with close margins (1-5 mm) in terms of LRFS, DFS and OS. There is a suggestion that a cutoff of < 2 mm might provide slightly more separation for DFS.

De-Identification of Radiomics Data Retaining Longitudinal Temporal Information.

We propose a de-identification system which runs in a standalone mode. The system takes care of the de-identification of radiation oncology patient's clinical and annotated imaging data including RTSTRUCT, RTPLAN, and RTDOSE. The clinical data consists of diagnosis, stages, outcome, and treatment information of the patient. The imaging data could be the diagnostic, therapy planning, and verification images. Archival of the longitudinal radiation oncology verification images like cone beam CT scans along with the initial imaging and clinical data are preserved in the process. During the de-identification, the system keeps the reference of original data identity in encrypted form. These could be useful for the re-identification if necessary.

Challenges in the Assessment of Medullary Bone Invasion in Oral Cavity Cancers and Its Prognostic Significance.

BACKGROUND: As per AJCC 8th edition TNM staging system, bone invasion is a poor prognostic marker that upstages oral cavity squamous carcinoma (OSCC) to pT4a. Cortical erosion alone of bone or tooth socket by a gingival primary is not sufficient to upstage a tumour. The differentiation of cortical erosion from invasion through the cortical bone into the medulla is often challenging, limiting accurate staging. This review aims to assess the difficulties in differentiating cortical erosion from medullary invasion and evaluate the prognostic significance of different patterns of bone involvement. METHODS: A retrospective review of OSCC with primary curative surgery and bone resection treated at a single-center over 10 years, was performed to assess the prognostic significance of bone invasion. Hematoxylin-eosin stained slides of a subset of cases were re-reviewed in a planned manner to assess difficulties in precise categorization (no invasion/erosion/cortical invasion and medullary invasion), evaluate interobserver agreement, and correlate with clinical outcome. RESULTS: Five hundred and ninety patients were included, with a median follow-up of 28 months. On univariate analysis, the 3-year local, nodal and distant metastasis control were not significantly different in the 3 groups of no invasion, erosion, and invasion (p = 0.43, 0.47, and 0.47, respectively). Overall survival (OS) at 3 years was 78.1% and disease-free-survival(DFS) was 63.7% in the entire cohort. On univariate analysis, there was significant difference in OS and DFS based on these groups. This did not translate into independent prognostic benefit on multivariable analysis (p = 0.75 and 0.19, respectively). The independent prognostic factors were margin positivity, tumor differentiation, perineural invasion and pathological nodal involvement. Planned re-review of a subset of 202 cases resulted in a change in bone involvement category in 26/202 cases, which was mainly due to difficulty in assessing cortico-medullary junction near the tooth socket and bone fragmentation. The assessment showed moderate to near complete agreement (kappa 0.59-0.82) between 2 observers. CONCLUSION: Our study shows that bone involvement is not an independent prognostic marker and there is no specific correlation of medullary invasion with outcome over those that showed cortical erosion. Several factors contribute to difficulties and interobserver variability in assessing bone involvement.

A re-look at the modified Ryan regression scoring system in esophageal cancer-validation of prognostic significance and comparison with other less commonly used systems.

BACKGROUND: Neoadjuvant chemoradiotherapy (NACRT) using CROSS protocol is currently the treatment of choice for esophageal cancer (EC). Tumor response grade (TRG) is a mandatory reporting criterion in most guidelines. One of the most commonly used TRG systems is the modified Ryan system. We aim to assess the TRG using modified Ryan and seven other systems (Mandard, Chireac, Swisher, Japanese esophageal society guidelines, modified rectal cancer regression grading (mRCRG), CROSS, and Becker) to evaluate their reproducibility and role as a prognostic marker. MATERIALS AND METHODS: Two pathologists independently reviewed all cases of post-NACRT (CROSS) EC, to score TRGs and other histological parameters. Inter-rater agreement assessment for different TRG systems and correlation with disease-free survival (DFS) was performed. RESULTS: Our series includes 93 patients with predominantly mid-esophageal squamous cell carcinoma. Complete pathological response (pCR) was noted in 47% (44/93) patients. The kappa inter-rater agreement score for the Ryan system was substantial (0.774), while it was almost perfect agreement for tumor percentage assessment-based systems (Swisher, CROSS, and Becker). Only the mRCRG TRGs correlated significantly with prognosis, while the Ryan system did not. Tumor stage and pCR status did not correlate with DFS, though the nodal stage was clinically significant. CONCLUSION: Though the inter-rater concordance was optimal for all the TRG systems studied, only the mRCRG system showed prognostic significance, while the commonly used modified Ryan system did not. It may be worthwhile to look at further evaluating other systems like mRCRG for inclusion in minimum dataset reporting.

Are oral cancers effectively palliated with radiotherapy? Outcomes of treatment with a modified QUAD SHOT regimen.

BACKGROUND: This study assessed a palliative radiotherapy regimen using daily radiation over 4 days for three courses in inoperable head and neck cancers, emphasizing oral primary cancers. METHODS: Retrospective data of 116 patients treated with a daily dose of 3.6-3.7 Gy in four fractions over 4 days to a total of three courses, with a 2-week gap after every course, were analyzed for survival outcomes. A subgroup analysis was done for oral cancer. RESULTS: Ninety-nine (85%) completed three courses. Overall subjective response rate was 77%. Median overall survival and progression-free survival were 12 months (95% confidence interval [CI]: 8-20) and 8 months (95% CI: 6-10), with numerically higher overall survival in oral cancer. The treatment was well tolerated, with no on-treatment hospitalization or grade 3-4 toxicities. CONCLUSION: The modified QUAD SHOT regimen is practical for palliation in head and neck cancers.

Enhancing quality assurance in radiotherapy for gynaecological cancers: implementation of an on-demand peer review process.

OBJECTIVES: Ensuring high-quality radiotherapy requires peer-reviewing target volumes. The Royal College of Radiologists recommends peer review specifically for individual target volumes in cases of gynaecological cancers. This study presents the outcomes of implementing an on-demand peer review system for gynaecological cancers within our institute. METHODS: The peer review process was planned for gynaecological cancer cases intended for curative radiotherapy. After junior clinical oncologists (COs) completed the segmentation, two senior COs specializing in gynaecological cancers conducted the peer review. All peer review outcomes were recorded prospectively. The audit process compliance, the proportion of patients requiring major and minor modifications in target volumes, the direction of changes, and the factors influencing these changes were reported. RESULTS: A total of 230 patients were eligible, and out of these, 204 (88.3%) patients underwent at least one peer review. Among the patients, 108 required major modifications in their target volumes. P-charts revealed a stabilization in the need for major modifications at the end of three months, indicating that 38.2% and 28% of patients still required major modifications for the nodal and primary CTV, respectively. Multivariable analysis demonstrated that major modifications were associated with the use of extended field radiotherapy and radical radiation in non-cervical primary cases. CONCLUSIONS: An on-demand peer review system was feasible and resulted in clinically meaningful, major modifications in the target volumes for 53% of patients. ADVANCES IN KNOWLEDGE: Gynaecological cancers require ongoing peer review to ensure quality of care in radiotherapy. A flexible on-demand system not only ensures that patient treatment start is not delayed but also has an important educational role for junior trainees.

18F-fluorodeoxyglucose PET-CT-guided pelvic chemoradiation therapy using helical tomotherapy for locally advanced carcinoma cervix without para-aortic nodal disease: Clinical and patient-reported outcomes from a prospective phase 2 study.

INTRODUCTION: Locally advanced carcinoma cervix (LACC) is a heterogeneous disease with variable combinations of primary tumour extensions with or without nodal involvement. Metabolic information from 18 fluro-deoxyglucose positron emission tomography combined with contrast-enhanced computerized tomography (FDG PET-CT) may potentially augment treatment decision-making for LACC. This study ascertained FDG-PET CT influence on chemoradiation therapy (CTRT) decisions in LACC. We report oncologic and patient-reported outcome measures (PROMs). METHODS: FDG PET-CT scans were reviewed independently by two nuclear medicine specialists and two radiation oncologists. Pelvic CTRT plan digressions were documented and therapy was adapted accordingly. Pelvis radiation (50 Gy/25#/5 weeks) using tomotherapy with weekly cisplatin was used in node-negative disease. Dose-escalated simultaneous integrated boost (SIB) 60 Gy/25#/5 weeks was delivered to involved pelvic nodes. All received brachytherapy. Post-treatment PET-CT scans were at 6 months. Functional assessment of cancer therapy scores were calculated at baseline, treatment completion, 3 months, 1 year and 3 years. RESULTS: Between November 2015 and January 2018, 85 patients were screened, and 77 consented. Extrapelvic disease was seen in 12 (16%) patients (9 para-aortic nodes, 2 distant metastases and 1 synchronous carcinoma breast); 60 patients were included in the final analysis. Decision changes were seen in 10/77 (13%) screened, 8/60 (13%) included and 32 (53.3%) received SIB. Post-treatment, 27 (45%) had grade 2 GI/GU/GYN toxicity, one (2%) had grade 3 GI and five (8.3%) had grade 3 neutropenia. At median overall survival of 54.2 months (95% CI 52.8-58.3), 5-year local failure, pelvic nodal and para-aortic nodal-free survival were 86.8% (95% CI 78.0-96.6), 85.2% (95% CI 76.1-95.3) and 85.2% (95% CI 76.2-95.4). Functional assessment of cancer therapy trial outcome index (FACT TOI) improved by 10.43 at 3 months with no further decline. Grade 3 toxicity was noted for abdominal pain in one (1.7%), cystitis in four (6.7%) and lymphoedema in one (1.7%) at 5 years. CONCLUSION: PET-CT resulted in major decision changes in 13%. PET-adapted CTRT was associated with acceptable toxicity, encouraging long-term survival and improvement in PROMS.

Definitive chemoradiation for oropharyngeal squamous carcinomas: Outcomes with intensity-modulated radiation therapy using simultaneous integrated boost, in a majorly p16 negative cohort.

BACKGROUND: The literature on modern-era outcomes of oropharyngeal squamous carcinoma (OPSCC) in India is limited. MATERIALS AND METHODS: We analyzed records of consecutive patients with OPSCC treated using a curative SIB IMRT regimen of 66 Gy/30#/6 weeks. RESULTS: One hundred fifteen patients from July 2011 to December 2018 were analyzed. Twenty of 69 patients tested positive for p16. In p16 positive patients, the K-M probability of being disease free and alive at 2 years, with at least one follow-up 3 months after treatment, was 83% (median not reached) compared with 48% if p16 was unknown/negative. Patients staged as IVB p16 negative had a 2-year DFS of 25%. Patients unfit for cisplatin and consequently received other agents had 2-year DFS estimated at 20%. CONCLUSIONS: Intensity-modulated radiation therapy (IMRT) with simultaneous integrated boost (SIB) and concurrent chemotherapy was feasible, with toxicity and disease control comparable to available literature. AJCC Stage IVB p16 negative disease had notably poor outcome.

Competency-Based Workforce Development and Education in Global Oncology.

The healthcare workforce plays a pivotal role in cancer care delivery, leadership, policy, education, and research in complex cancer systems. To ensure quality and relevance, health professionals must have the necessary competencies to deliver patient-centered and efficient care, coupled with the ability to work in teams and manage health resources wisely. This paper aims to review the concept of competency-based medical education (CBME) in the context of oncology to provide insights and guidance for those interested in adopting or adapting competency-based education in training programs. The results of a scoping review of CBME in oncology are presented here to describe the current status of CBME in oncology. The literature describing the implementation and evaluation of CBME in oncology training programs for medical professionals internationally is summarized and key themes identified to provide practical guidance for educators. Further, the paper identifies critical competencies for oncology education and training globally and presents recommendations and opportunities for collaboration in competency-based education and training in oncology. The authors argue for increased global collaboration and networking in the realm of CBME to facilitate the establishment of a competent global cancer care workforce.

ICON-P - A double-blind evaluation of quality improvements with individualized CONstraints from low-cost knowledge-based radiation therapy planning in prostate cancer.

Purpose: /Objective(S)A low-cost, prior knowledge-based individualized dose-constraint generator for organs-at-risk has been developed for prostate cancer radiation therapy (RT) planning. In this study, we aimed to evaluate the feasibility and improvements in organs-at-risk (OAR) doses in prostate cancer RT planning using this tool served on a web application. Materials And Methods: A set of previously treated prostate cancer cases planned and treated with generic constraints were replanned using individualized dose constraints derived from a library of cases with similar volumes of target, OAR, and overlap regions and served on the web-based application. The goal was to assess the reduction in mean dose, specified dose volumes (V59Gy, V56Gy, V53Gy, V47Gy, and V40Gy), and generalized equivalent uniform dose (gEUD) to the rectum and bladder. Planners and assessors were blinded to the initial achieved doses and penalties. Sample size estimation was based on improvement in V53Gy for the rectum and bladder with a paired evaluation. Results: Twenty-four patients were replanned. All the plans had a PTV D95 of at least 97% of the prescribed dose. The individualized OAR constraints could be met for 87.5% of patients for all dose levels. The mean dose, V59Gy, V53Gy, and V47Gy for the bladder was reduced by 7.5 Gy, 1.12%, 5.51%, and 10.53% respectively. Similarly for the rectum, the mean dose, V59Gy, V53Gy, V47Gy and was reduced by 5.5 Gy, 4.34%, 6.97%, and 11.61% respectively. All dose reductions were statistically significant. The gEUD of the bladder was reduced by 2.47 Gy (p < 0.001) and the rectum by 3.21 Gy (p < 0.001). Conclusion: Treatment planning based on individualized dose constraints served on a web application is feasible and leads to improvement at clinically important dose volumes in prostate cancer RT planning. This application can be served publicly for improvements in RT plan quality.

Alleviating Morbidity From Locally Advanced Breast Cancer Using a Practical and Short Radiation Therapy Regimen: Results of the HYPORT Palliative Studies.

PURPOSE: Locally advanced breast cancers lead to debilitating local symptoms. Treatment of these women encountered commonly in less resourced countries is not backed by strong evidence. We formulated the  HYPORT and HYPORT B phase 1/2 studies to evaluate the safety and efficacy of hypofractionated palliative breast radiation therapy. METHODS AND MATERIALS: Two studies (35 Gy/10 fractions; HYPORT ) and (26 Gy to breast/32 Gy tumor boost in 5 fractions; HYPORT B) were designed with increasing hypofractionation to save overall treatment time from 10 to 5 days. We report the acute toxicity, symptomatic, metabolic response, and quality of life (QOL) changes after radiation therapy. RESULTS: Fifty-eight patients, the majority of whom were pretreated with systemic therapy, completed the treatment. No grade 3 toxicity was reported. Response assessment at 3 months showed improvement in ulceration (58% vs 22%, P = .013) and bleeding (22% vs 0%, P = .074) within the HYPORT study. Similarly, in the HYPORT B study, ulceration (64% and 39%, P = .2), fungating (26% and 0%, P = .041), bleeding (26% and 4.3%, P = .074), and discharge (57% and 8.7%, P = .003) was reduced. Metabolic response was noted in 90% and 83% of patients, respectively, in the 2 studies. Improvement in the QOL scores were evident in both studies. Only 10% of the patients relapsed locally within 1 year. CONCLUSIONS: Palliative ultrahypofractionated radiation therapy to the breast is well tolerated, is effective, and results in a durable response with improved QOL. This could be considered a standard for locoregional symptom control.

Multi-drug resistant (MDR) and extensively drug-resistant (XDR) bacteraemia rates among cancer patients in an oncology hospital in eastern India: an 11-year retrospective observational study.

Background: Trend analysis of bacteraemias caused by multi-drug resistant (MDR) and extensively drug resistant (XDR) bacteria helps to assess efficacy of infection prevention and control (IPC) practices. Data on the trends of MDR and XDR bacteraemias are lacking from cancer patients in India. Aims: To report antibiotic resistance rates over time in bacteraemias and to assess the effect of IPC practices where patient isolation facilities were limited on the rates and trends of MDR and XDR bacteraemias from a cancer centre in eastern India. Methods: A retrospective observational study was conducted in a specialist cancer hospital in India from 2011 to 2021. The study included both patients with haematological and solid organ malignancy. Data on blood cultures and surveillance culture samples were analysed. Blood cultures were processed using BACT/ALERT® (bioMérieux, Marcy-l'Étoile, France) and the identification and antibiotic susceptibilities of bacteria were performed using VITEK® 2 (bioMérieux, Marcy-l'Étoile, France). Surveillance cultures for MDR/XDR bacteria were performed on a subset of patients and processed based on a modified method described previously. Findings: 3rd-generation cephalosporin-resistant Gram negative bacilli were the commonest cause of MDR bacteraemia (57.6%) followed by carbapenem resistant organisms (CRO) (35.7%). Bacteraemias caused by vancomycin-resistant enterococci (VRE), meticillin-resistant Staphylococcus aureus (MRSA) and colistin-resistant Gram negative bacilli were responsible for 1.3%, 2.3% and 3.0% of laboratory confirmed bloodstream infections (BSI) respectively. The ranges of the rates of MDR/XDR BSI per 1000 in-patients during the study period were: MRSA (1-1.18), VRE (0-0.88), 3rd generation cephalosporin-resistant Gram negative bacilli (10.10-20.32), CRO (5.05-13.07) and colistin-resistant Gram negative bacilli (E. coli, Klebsiella, Pseudomonas aeruginosa, Acinetobacter spp (0-1.3). Surveillance cultures collected from a subset of patients showed ranges of MRSA detection in 0-2.11%, VRE in 1.67%-7.49%, 3rd generation cephalosporin-resistant Gram negative bacilli in 55%-89.91% and carbapenem resistant Gram negative bacilli in 18.33%-31.11% of patients. Conclusion: This is one of few studies providing trend data for MDR/XDR bacteraemia rates among cancer patients in India over a decade. In a high prevalence setting it was possible to keep the rates of MDR/XDR bacteraemia controlled with IPC strategies and without adequate isolation facilities. The results are of potential interest to policy makers, IPC specialists and clinicians.

ARCHERY: a prospective observational study of artificial intelligence-based radiotherapy treatment planning for cervical, head and neck and prostate cancer - study protocol.

INTRODUCTION: Fifty per cent of patients with cancer require radiotherapy during their disease course, however, only 10%-40% of patients in low-income and middle-income countries (LMICs) have access to it. A shortfall in specialised workforce has been identified as the most significant barrier to expanding radiotherapy capacity. Artificial intelligence (AI)-based software has been developed to automate both the delineation of anatomical target structures and the definition of the position, size and shape of the radiation beams. Proposed advantages include improved treatment accuracy, as well as a reduction in the time (from weeks to minutes) and human resources needed to deliver radiotherapy. METHODS: ARCHERY is a non-randomised prospective study to evaluate the quality and economic impact of AI-based automated radiotherapy treatment planning for cervical, head and neck, and prostate cancers, which are endemic in LMICs, and for which radiotherapy is the primary curative treatment modality. The sample size of 990 patients (330 for each cancer type) has been calculated based on an estimated 95% treatment plan acceptability rate. Time and cost savings will be analysed as secondary outcome measures using the time-driven activity-based costing model. The 48-month study will take place in six public sector cancer hospitals in India (n=2), Jordan (n=1), Malaysia (n=1) and South Africa (n=2) to support implementation of the software in LMICs. ETHICS AND DISSEMINATION: The study has received ethical approval from University College London (UCL) and each of the six study sites. If the study objectives are met, the AI-based software will be offered as a not-for-profit web service to public sector state hospitals in LMICs to support expansion of high quality radiotherapy capacity, improving access to and affordability of this key modality of cancer cure and control. Public and policy engagement plans will involve patients as key partners.

Federated benchmarking of medical artificial intelligence with MedPerf.

Medical artificial intelligence (AI) has tremendous potential to advance healthcare by supporting and contributing to the evidence-based practice of medicine, personalizing patient treatment, reducing costs, and improving both healthcare provider and patient experience. Unlocking this potential requires systematic, quantitative evaluation of the performance of medical AI models on large-scale, heterogeneous data capturing diverse patient populations. Here, to meet this need, we introduce MedPerf, an open platform for benchmarking AI models in the medical domain. MedPerf focuses on enabling federated evaluation of AI models, by securely distributing them to different facilities, such as healthcare organizations. This process of bringing the model to the data empowers each facility to assess and verify the performance of AI models in an efficient and human-supervised process, while prioritizing privacy. We describe the current challenges healthcare and AI communities face, the need for an open platform, the design philosophy of MedPerf, its current implementation status and real-world deployment, our roadmap and, importantly, the use of MedPerf with multiple international institutions within cloud-based technology and on-premises scenarios. Finally, we welcome new contributions by researchers and organizations to further strengthen MedPerf as an open benchmarking platform.