Rohan Sanghera

Rohan Sanghera

Auto-generated profile, based on publicly available information

Rohan Sanghera

School of Clinical Medicine, University of Cambridge, Cambridge, UK

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
21 April 2023 in JMIR Medical Education
Verified
Journal Article
Open access
ATArun James Thirunavukarasu
Arun James Thirunavukarasu
RHRefaat Hassan
Refaat Hassan
SMShathar Mahmood
Shathar Mahmood
RSRohan Sanghera
Rohan Sanghera
KBKara Barzangi
Kara Barzangi
SSSachin Shah
Sachin Shah
Background: Large language models exhibiting human-level performance in specialized tasks are emerging; examples include Generative Pretrained Transformer 3.5, which underlies the processing of ChatGPT. Rigorous trials are required to understand the capabilities of emerging technology, so that innovation can be directed to benefit patients and practitioners. Objective: Here, we evaluated the strengths and weaknesses of ChatGPT in primary care using the Membership of the Royal College of General Practitioners Applied Knowledge Test (AKT) as a medium. Methods: AKT questions were sourced from a web-based question bank and 2 AKT practice papers. In total, 674 unique AKT questions were inputted to ChatGPT, with the model’s answers recorded and compared to correct answers provided by the Royal College of General Practitioners. Each question was inputted twice in separate ChatGPT sessions, with answers on repeated trials compared to gauge consistency. Subject difficulty was gauged by referring to examiners’ reports from 2018 to 2022. Novel explanations from ChatGPT—defined as information provided that was not inputted within the question or multiple answer choices—were recorded. Performance was analyzed with respect to subject, difficulty, question source, and novel model outputs to explore ChatGPT’s strengths and weaknesses. Results: Average overall performance of ChatGPT was 60.17%, which is below the mean passing mark in the last 2 years (70.42%). Accuracy differed between sources (P=.04 and .06). ChatGPT’s performance varied with subject category (P=.02 and .02), but variation did not correlate with difficulty (Spearman ρ=–0.241 and –0.238; P=.19 and .20). The proclivity of ChatGPT to provide novel explanations did not affect accuracy (P>.99 and .23). Conclusions: Large language models are approaching human expert–level performance, although further development is required to match the performance of qualified primary care physicians in the AKT. Validated high-performance models may serve as assistants or autonomous clinical tools to ameliorate the general practice workforce crisis.

17 April 2024 in PLOS Digital Health
Verified
Journal Article
Open access
PDF file
ATArun James Thirunavukarasu
Arun James Thirunavukarasu
SMShathar Mahmood
Shathar Mahmood
AMAndrew Malem
Andrew Malem
WFWilliam Paul Foster
William Paul Foster
RSRohan Sanghera
Rohan Sanghera
DTDarren Shu Jeng Ting
Darren Shu Jeng Ting
Large language models (LLMs) underlie remarkable recent advanced in natural language processing, and they are beginning to be applied in clinical contexts. We aimed to evaluate the clinical potential of state-of-the-art LLMs in ophthalmology using a more robust benchmark than raw examination scores. We trialled GPT-3.5 and GPT-4 on 347 ophthalmology questions before GPT-3.5, GPT-4, PaLM 2, LLaMA, expert ophthalmologists, and doctors in training were trialled on a mock examination of 87 questions. Performance was analysed with respect to question subject and type (first order recall and higher order reasoning). Masked ophthalmologists graded the accuracy, relevance, and overall preference of GPT-3.5 and GPT-4 responses to the same questions. The performance of GPT-4 (69%) was superior to GPT-3.5 (48%), LLaMA (32%), and PaLM 2 (56%). GPT-4 compared favourably with expert ophthalmologists (median 76%, range 64–90%), ophthalmology trainees (median 59%, range 57–63%), and unspecialised junior doctors (median 43%, range 41–44%). Low agreement between LLMs and doctors reflected idiosyncratic differences in knowledge and reasoning with overall consistency across subjects and types (p>0.05). All ophthalmologists preferred GPT-4 responses over GPT-3.5 and rated the accuracy and relevance of GPT-4 as higher (p<0.05). LLMs are approaching expert-level knowledge and reasoning skills in ophthalmology. In view of the comparable or superior performance to trainee-grade ophthalmologists and unspecialised junior doctors, state-of-the-art LLMs such as GPT-4 may provide useful medical advice and assistance where access to expert ophthalmologists is limited. Clinical benchmarks provide useful assays of LLM capabilities in healthcare before clinical trials can be designed and conducted.

18 May 2024 in npj Digital Medicine
Verified
Journal Article
Open access
PDF file
ATArun James Thirunavukarasu
Arun James Thirunavukarasu
NJNikhil Jain
Nikhil Jain
RSRohan Sanghera
Rohan Sanghera
FLFederico Lattuada
Federico Lattuada
SMShathar Mahmood
Shathar Mahmood
RBRupert Bourne
Rupert Bourne
Subjectivity and ambiguity of visual field classification limits the accuracy and reliability of glaucoma diagnosis, prognostication, and management decisions. Standardised rules for classifying glaucomatous visual field defects exist, but these are labour-intensive and therefore impractical for day-to-day clinical work. Here a web-application, Glaucoma Field Defect Classifier (GFDC), for automatic application of Hodapp-Parrish-Anderson, is presented and validated in a cross-sectional study. GFDC exhibits perfect accuracy in classifying mild, moderate, and severe glaucomatous field defects. GFDC may thereby improve the accuracy and fairness of clinical decision-making in glaucoma. The application and its source code are freely hosted online for clinicians and researchers to use with glaucoma patients.

18 October 2019 in Cureus
Verified
Journal Article
Open access
PDF file
CSChandan Sanghera
Chandan Sanghera
RSRohan Sanghera
Rohan Sanghera
Immunotherapy is widely regarded to have the ability to transform the treatment of cancer, with immune checkpoint inhibitors already in use for cancers such as advanced melanoma and non-small cell lung cancer (NSCLC). However, despite its potential, the widespread adoption of immunotherapy for the treatment of other cancers has been largely limited. This can be partly attributed to additional immunosuppressive mechanisms in the tumor microenvironment that help promote and maintain a state of T cell exhaustion. As such, the exploration of combinatory immunotherapies is an active area of research and includes the combination of immune checkpoint inhibitors with cytotoxic therapies, cancer vaccines and monoclonal antibodies against other co-inhibitory and co-stimulatory receptors. Strategies are also being employed to improve the homing, extravasation and survival of chimeric antigen receptor (CAR)-T cells in the tumor microenvironment. Furthermore, the development of immunotherapies targeted to one or multiple neoantigens unique to a specific tumor may act to enhance anti-tumor immunity, as well as reduce immune-related adverse events (irAEs). As immunotherapy evolves to become a mainstay treatment for cancer, it is imperative that optimum treatment regimens that maximize efficacy and limit toxicity are developed. Foremost, appropriate biomarkers must be identified to help tailor combinatory immunotherapies to the individual patient and hence pave the way to a new era of personalized medicine.

7 May 2025 in British Journal of Ophthalmology
Verified
Journal Article
Closed access
PDF file
ATArun James Thirunavukarasu
Arun James Thirunavukarasu
NJNikhil Jain
Nikhil Jain
HYHelmut C Y Yu
Helmut C Y Yu
GNGeorge Nishimura
George Nishimura
ATAnsh Tandon
Ansh Tandon
RBRupert R A Bourne
Rupert R A Bourne
Background/aims Many countries provide social support to patients with severe sight impairment (blindness). In the UK, these benefits require a certificate of visual impairment (CVI) which requires referral by a consultant ophthalmologist. Many eligible patients do not receive a CVI due to personal choice or lack of consideration or communication by their doctor. This study investigated the frequency and reasons for missed certification in glaucoma. Methods A retrospective cross-sectional survey was undertaken of all patients with glaucoma attending a tertiary referral clinic over a 12-month period. Semi-automated screening using objective visual acuity and perimetry parameters was facilitated by a web application, GFDC (Glaucoma Field Defect Classifier). CVI-eligible patients’ records were analysed to determine the reasons for missed registration, including incorrect screening classification. Results Of 5620 individual patients consulted, 919 were classified as sight impaired, and 64 were classified as severely sight impaired (blind). Of the patients classified as blind, 7 (11%) were misclassified, and 36 (56%) were on the CVI register already. 21 of 57 eligible patients with glaucoma-induced blindness (37%) were unregistered. Reasons for missed registration included administrative failure (23.8%), lack of consent (9.5%), reversible visual impairment (19.0%), frailty and comorbidity (71.4%), and mental health diagnoses (38.1%). Conclusion A semi-automated algorithm can be used to screen large numbers of patients for CVI eligibility due to blindness. Many eligible patients are not registered, with risk factors including frailty, comorbidity and reversible causes of visual impairment. This algorithm could be used to prompt ophthalmologists to consider registration or used as an alternative referral mechanism. Screening for CVI-eligible patients with an objective algorithm may ameliorate the inequity associated with subjective and variable decision-making.

6 August 2023 in medRxiv
Verified
Preprint
Open access
PDF file
ATArun James Thirunavukarasu
Arun James Thirunavukarasu
SMShathar Mahmood
Shathar Mahmood
AMAndrew Malem
Andrew Malem
WFWilliam Paul Foster
William Paul Foster
RSRohan Sanghera
Rohan Sanghera
DTDarren Shu Jeng Ting
Darren Shu Jeng Ting
Objective: To evaluate the clinical potential of large language models (LLMs) in ophthalmology using a more robust benchmark than raw examination scores. Materials and methods: GPT-3.5 and GPT-4 were trialled on 347 questions before GPT-3.5, GPT-4, PaLM 2, LLaMA, expert ophthalmologists, and doctors in training were trialled on a mock examination of 87 questions. Performance was analysed with respect to question subject and type (first order recall and higher order reasoning). Masked ophthalmologists graded the accuracy, relevance, and overall preference of GPT-3.5 and GPT-4 responses to the same questions. Results: The performance of GPT-4 (69%) was superior to GPT-3.5 (48%), LLaMA (32%), and PaLM 2 (56%). GPT-4 compared favourably with expert ophthalmologists (median 76%, range 64-90%), ophthalmology trainees (median 59%, range 57-63%), and unspecialised junior doctors (median 43%, range 41-44%). Low agreement between LLMs and doctors reflected idiosyncratic differences in knowledge and reasoning with overall consistency across subjects and types ( p >0.05). All ophthalmologists preferred GPT-4 responses over GPT-3.5 and rated the accuracy and relevance of GPT-4 as higher ( p <0.05). Discussion: In view of the comparable or superior performance to trainee-grade ophthalmologists and unspecialised junior doctors, state-of-the-art LLMs such as GPT-4 may provide useful medical advice and assistance where access to expert ophthalmologists is limited. Clinical benchmarks provide useful assays of LLM capabilities in healthcare before clinical trials can be designed and conducted. Conclusion: LLMs are approaching expert-level knowledge and reasoning skills in ophthalmology. Further research is required to develop and validate clinical applications to improve eye health outcomes.