Medical Diagnostics R&D

Bostel Technologies

SUMMARY

• Bostel sought to explore and develop an ML-powered approach for diagnosing skin cancer.

• It was unclear whether ML could derive from dermatoscopic imagery a clinical-grade diagnosis and communicate results appropriately to a clinician.

• Hop acted as trusted technical partner, ensuring work was rigorous enough to satisfy regulatory requirements and investors, and building production-level infrastructure for the clinical trial process.

• This partnership resulted in a successful clinical trial, several publications, and multiple issued patents. 

THE COMPANY

Bostel Technologies is a clinical-stage skin cancer diagnostic company, with founders and investors from the dermatology and pharmaceutical fields. At the time Bostel engaged Hop’s services, the leadership team had identified an opportunity for a novel diagnostic approach powered by machine learning, and sought to leverage their deep clinical network to validate and meet an overwhelming market demand.

THE CHALLENGE

Bostel faced a variety of interesting and novel technical challenges on the path to commercialization. Key amongst them: could machine learning be used to provide clinical-grade sensitivity and specificity for early-stage melanoma detection from dermatoscopic imagery? And if so, could the nuanced results of the model be communicated in a clinically appropriate way to the clinician without interfering with the patient experience?

Underlying these technical challenges was a deeper organizational one: although Bostel had deep domain expertise, they needed a trusted technical partner to ensure their technical work was rigorous, professional, and would clear any necessary regulatory approval or investor due-diligence processes.

THE SOLUTION

Hop partnered with Bostel and various clinical and academic collaborators to ensure the R&D process was both thorough and rigorous. Throughout this process, Hop kept the Bostel executive team informed with regular research reports. For their IP to be commercially viable, Bostel needed an approach that was both novel and able to meet key business targets.

Once a promising approach was identified in the laboratory, Hop developed the infrastructure necessary to evaluate the technology in both clinical and usability trials. Given the critical needs of clinical use, Hop’s experience with production machine learning was invaluable in ensuring that the technology stack was reliable and resilient enough for a smooth and successful clinical trial. 

THE RESULTS

Bostel and Hop’s collaboration resulted in a successful clinical trial, several publications [1-4] in top-tier and high-impact venues (Lancet, ICAD, others), and multiple issued patents. The commercialization of this technology is currently underway with several key partners.

[1] Walker, B. N., et al. Dermoscopy diagnosis of cancerous lesions utilizing dual deep learning algorithms via visual and audio (sonification) outputs: Laboratory and prospective observational studies, The Lancet, EBioMEdicine; February 2019, Volume 40, Pages 176–183.

[2] Dascalu, A., et al. Skin cancer detection by deep learning and sound analysis algorithms: A prospective clinical study of an elementary dermoscope. The Lancet, EBioMEdicine; May 2019, Volume 43, Pages 107–113.

[3] Walker, B. N., et al. Hearing artificial intelligence: Sonification guidelines & results from a case-study in melanoma diagnosis and prospective observational studies. Proceedings of the 25th International Conference on Auditory Display (ICAD 2019) 23-27 June 2019, Northumbria University, Newcastle upon Tyne, UK.

[4] Dascalu, A., Walker, B.N., Oron, Y. et al. Non-melanoma skin cancer diagnosis: a comparison between dermoscopic and smartphone images by unified visual and sonification deep learning algorithms. Journal of Cancer Research and Clinical Oncology (2021). https://doi.org/10.1007/s00432-021-03809-x.

Looking for a technical partner to bring your research idea to life? Contact us to learn how Hop can help.