💡 Article Title: Nabla Raises $24 Million to Develop AI Copilot for Doctors

Paris-Based Startup Nabla Announces $24 Million Series B Funding Round for its AI Doctor Copilot

Nabla brings in $24 million more for its AI doctor assistant that writes clinical notes automatically | ENBLE


Paris-based startup Nabla has recently secured a staggering $24 million in a Series B funding round led by Cathay Innovation, with participation from ZEBOX Ventures, the corporate VC fund of CMA CGM. This funding round comes not long after Nabla signed a significant partnership agreement with Permanente Medical Group, a division of the renowned U.S. healthcare giant, Kaiser Permanente. With this latest financial boost, Nabla has now achieved a valuation of $180 million. The company may also secure further investments from U.S. investors during this funding round.

🤖 A Game-Changing AI Copilot

Nabla has been tirelessly working on the development of an AI copilot for doctors and other medical staff. The best way to envision this innovation is as a quiet and diligent work partner that sits in the corner of the room, diligently taking notes and crafting medical reports.

The masterminds behind this ingenious startup are Alexandre Lebrun, Delphine Groll, and Martin Raison. Lebrun, the CEO of Nabla, was also the CEO of Wit.ai, an AI assistant startup which Facebook acquired. Later, he became the head of engineering at Facebook’s AI research lab, FAIR.

🩺 How Does Nabla Work?

A few weeks ago, I had the privilege of witnessing a live demo of Nabla in action. A real doctor and a fake patient, pretending to have back pains, conducted a consultation. The doctor simply hit the start button on Nabla’s interface, and then forgot about their computer throughout the consultation.

During the consultation, besides the physical examination, there was an extensive discussion that covered various subjects such as the patient’s medical history and the reason for their visit. At the end of the consultation, recommendations and prescriptions may be given.

Nabla utilizes speech-to-text technology to transform the conversation into a written transcript. This incredible tool works seamlessly for both in-person consultations and telehealth appointments.

Once the patient has left, the doctor presses the stop button. Nabla then employs a sophisticated language model, refined with medical data and health-related conversations, to identify critical data points in the consultation. This includes medical vitals, drug names, and pathologies.

Within a mere minute or two, Nabla generates a comprehensive medical report that encompasses a summary of the consultation, prescriptions, and follow-up appointment letters. Furthermore, doctors can customize these reports according to their specific needs, allowing for a personalized format. For instance, they can request more concise or more detailed notes. Alternatively, they can even ask Nabla to generate notes following the widely-used SOAP note pattern (Subjective, Objective, Assessment, and Plan).

My initial skepticism about the effectiveness of Nabla quickly dissipated when I witnessed its accuracy firsthand. Even in a crowded room with Nabla running on a laptop a few meters away from the demo presenters, this extraordinary tool efficiently generated an accurate transcript and a useful report.

🕹️ Nabla Copilot: A Time-Saving Assistant

Contrary to popular fears of replacing doctors, Nabla’s objective is not to eliminate the human element from the medical loop. Rather, the company aims to alleviate doctors from excessive administrative work, enabling them to focus more on patient care.

“We don’t want to try to replace doctors. You’ve seen companies like Babylon burning $1 billion trying to do chatbots and trying to automate things right away and remove doctors from the loop,” Lebrun emphasized. “With Nabla Copilot, doctors are the pilots, and we work by their side.”

Lebrun brilliantly compared this collaboration to automation in autonomous vehicles. Currently, we are at level two, with Nabla soon approaching level three with clinical assurance support. Level four involves clinical decision support, but it requires FDA approval due to the complexity of certain medical decisions that cannot be easily explained.

Although Lebrun cautiously acknowledged that there could be situations or markets where autonomous healthcare without physicians might be relevant, he firmly believes that in most cases, doctors’ role should always involve empathy, surgical procedures, and critical decisions.

While Nabla is based in France, the majority of its customers are situated in the United States. Following its successful implementation across Permanente Medical Group, thousands of doctors actively utilize Nabla on a daily basis.

🔒 Nabla’s Privacy Model

Given the sensitive nature of the data it handles, Nabla is acutely aware of the importance of data protection. Consequently, the company prudently refrains from storing audio or medical notes on its servers without the explicit consent of both the doctor and the patient.

Nabla’s focus lies in data processing rather than data storage. Once a consultation concludes, the audio file is discarded, while the transcript is stored securely in the Electronic Health Record (EHR) system that doctors already utilize for their patient files.

To achieve this, Nabla uses a combination of both an off-the-shelf speech-to-text API from Microsoft Azure and its own speech-to-text model, a refined adaptation based on the open-source Whisper model. Nabla’s ML engineer, Grégoire Retourné, explained that the text gradually darkens during the transcription process, indicating that it has been verified and corrected with medication names or medical conditions using Nabla’s fine-tuned speech-to-text algorithm.

After the transcript is pseudonymized (replacing personally identifiable information with variables), it goes through processing by a large language model (LLM). Historically, Nabla has primarily used GPT-3 and GPT-4. As an enterprise customer, Nabla has stipulated to OpenAI that its data cannot be stored or used for training the language model.

Furthermore, Nabla is experimenting with a fine-tuned version of Llama 2. Lebrun envisions employing narrow models more frequently in the future, emphasizing their widespread usage for specific medical purposes.

Once the LLM has completed processing the transcript, Nabla de-pseudonymizes the output. Doctors can access and securely store the notes on their computers in the local web browser storage file, with the ability to export notes to EHRs.

Additionally, doctors have the option to grant approval and seek patient consent to share medical notes with Nabla for the purpose of correcting transcription errors. Nabla’s growing dataset from processing over 3 million consultations per year, in three languages, will undoubtedly lead to substantial improvement and refinement of its services.


💡 Q&A

Q1: Is Nabla aiming to replace doctors with its AI copilot?

Nabla is determined not to replace doctors but rather empower and assist them in their medical practice. The goal is to streamline administrative work and free up valuable time for doctors to focus more on patient care.

Q2: Does Nabla store audio or medical notes on its servers?

No, Nabla prudently refrains from storing such sensitive data without explicit consent from both the doctor and patient. Audio files are discarded after consultations, while the securely pseudonymized transcripts are stored in the doctors’ existing Electronic Health Record (EHR) systems.

Q3: How does Nabla ensure the accuracy and quality of its transcription?

Nabla utilizes a combination of an off-the-shelf speech-to-text API and its own refined speech-to-text model. The model is fine-tuned for medical data, ensuring accurate and reliable transcriptions. Additionally, doctors have the option to review and edit the generated notes before they are filed in the EHR system.

🌍 Future Implications

Nabla’s groundbreaking AI copilot for doctors has immense potential to revolutionize the healthcare industry. As technology continues to advance, we may witness future developments that could eventually lead to autonomous healthcare. However, Nabla’s CEO, Alexandre Lebrun, remains cautious about removing physicians from the equation entirely. For now, Nabla intends to complement doctors by providing invaluable support, allowing them to focus more on empathetic patient care, intricate surgical procedures, and critical medical decisions.

With Nabla’s presence predominantly in the United States, the startup is well-positioned to make significant strides in transforming the healthcare experience for doctors and patients alike. As Nabla continues to process millions of consultations and expand its dataset, we can anticipate rapid improvement and increased reliability in its AI copilot technology.

📚 References

✅ If you found this article informative, please share it on social media and continue the conversation.

[Facebook] [Twitter] [LinkedIn]