The artificial intelligence in healthcare is revolutionizing the industry transforming everything from surgical procedures and medical image AI analysis to disease diagnostics and even state-level healthcare management. 

AI in medicine offers unprecedented precision, efficiency, and patient-oriented assistance by providing advanced decision support tools and systems. It enhances diagnostic accuracy, optimizes treatment strategies, and improves clinical outcomes.

AI tools have the potential to surpass human capabilities in various healthcare aspects, utilizing vast datasets to enhance accuracy, reduce costs, save time, minimize errors, and revolutionize personalized medicine.

Before delving into the practical application of health and tech symbiosis, it’s essential to assess the tremendous market growth potential, as illustrated here with the example of GenAI.

Source: Lemberg Solutions

Use Cases

The integration of Artificial Intelligence into healthcare is revolutionizing patient care and service delivery, offering significant benefits for both patients and healthcare providers. There are diverse use cases and areas where AI’s practical applications have a substantial impact, addressing industry-specific needs and enhancing operational efficiency.

By late 2023 and early 2024, the narrative of multimodality gained even more momentum in the AI sphere. Considering that medicine itself is essentially multimodal, this aligns perfectly with the canvas of practical application.

Sources of “signal” include textual and vocal descriptions, medical documentation, and established treatment protocols, research archives, medical information from photos, electronic health records, sensors, microphones, wearable devices, genomics code, and more.

The value of practical decisions depends on their ability to be suitable for integration with a medical data center, ensuring seamless incorporation of AI into healthcare processes.

1. Text & Language Processing

  • Recognition of diagnoses based on textual/audial descriptions
  • Identification of keywords or key points in symptom descriptions
  • Structuring, summarizing, and detecting key points and events in

textual descriptions

  • Symptom search in free-text, particularly for call centers

2. Alerts, Risk Predictions and the selection of optimal treatment strategies and tactics.

  • Early determination and classification of risks
  • Search for co-existing diseases

For instance, Google’s DeepMind division has pioneered an AI-powered system capable of forecasting acute kidney injury (AKI) in hospitalized patients up to 48 hours in advance. This case has been ongoing for 5 years now.

Particularly powerful (albeit complex) are the AI use cases in genetic computations. Genetics is closely related to mathematics, thus leveraging models in genetic analyses and forecasts (inheritance, etc.) appears judicious. Analyzing genetic predispositions, potential health risks, and personalized medicine options based on an individual’s genetics.

A schematic example of modelling in genetics. Source: BioMed Central

Tempus Labs has developed a platform that utilizes machine learning to analyze genomic data and identify patients at risk of developing specific diseases.

A powerful case study emerges when individual private datasets about a particular patient are thoughtfully combined (such as an online medical record containing all previously conducted analyses and descriptions, individual body characteristics, and adaptation to the necessary protocol and treatment course, therapeutic drug doses, etc.).

This also applies to individual rehabilitation programs for sick and injured individuals according to their historical data pool.

Collecting, analyzing, and interpreting patient vital signs and health indicators

1/ ECG Analysis & Interpretation

  • Forecasting heart abnormalities
  • Multi-label ECG classification
  • End-to-end risk prediction of atrial fibrillation (via Deep Neural Networks), including combination of ECG + respiratory modulation estimation.
  • Prediction of vascular aging and its correlation with smoking habits 
  • Textual interpretation of key ECG data flow
  • Detection of ECG noiseand alerting patients or medical personnel
  • Non-invasive detection of hyperglycemia using ECG and Deep Learning

2/ AI in Medical Imaging, Visual and Audio Investigations Analysis and Summarizing: Ultrasound (US), X-ray, Fluoroscopy, Computerized Tomography (CT), Magnetic Resonance Imaging (MRI), Dermatoscopic examinations, microbiological and histological investigations (AI processing of microscopy images), and other examinations involving the collection and processing of visual information.

By training models on specialized datasets of images and analyses, one can achieve high accuracy, often eliminating human factors and errors, especially in recognizing small pathologies on images. In this context, the example of “contouring” CT scans for precise radiation therapy calculation in oncological practice is often cited. This process can take up to 7 hours for a single patient.

Med-PaLM by Google. Source:

Another study has shown that the use of artificial intelligence for mammogram readings allows for the detection of 20% more cases of cancer.

Regarding sound, the spectrum of AI use cases ranges from collecting and processing audio signals from household microphones in smartphones (virtual stethoscopes / phonendoscopes) to employing professional systems for more in-depth investigations.

3/ Processing and deciphering data collected from blood and other substances involve combining a vast array of normative indicators for specific genders/age groups/etc., along with patient-specific data over time, representing one of the most apparent yet potent AI use cases.

4/ Health Trackers and specialized sensors

Examples include diabetic trackers, process automation of pregnancy diaries, pulse meters, and blood pressure analyzers, temperature sensors, oxygenation, etc.

In this scenario, the spectrum of use cases is exceedingly broad, ranging from everyday Health Trackers to specialized platforms designed for niche medical research purposes. AI plays a pivotal role here, intertwined with embedded technologies and, in many instances, with IoT integration.

It is worth highlighting the extensive applications of such analytical models in forensic medicine. The thematic exploration of “AI Sherlock” use cases deserves for a dedicated article.

Additionally, it is crucial to underscore the must-have core feature that should be inherent in medical AI solutions across all the aforementioned diagnostic domains. This feature must be meticulously developed and proven to achieve high accuracy in performance:

Anomaly & Signal Detection:

  • Anomaly detection within textual and signal data
  • Signal processing and AI-driven data mining within device measurements

3. Physical Assistance in Surgery

Focus: Revolutionizing preoperative planning, intraoperative assistance, and postoperative care.

The promising fusion of robotics and AI entails executing ultra-precise surgeries through robotic surgery, minimally invasive micro- and nanosurgery (controlled surgical microbots with AI onboard).

It is anticipated that with an average annual growth rate of 15.7%, the market for robotic surgical systems utilizing AI technology will reach $7.2 billion by 2033.

4. Specialized Pharmacological Modeling

  • Development and “Fine-tuning” of Medications
  • Acceleration and Optimization of Clinical Laboratory Studies

According to the McKinsey Global Institute, machine learning (ML) and artificial intelligence (AI) in the pharmaceutical sector have the potential to contribute approximately $100 billion annually to the U.S. healthcare system.

One of the most prominent examples is the accelerated development and testing of the next generation of COVID-19 vaccines (previously, such testing took years and decades).

5. Technical + Software Solutions for Mobile Healthcare

Google, in its research, highlights the shift towards mobile medicine. As a result, numerous startups and working groups worldwide are developing mobile ML solutions that can be brought to the patient, rather than bringing the patient to the clinic (ML-powered).

According to Google Research, an inexpensive ultrasound device powered by batteries and a smartphone was tested, demonstrating accuracy comparable to existing clinical standards for professional sonographers in diagnosing fetal indicators.

Source: Google Research

Another important use case involves the utilization of AI-based solutions (often in conjunction with hardware components) for emergency and field medicine (use in combat zones, emergency situations, etc.).

6. Use in the Educational Process in Medical Universities and for Assessing the Knowledge of Doctors and Students

The discussion primarily revolved around the testing process of the Google Med-PaLM 2 model. Questions were posed to the model, answers were cross-checked, and calibration was conducted.

These AI use cases extend beyond just education and knowledge assessment, offering facilitation through a certain level of simplification and a form of gamification. Another benefit lies in using such medical AI solutions to create credible educational simulations.

Patient & Data Management, Administrative Application

  • Grouping of patients
  • Matching doctors based on symptom descriptions
  • Data collection from doctors and its standardization
  • Call center support for psychologists, providing real-time conversation scripts for patient interactions
  • Managing logistics and operational processes at the level of medical institutions
  • Assistant for HR, Corporate wellness, and compliance with medical regulations for employers

This also opens up a wide range of practical applications: from CV + OCR to the analysis and management of textual information through NLP / LLM or proprietary models.

Optimizing the workload of professionals who provide basic consultations, register patients, schedule appointments, and perform various reception functions. In this case, practical use cases include a spectrum from basic chatbots to advanced multimodal agents with various API enhancements.

Regard, previously known as HealthTensor, a machine learning-powered tool, can analyze patient data to identify patterns that may indicate the presence of co-existing diseases. They have now shifted their focus towards automating clinical tasks for clinicians and admins. Through integration with EHR, Regard scans and organizes the patient’s entire medical history, assisting doctors in making data-driven decisions. Another focus of this medical AI solution is hospital finances, patient safety, coding queries, insurance document processing, etc.

7. Telemedicine and Mobile Applications with Multimodal LLM Agents (Assistants Under the Hood)

  • “The Young Doctor’s Guide”
  • Everyday Medical Question-Answering, Symptom Checker, Health Advisor, medical “translators” and interpreters, etc.
  • Medicine + dietetics: selection of permissible nutrition
  • Mobile applications/agents to improve quality of life for certain conditions such as “Allergy Assistant,” “Diabetes Assistant”
  • Individual programmed calendar for screenings and health monitoring
  • Virtual assistants for the care of elderly and people with special needs, etc.
  • Search and selection of service providers, such as hospitals (Geolocation-based integration with maps and directories).
  • Mental Health App. This topic has continued to gain popularity over the past few years.

Closing Remarks

Before concluding, it’s worth noting that all of this has long surpassed the realms of experimentation and theorizing. Both industry giants like Google and smaller startups are actively engaged in projects at the intersection of Healthtech and AI.

A glance at the Google Med-PaLM 2 presentation reveals that with a model accuracy of 67.2%, it outpaced its closest competitors by a significant margin (considering Google’s penchant for cherry-picking in public presentations).

Source: Google Research

In the near future, it will be intriguing to witness a new set of competitors and their metrics of accuracy and efficiency. Over the past year since this presentation, model capabilities have significantly advanced, partly due to accelerated self-improvement, self-induction, and AutoML.

Utilizing the synergy of AI and healthcare extends beyond mere process automation and cost reduction.  It entails comprehensive enhancement of the effectiveness of tools and resources aimed at preserving and improving human health. It also anticipates breakthroughs and evolution in the development of medical technologies.

Are you in need of AI+Healthtech solutions? Book a quick intro call. We’re confident we have something valuable to offer.


Author: azakharchenko