Post Reply
1-20 of 413
https://www.youtube.com/watch?v=wxx-i_H_WZI
The clever part of Alibaba DAMO PANDA is that it is not simply looking for a big, obvious pancreatic tumour. It is trained to find very subtle changes in ordinary CT images, including non-contrast CT, where pancreatic cancer can be extremely difficult for humans to see.
The original research describes PANDA as a three-stage deep-learning system.
The AI first asks:
This is important because the pancreas is relatively small and surrounded by other organs. The system can then concentrate its computational attention on the pancreatic region rather than analysing the entire CT equally.
Pancreatic cancer can produce subtle changes such as:
It isn' t thinking:
It can distinguish between:
PDAC
(pancreatic ductal adenocarcinoma)
and other pancreatic lesions such as:
Normally, doctors can use contrast-enhanced CT because contrast makes blood vessels and many tumours much easier to distinguish.
But PANDA' s researchers wanted the AI to work with non-contrast CT, which is much more commonly available and doesn' t require contrast injection.
So they used a clever training technique.
Contrast CT
&darr
Radiologists identify/annotate the tumour
&darr
Computer algorithm maps the tumour location onto the corresponding non-contrast CT
&darr
AI learns:
So the AI essentially learned to see signals humans don' t routinely perceive in the non-contrast image.
A patient might come into hospital for something completely different.
For example:
Diabetes / abdominal discomfort / pneumonia
&darr
Doctor orders ordinary CT
&darr
Radiologist examines CT for the original clinical question
&darr
Nothing obviously suspicious for pancreatic cancer
&darr
PANDA examines the same CT
&darr
🚨 Pancreatic abnormality detected
&darr
Doctor investigates
&darr
Contrast CT / MRI / EUS / biopsy as appropriate
&darr
Cancer confirmed
&darr
Treatment.
At Ningbo University Affiliated People' s Hospital, PANDA reportedly analysed more than 180,000 abdominal/chest CTs and helped identify about two dozen pancreatic cancers, including 14 early-stage cases.
It doesn' t look at a scan and consciously say:
Conceptually:
Millions of image features
&darr
learned patterns
&darr
probability of lesion
&darr
probability of PDAC / other lesion
&darr
location/segmentation
&darr
alert to doctor
The Nature Medicine study says PANDA outputs both a segmentation mask of the pancreas/lesion and classification probabilities.
So the doctor can see where the AI thinks the abnormality is, rather than receiving merely:
The CT scan already exists.
You don' t necessarily need:
❌ new CT appointment
❌ new radiation exposure
❌ expensive screening programme
❌ specialist looking at every scan again
Instead:
&darr
AI re-analyses them
&darr
small number of suspicious cases
&darr
human specialist reviews
&darr
additional testing only for those patients.
That' s opportunistic cancer detection.
The research specifically explored routine abdominal and chest non-contrast CTs across physical examination, emergency, outpatient and inpatient settings.
The PANDA Pro version is designed to improve differentiation between pancreatic cancer, pancreatitis and pancreatic cystic lesions and reduce interference from neighbouring structures such as the bile duct and duodenum.
DAMO' s current medical-AI platform also reports pancreatic screening performance of approximately 92.9% sensitivity and 99.9% specificity in its published/marketed evaluation metrics.
But don' t interpret 99.9% specificity as " 99.9% accurate diagnosis." These are different statistical measures, and real-world deployment still requires physician confirmation and further diagnostic testing.
Not just pancreas.
The same architecture could potentially screen for:
Pancreas &rarr liver &rarr lung &rarr kidney &rarr colon &rarr cardiovascular abnormalities &rarr other cancers
DAMO itself is already describing a broader multi-cancer direction using non-contrast CT.
Then you add an AI agent:
It becomes an AI-powered early-detection hospital.
And that, in my view, is potentially much more strategically important to Raffles Medical than humanoid robots.
Think of PANDA as a 3-step visual system
CT scan &rarr find pancreas &rarr find abnormal tissue &rarr classify itThe original research describes PANDA as a three-stage deep-learning system.
1. First, it finds the pancreas
A CT contains hundreds of images/slices showing the whole abdomen or chest.The AI first asks:
" Where exactly is the pancreas?"It creates a segmentation mask around the pancreas.
This is important because the pancreas is relatively small and surrounded by other organs. The system can then concentrate its computational attention on the pancreatic region rather than analysing the entire CT equally.
2. Then it looks for tiny visual abnormalities
This is the really impressive part.Pancreatic cancer can produce subtle changes such as:
- very small differences in tissue texture
- changes in the pancreatic contour
- small masses
- distortion of normal pancreatic structure
- abnormalities around the pancreas
- changes that may be difficult to distinguish from surrounding tissue
It isn' t thinking:
" I see a 2-cm tumour."Instead, it is more like:
" The statistical pattern of pixels/voxels in this region looks different from thousands of examples of normal pancreas."That distinction is extremely important.
3. It then asks: " What is this abnormality?"
If PANDA detects something suspicious, a third network attempts to classify it.It can distinguish between:
PDAC
(pancreatic ductal adenocarcinoma)
and other pancreatic lesions such as:
- pancreatic neuroendocrine tumour
- IPMN
- mucinous cystic neoplasm
- serous cystic neoplasm
- solid pseudopapillary tumour
- chronic pancreatitis
- other lesions.
But how can it do this from a " bad" non-contrast CT?
This is probably the most fascinating part.Normally, doctors can use contrast-enhanced CT because contrast makes blood vessels and many tumours much easier to distinguish.
But PANDA' s researchers wanted the AI to work with non-contrast CT, which is much more commonly available and doesn' t require contrast injection.
So they used a clever training technique.
They had more than 2,000 pancreatic patients.
For these patients they had:Contrast CT
&darr
Radiologists identify/annotate the tumour
&darr
Computer algorithm maps the tumour location onto the corresponding non-contrast CT
&darr
AI learns:
" This subtle pattern on the ordinary CT corresponds to a confirmed pancreatic lesion."The Nature Medicine paper describes this transfer of lesion annotations from contrast-enhanced CT to non-contrast CT.
So the AI essentially learned to see signals humans don' t routinely perceive in the non-contrast image.
This explains the Ningbo cases
This is why the Ningbo deployment is so interesting.A patient might come into hospital for something completely different.
For example:
Diabetes / abdominal discomfort / pneumonia
&darr
Doctor orders ordinary CT
&darr
Radiologist examines CT for the original clinical question
&darr
Nothing obviously suspicious for pancreatic cancer
&darr
PANDA examines the same CT
&darr
🚨 Pancreatic abnormality detected
&darr
Doctor investigates
&darr
Contrast CT / MRI / EUS / biopsy as appropriate
&darr
Cancer confirmed
&darr
Treatment.
At Ningbo University Affiliated People' s Hospital, PANDA reportedly analysed more than 180,000 abdominal/chest CTs and helped identify about two dozen pancreatic cancers, including 14 early-stage cases.
The important thing: PANDA does NOT " see cancer" like a human
This is worth understanding.It doesn' t look at a scan and consciously say:
" There' s pancreatic cancer."It' s a mathematical model.
Conceptually:
Millions of image features
&darr
learned patterns
&darr
probability of lesion
&darr
probability of PDAC / other lesion
&darr
location/segmentation
&darr
alert to doctor
The Nature Medicine study says PANDA outputs both a segmentation mask of the pancreas/lesion and classification probabilities.
So the doctor can see where the AI thinks the abnormality is, rather than receiving merely:
" Cancer: YES."
Why this is potentially revolutionary
Here' s the part I think connects directly to your Raffles question.The CT scan already exists.
You don' t necessarily need:
❌ new CT appointment
❌ new radiation exposure
❌ expensive screening programme
❌ specialist looking at every scan again
Instead:
Existing CT + AI
100,000 existing CT scans&darr
AI re-analyses them
&darr
small number of suspicious cases
&darr
human specialist reviews
&darr
additional testing only for those patients.
That' s opportunistic cancer detection.
The research specifically explored routine abdominal and chest non-contrast CTs across physical examination, emergency, outpatient and inpatient settings.
And there is a second-generation opportunity
The current PANDA system is already being developed further.The PANDA Pro version is designed to improve differentiation between pancreatic cancer, pancreatitis and pancreatic cystic lesions and reduce interference from neighbouring structures such as the bile duct and duodenum.
DAMO' s current medical-AI platform also reports pancreatic screening performance of approximately 92.9% sensitivity and 99.9% specificity in its published/marketed evaluation metrics.
But don' t interpret 99.9% specificity as " 99.9% accurate diagnosis." These are different statistical measures, and real-world deployment still requires physician confirmation and further diagnostic testing.
The really big idea for Raffles
Imagine Raffles Hospital puts an AI layer over every CT scanner.Not just pancreas.
The same architecture could potentially screen for:
Pancreas &rarr liver &rarr lung &rarr kidney &rarr colon &rarr cardiovascular abnormalities &rarr other cancers
DAMO itself is already describing a broader multi-cancer direction using non-contrast CT.
Then you add an AI agent:
CT completed.&darr
AI finds pancreatic abnormality.&darr
AI checks previous scans.&darr
" There was a smaller abnormality 18 months ago."&darr
AI alerts gastroenterologist.&darr
AI orders/coordinates appropriate follow-up subject to clinician approval.&darr
Patient receives appointment.&darr
MRI/EUS performed.&darr
Diagnosis.That is no longer just " AI medical imaging."
It becomes an AI-powered early-detection hospital.
And that, in my view, is potentially much more strategically important to Raffles Medical than humanoid robots.
chartiskao ( Date: 04-Sep-2026 09:43) Posted:
|
if AI fails the indians too over AIs
India is already moving toward AI-led hospital transformation, and in some areas it is arguably moving faster than Singapore' s private-hospital sector. But there is a big difference between AI adoption and a genuinely AI-native hospital.
Apollo says it has moved AI into clinical, operational and enterprise functions, rather than treating AI as isolated experiments. Its current platform includes AI disease-risk models, AI clinical decision support, generative-AI clinical summaries, connected wards/ICUs/homes and AI-enabled EMRs.
It has also developed 12 clinical AI algorithms covering conditions including chronic kidney disease, COPD and liver fibrosis, with the models validated against more than 1.7 million patient records.
And Apollo is now partnering with Roche to integrate AI-driven clinical decision support with its EMR + laboratory + hospital information systems.
That is much closer to the model we were discussing for Raffles.
Stage 1 &mdash Digital hospital
EMR &rarr digital records &rarr cloud &rarr telemedicine
&darr
Stage 2 &mdash AI tools
Radiology AI
Risk prediction
Disease detection
Clinical decision support
&darr
Stage 3 &mdash AI operating system
AI agents
Automated documentation
Predictive hospital operations
Connected wards
Continuous monitoring
&darr
Stage 4 &mdash Autonomous hospital
AI agents + computer vision + robotics + digital twin + autonomous logistics
India is clearly at Stage 2&ndash 3.
Some leading hospitals are beginning to approach Stage 4.
That gives it something Raffles cannot easily replicate:
massive data + massive patient volume + multiple hospitals + diagnostics + pharmacy + telemedicine.
This creates a potentially powerful AI feedback loop:
Millions of patients
&rarr clinical data
&rarr AI models
&rarr better diagnosis/workflow
&rarr more patients
&rarr more data
&rarr better models.
Apollo is already developing AI copilots for clinicians, nurses and patients, while its " Ask Apollo" system functions as an AI health assistant across the patient' s healthcare journey.
Reuters reported that Apollo was using AI for things such as:
That is a much more economically meaningful application than a humanoid robot simply greeting patients.
That changes the economics.
Suppose an AI agent can save a doctor:
2 hours/day
For 10,000 doctors:
20,000 hours/day
That is equivalent to an enormous amount of additional healthcare capacity.
So AI in India isn' t merely:
There are:
So India has a paradox:
Excellent AI capability + huge healthcare demand
but
fragmented healthcare infrastructure.
But India may have the better environment for developing AI healthcare at enormous scale.
And Apollo is the company I would watch.
Its new hospitals are increasingly being designed around a digital operating model, rather than taking an old hospital and adding AI afterwards. Apollo' s new 400-bed Hyderabad hospital, for example, integrates its unified EMR/HIS platform with AI-assisted clinical tools and advanced diagnostic, surgical and rehabilitation technology.
That is very close to the " AI-native hospital" concept we were discussing for Raffles.
And there is an even more interesting development: a new AIG Hospitals campus in Visakhapatnam is being built with an AI centre incorporated from the outset, rather than AI being an afterthought.
So yes: India is not waiting for the AI hospital of the future. The leading hospital groups are already building pieces of it. The next competitive leap will be whether they can combine those pieces into one autonomous hospital operating system.
The clearest example: Apollo Hospitals
Apollo Hospitals is probably the Indian example I would watch most closely.Apollo says it has moved AI into clinical, operational and enterprise functions, rather than treating AI as isolated experiments. Its current platform includes AI disease-risk models, AI clinical decision support, generative-AI clinical summaries, connected wards/ICUs/homes and AI-enabled EMRs.
It has also developed 12 clinical AI algorithms covering conditions including chronic kidney disease, COPD and liver fibrosis, with the models validated against more than 1.7 million patient records.
And Apollo is now partnering with Roche to integrate AI-driven clinical decision support with its EMR + laboratory + hospital information systems.
That is much closer to the model we were discussing for Raffles.
India' s progression
I would roughly describe India' s hospital AI evolution like this:Stage 1 &mdash Digital hospital
EMR &rarr digital records &rarr cloud &rarr telemedicine
&darr
Stage 2 &mdash AI tools
Radiology AI
Risk prediction
Disease detection
Clinical decision support
&darr
Stage 3 &mdash AI operating system
AI agents
Automated documentation
Predictive hospital operations
Connected wards
Continuous monitoring
&darr
Stage 4 &mdash Autonomous hospital
AI agents + computer vision + robotics + digital twin + autonomous logistics
India is clearly at Stage 2&ndash 3.
Some leading hospitals are beginning to approach Stage 4.
Apollo is particularly interesting because of scale
Apollo has 74+ hospitals, 6,600+ pharmacies, 2,182 diagnostic centres and 800+ telemedicine centres according to its latest announcement.That gives it something Raffles cannot easily replicate:
massive data + massive patient volume + multiple hospitals + diagnostics + pharmacy + telemedicine.
This creates a potentially powerful AI feedback loop:
Millions of patients
&rarr clinical data
&rarr AI models
&rarr better diagnosis/workflow
&rarr more patients
&rarr more data
&rarr better models.
And India is now moving toward AI agents
This is important relative to your previous Raffles question.Apollo is already developing AI copilots for clinicians, nurses and patients, while its " Ask Apollo" system functions as an AI health assistant across the patient' s healthcare journey.
Reuters reported that Apollo was using AI for things such as:
- medical documentation
- diagnosis/treatment suggestions
- discharge summaries
- nursing scheduling
- antibiotic selection
That is a much more economically meaningful application than a humanoid robot simply greeting patients.
India also has a major advantage
India has a huge healthcare labour shortage and enormous patient population.That changes the economics.
Suppose an AI agent can save a doctor:
2 hours/day
For 10,000 doctors:
20,000 hours/day
That is equivalent to an enormous amount of additional healthcare capacity.
So AI in India isn' t merely:
" Let' s make hospitals technologically impressive."It can become:
" How do we produce more healthcare with the doctors and nurses we already have?"That' s a much stronger economic incentive.
But India has a weakness
India' s healthcare system is extremely fragmented.There are:
- world-class private hospitals
- small private hospitals
- government hospitals
- diagnostic centres
- rural facilities
- different IT systems
- inconsistent digitisation
So India has a paradox:
Excellent AI capability + huge healthcare demand
but
fragmented healthcare infrastructure.
This makes the Raffles vs India comparison interesting
| Raffles Singapore | Leading Indian hospitals | |
|---|---|---|
| Digital infrastructure | ⭐ ⭐ ⭐ ⭐ ⭐ | ⭐ ⭐ ⭐ ⭐ |
| AI clinical adoption | ⭐ ⭐ ⭐ ⭐ | ⭐ ⭐ ⭐ ⭐ ⭐ |
| Patient volume | ⭐ ⭐ | ⭐ ⭐ ⭐ ⭐ ⭐ |
| AI development | ⭐ ⭐ ⭐ | ⭐ ⭐ ⭐ ⭐ ⭐ |
| Data scale | ⭐ ⭐ ⭐ | ⭐ ⭐ ⭐ ⭐ ⭐ |
| Labour-saving incentive | ⭐ ⭐ ⭐ ⭐ | ⭐ ⭐ ⭐ ⭐ ⭐ |
| Robotics potential | ⭐ ⭐ ⭐ ⭐ | ⭐ ⭐ ⭐ ⭐ |
| AI agents | ⭐ ⭐ ⭐ | ⭐ ⭐ ⭐ ⭐ |
| Hospital digital twin | ⭐ ⭐ ⭐ | ⭐ ⭐ ⭐ |
| Scaling AI nationally | ⭐ ⭐ ⭐ ⭐ | ⭐ ⭐ ⭐ ⭐ ⭐ |
The surprising conclusion
Singapore may have the better environment for building a highly controlled, integrated " smart hospital."But India may have the better environment for developing AI healthcare at enormous scale.
And Apollo is the company I would watch.
Its new hospitals are increasingly being designed around a digital operating model, rather than taking an old hospital and adding AI afterwards. Apollo' s new 400-bed Hyderabad hospital, for example, integrates its unified EMR/HIS platform with AI-assisted clinical tools and advanced diagnostic, surgical and rehabilitation technology.
That is very close to the " AI-native hospital" concept we were discussing for Raffles.
And there is an even more interesting development: a new AIG Hospitals campus in Visakhapatnam is being built with an AI centre incorporated from the outset, rather than AI being an afterthought.
So yes: India is not waiting for the AI hospital of the future. The leading hospital groups are already building pieces of it. The next competitive leap will be whether they can combine those pieces into one autonomous hospital operating system.
chartiskao ( Date: 04-Sep-2026 09:40) Posted:
|
https://www.youtube.com/watch?v=X73dGEM_29w
Raffles already has several building blocks: an integrated electronic medical-record system, AI-aided colonoscopy, robotics-assisted surgery, advanced radiology/nuclear medicine, and a digital patient platform.
The next step could be to connect these into an AI operating layer for the entire hospital.
For example:
Patient arrives &rarr AI agent coordinates the whole journey
So Raffles doesn' t necessarily need to invent everything itself.
Raffles already has a real example: GI Genius, an AI system used during colonoscopy to help doctors identify polyps and tumours. Raffles was the first private hospital in Singapore to implement AI in colonoscopy.
The next layer could be:
Radiology AI
AI examines:
The important concept is:
AI doesn' t replace the radiologist &rarr AI becomes the radiologist' s second pair of eyes.
Cameras + computer vision could monitor hospital environments without requiring someone to constantly watch screens.
For example:
That is unnecessarily difficult.
The first applications should be boring but economically valuable.
For example:
That frees nurses and healthcare workers from transportation work.
Imagine:
Doctor orders blood test
&darr
AI agent creates request
&darr
Autonomous robot collects sample
&darr
Robot travels to laboratory
&darr
Laboratory system processes sample
&darr
AI analyses result
&darr
Abnormal result is flagged
&darr
Doctor receives notification
&darr
AI agent schedules follow-up
That' s an end-to-end automated workflow.
The human remains responsible for the medical decision.
Create a digital twin of the physical hospital.
The digital twin would contain a real-time representation of:
The progression could be:
Robot-assisted
&rarr AI-assisted
&rarr AI-guided
&rarr eventually partially autonomous
For example, AI could analyse the patient' s anatomy before surgery and create a 3D model.
The surgeon then uses the robotic system with AI providing:
Instead of looking at:
AI agent + robot + computer vision + digital twin + medical AI
as five separate technologies, Raffles could connect them.
detects the fall.
&darr
AI agent
identifies patient and medical history.
&darr
Robot
travels to the room and provides basic assistance / alerts staff.
&darr
Medical AI
assesses vital signs and clinical information.
&darr
Digital twin
updates the hospital' s real-time patient/location model.
&darr
AI agent
coordinates doctor, nurse, imaging and laboratory.
&darr
Robotics
transports the patient or equipment.
&darr
Doctor
makes the final clinical decision.
That' s an AI-native hospital.
Raffles says its integrated electronic medical-record system allows doctors across its ecosystem to access patient records, while its 2025 annual report describes further digital-system development and progressive onboarding to Singapore' s National Electronic Health Record.
And it already has:
The important distinction is that Raffles doesn' t need to become a robotics manufacturer.
It can become the hospital testbed and operating environment where robotics, AI, computer vision and autonomous systems are integrated.
Their January 2025 partnership explicitly includes sharing quality indicators and patient outcomes, using data analytics, and jointly managing hospitalisation bills.
That creates a possible future loop:
AIA insurance data
&rarr identify high-risk patients
&rarr Raffles preventative intervention
&rarr AI-assisted treatment
&rarr lower complications
&rarr shorter hospital stays
&rarr better outcomes
&rarr lower insurance claims
&rarr AIA can price/manage risk better
&rarr Raffles gets more patients
&rarr more clinical data
&rarr better AI
That is potentially much more valuable economically than simply putting a humanoid robot in the hospital lobby.
2. Medical AI &mdash ⭐ ⭐ ⭐ ⭐ ⭐
3. Computer vision &mdash ⭐ ⭐ ⭐ ⭐
4. Autonomous logistics robots &mdash ⭐ ⭐ ⭐ ⭐
5. Digital twin &mdash ⭐ ⭐ ⭐ ⭐
6. Humanoid robots &mdash ⭐ ⭐
The humanoid is the visible part.
The real transformation is the AI agent + hospital data + computer vision + robotics + digital twin underneath it.
And Singapore' s healthcare policy direction is already explicitly moving toward AI-led transformation, while emphasising that AI should improve what clinicians and care teams can actually accomplish rather than being deployed simply for its own sake.
The next step could be to connect these into an AI operating layer for the entire hospital.
1. AI agents &mdash the biggest near-term opportunity
Think of an AI agent as a digital hospital employee, rather than a chatbot.For example:
Patient arrives &rarr AI agent coordinates the whole journey
Appointment &rarr registration &rarr medical records &rarr doctor &rarr imaging &rarr laboratory &rarr pharmacy &rarr billing &rarr discharge &rarr follow-up.The agent could:
- retrieve relevant medical history
- prepare a doctor briefing before consultation
- identify missing test results
- coordinate appointments
- monitor whether tests have been completed
- prepare discharge instructions
- arrange follow-up
- remind patients about medication
- flag abnormal results to clinicians
So Raffles doesn' t necessarily need to invent everything itself.
2. Medical AI
This is where Raffles can directly improve clinical productivity.Raffles already has a real example: GI Genius, an AI system used during colonoscopy to help doctors identify polyps and tumours. Raffles was the first private hospital in Singapore to implement AI in colonoscopy.
The next layer could be:
Radiology AI
AI examines:
- CT
- MRI
- X-ray
- ultrasound
- PET
- mammography
The important concept is:
AI doesn' t replace the radiologist &rarr AI becomes the radiologist' s second pair of eyes.
3. Computer vision
This could be extremely valuable inside a hospital.7
For example:
Patient safety
AI could detect:- patient attempting to get out of bed
- falls
- unusual movement
- patient wandering
- overcrowding
- unsafe behaviour
Operating theatre
Computer vision could potentially recognise:- surgical instruments
- procedural stages
- missing instruments
- sterile-field breaches
- workflow deviations
4. Humanoid robots
I would not start with humanoid robots doing surgery.That is unnecessarily difficult.
The first applications should be boring but economically valuable.
For example:
Robot porter
A robot moves:- medicines
- laboratory samples
- documents
- linen
- meals
- equipment
That frees nurses and healthcare workers from transportation work.
Reception robot
A humanoid/embodied AI could:- greet patients
- provide directions
- translate languages
- explain registration
- escort patients
- answer routine questions
5. Autonomous systems
This is where the hospital starts behaving like an autonomous logistics network.Imagine:
Doctor orders blood test
&darr
AI agent creates request
&darr
Autonomous robot collects sample
&darr
Robot travels to laboratory
&darr
Laboratory system processes sample
&darr
AI analyses result
&darr
Abnormal result is flagged
&darr
Doctor receives notification
&darr
AI agent schedules follow-up
That' s an end-to-end automated workflow.
The human remains responsible for the medical decision.
6. Digital twin of Raffles Hospital
This is probably the most interesting long-term concept.Create a digital twin of the physical hospital.
6
- beds
- operating theatres
- emergency department
- imaging equipment
- doctors
- nurses
- patients
- robots
- ambulances
- laboratories
- pharmacies
- appointment schedules
" What happens if Emergency Department arrivals increase 30% tonight?"Or:
" What happens if MRI machine #2 is unavailable?"Or:
" Where should we deploy another nurse?"Or:
" Can we perform another 10 surgeries per week without expanding the building?"That' s much more powerful than simply putting AI into individual departments.
7. AI + robotics-assisted surgery
Raffles already has robotics-assisted surgery and a Robotics Centre its annual-report materials describe robotic technology being used to assist surgeons with minimally invasive surgery.The progression could be:
Robot-assisted
&rarr AI-assisted
&rarr AI-guided
&rarr eventually partially autonomous
For example, AI could analyse the patient' s anatomy before surgery and create a 3D model.
The surgeon then uses the robotic system with AI providing:
- anatomical identification
- optimal trajectory suggestions
- tissue recognition
- movement constraints
- real-time warnings
- surgical documentation
8. The really powerful model: combine everything
This is where I think your idea becomes interesting.Instead of looking at:
AI agent + robot + computer vision + digital twin + medical AI
as five separate technologies, Raffles could connect them.
Example: elderly patient falls
Computer visiondetects the fall.
&darr
AI agent
identifies patient and medical history.
&darr
Robot
travels to the room and provides basic assistance / alerts staff.
&darr
Medical AI
assesses vital signs and clinical information.
&darr
Digital twin
updates the hospital' s real-time patient/location model.
&darr
AI agent
coordinates doctor, nurse, imaging and laboratory.
&darr
Robotics
transports the patient or equipment.
&darr
Doctor
makes the final clinical decision.
That' s an AI-native hospital.
9. Why Raffles could actually be well positioned
There are several foundations already in place.Raffles says its integrated electronic medical-record system allows doctors across its ecosystem to access patient records, while its 2025 annual report describes further digital-system development and progressive onboarding to Singapore' s National Electronic Health Record.
And it already has:
| Technology | Raffles starting point | Next stage |
|---|---|---|
| Medical AI | AI colonoscopy | AI across imaging/diagnostics |
| Robotics | Robotic surgery | Autonomous hospital logistics |
| Digital records | Integrated EMR | AI clinical agents |
| Patient app | Raffles Connect | Personal healthcare AI agent |
| Computer vision | Potential new layer | Patient safety + workflow |
| Digital twin | Opportunity | Whole-hospital simulation |
| AI agents | Emerging opportunity | Autonomous administrative workflow |
It can become the hospital testbed and operating environment where robotics, AI, computer vision and autonomous systems are integrated.
And this is where AIA becomes interesting
If by your earlier question you meant AIA Singapore + Raffles Hospital, rather than " AI" , there is potentially an even bigger strategic connection.Their January 2025 partnership explicitly includes sharing quality indicators and patient outcomes, using data analytics, and jointly managing hospitalisation bills.
That creates a possible future loop:
AIA insurance data
&rarr identify high-risk patients
&rarr Raffles preventative intervention
&rarr AI-assisted treatment
&rarr lower complications
&rarr shorter hospital stays
&rarr better outcomes
&rarr lower insurance claims
&rarr AIA can price/manage risk better
&rarr Raffles gets more patients
&rarr more clinical data
&rarr better AI
That is potentially much more valuable economically than simply putting a humanoid robot in the hospital lobby.
My ranking for Raffles
1. AI agents &mdash ⭐ ⭐ ⭐ ⭐ ⭐2. Medical AI &mdash ⭐ ⭐ ⭐ ⭐ ⭐
3. Computer vision &mdash ⭐ ⭐ ⭐ ⭐
4. Autonomous logistics robots &mdash ⭐ ⭐ ⭐ ⭐
5. Digital twin &mdash ⭐ ⭐ ⭐ ⭐
6. Humanoid robots &mdash ⭐ ⭐
The humanoid is the visible part.
The real transformation is the AI agent + hospital data + computer vision + robotics + digital twin underneath it.
And Singapore' s healthcare policy direction is already explicitly moving toward AI-led transformation, while emphasising that AI should improve what clinicians and care teams can actually accomplish rather than being deployed simply for its own sake.
chartiskao ( Date: 04-Sep-2026 07:05) Posted:
|