© Aldeca Productions – stock.adobe.com
Article • Developments in brain imaging for psychiatry
Radiologists explore new frontiers of the mind
Psychiatrists and neuroimaging experts gathered to explore one of the last remaining frontiers in radiology – the human mind – at the annual meeting of the French Society of Radiology (JFR) last October in Paris. Their conversation oscillated between neurons and narratives, algorithms and emotions – a vivid reminder that psychiatry is not only about understanding the brain, but about listening to the invisible.
By Mélisande Rouger
The experts’ message was bold and humble: psychiatry is entering a new era, powered by artificial intelligence (AI) and brain imaging. But amid these disruptive changes, one truth remains constant – the patient must stay at the centre.
From invisible suffering to visible patterns
For decades, psychiatry has relied on words. Diagnosis has been rooted in interviews, intuition, and empathy – an encounter between two minds, one seeking to name what the other feels.
But as Edouard Duchesnay, researcher at Neurospin in Saclay, explained, mental illness also leaves subtle, measurable marks in the brain’s structure – almost invisible variations in grey matter, connectivity, or cortical development. ‘A patient’s state of mind originates in the brain,’ he said. ‘So we can image it, and perhaps one day, predict its evolution.’
Imagine a fourteen-year-old who struggles silently. If brain imaging could show whether that adolescent’s neural development follows a trajectory toward resilience or psychosis, clinicians could act earlier, adapt care, and perhaps prevent suffering before it surfaces
Edouard Duchesnay
His team develops algorithms capable of identifying differences of just two or three percent in brain tissue or white matter pathways. These imperceptible irregularities, when analyzed across large cohorts, can reveal patterns of vulnerability or resilience.
The goal is not to replace psychiatry’s human gaze with cold mathematics, but to give it new tools: to understand who might be at risk, who might recover, and why. ‘Nothing predicts the future,’ Duchesnay said, ‘but data can help us see trajectories that words alone can’t describe.’
These models – built on thousands of MRI scans – can already anticipate the onset of psychosis or predict how a bipolar patient will respond to lithium years in advance. It’s not magic, Duchesnay insists, but mathematics in the service of care.
The promise and the paradox of prediction
The ambition is clear: to move from reactive to predictive psychiatry – a discipline capable of detecting vulnerability before crisis. ‘Imagine,’ Duchesnay said, ‘a fourteen-year-old who struggles silently. If brain imaging could show whether that adolescent’s neural development follows a trajectory toward resilience or psychosis, clinicians could act earlier, adapt care, and perhaps prevent suffering before it surfaces.’
Yet caution prevails. For him, AI in psychiatry is not clairvoyance – it’s pattern recognition, grounded in biology but guided by humility. ‘The danger,’ he warned, ‘is to believe that numbers know more than people.’ Each dataset, no matter how vast, is only a reflection – a statistical echo of human complexity. That’s why, even in the heart of the algorithm, the art of interpretation remains indispensable.
Images give us an entry point, but they don’t replace the story the patient tells. The goal is to align both – what the scan reveals, and what the person lives
David Attali
Cachia, a professor of cognitive neuroscience at Paris Cité University, emphasized how MRI helps decode the architecture of thought. ‘We’re not looking for illness,’ he said, ‘we’re looking for balance – how the brain’s structure supports or disrupts it.’
Tomas Mastellari, head of the psychiatry clinic at Lille University Hospital, brought a clinician’s sensitivity to the discussion: ‘A brain scan can’t feel suffering,’ he said, ‘but it can show us where suffering leaves its trace.’
For Marion Plaze, a psychiatrist at Sainte-Anne Hospital in Paris, imaging is reshaping how medicine defines disorders such as depression or schizophrenia. These are no longer fixed categories, but dynamic states within a spectrum of neural and emotional activity. AI, she suggested, could refine these spectrums – allowing treatment to match each individual’s biology and rhythm.
David Attali, head of the psychiatry clinic at Paris University Hospital Group Psychiatry & Neuroscience, insisted that technology’s real power lies in connecting data with dialogue. ‘Images give us an entry point,’ he said, ‘but they don’t replace the story the patient tells. The goal is to align both – what the scan reveals, and what the person lives.’
Psychiatry in transition: a new language of the brain
While Duchesnay’s work represents the technical frontier, the next challenge lies in translation – turning data into understanding, and understanding into care. That was the focus of a roundtable gathering four leading psychiatrists: Arnaud Cachia, Tomas Mastellari, Marion Plaze, and David Attali. Together, they explored how brain imaging can illuminate psychiatry without overshadowing it.
Between data and empathy
Across all voices, one conviction stood firm: psychiatry must remain a human science. AI may classify patterns, but it can neither grasp paradoxes nor read irony, guilt, or love.
Cachia warned of the temptation to turn psychiatry into a science of patterns rather than of people. ‘If we see only the image and not the individual,’ he said, ‘we risk losing what psychiatry was built upon – the encounter.’
Their dialogue, full of humor and empathy, bridged the worlds of imaging and imagination.
The psychiatrists spoke of their patients not as data points, but as complex stories in motion – each MRI slice a fragment of experience, each voxel a memory, each anomaly a question rather than an answer.
This blend of precision and compassion marked the tone of the entire session: an invitation to use AI not as a substitute for intuition, but as a mirror that sharpens it.
The mind, illuminated
In the end, both the researchers’ algorithms and the psychiatrists’ reflections converged toward a shared horizon – that of an integrated psychiatry, where technology supports empathy and science deepens understanding.
AI, when used wisely, becomes not a machine of prediction, but an instrument of attention.
The brain, once a black box, is slowly opening to science. What it reveals is not only the anatomy of disease, but the resilience of those who live within it, they concluded.
Profile:
Edouard Duchesnay is Affiliate Professor of machine learning at Université Evry Paris-Saclay in France. He is the leader of the “Signatures of brain disorders” team at NeuroSpin (CEA, Université Paris-Saclay), where he supervises the design of machine learning and statistical models to uncover neural signatures predictive of clinical trajectories in psychiatric disorders.
14.03.2026
