Artificial Intelligence Predictive Model for Hormone Therapy Use in Prostate Cancer.

Background: Androgen deprivation therapy (ADT) with radiotherapy can benefit patients with localized prostate cancer. However, ADT can negatively impact quality of life and there remain no validated predictive models to guide its use. Methods: Digital pathology image and clinical data from pre-treatment prostate tissue from 5,727 patients enrolled on five phase III randomized trials treated with radiotherapy +/- ADT were used to develop and validate an artificial intelligence (AI)-derived predictive model to assess ADT benefit with the primary endpoint of distant metastasis. After the model was locked, validation was performed on NRG/RTOG 9408 (n = 1,594) that randomized men to radiotherapy +/- 4 months of ADT. Fine-Gray regression and restricted mean survival times were used to assess the interaction between treatment and predictive model and within predictive model positive and negative subgroup treatment effects. Results: In the NRG/RTOG 9408 validation cohort (14.9 years of median follow-up), ADT significantly improved time to distant metastasis (subdistribution hazard ratio [sHR] = 0.64, 95%CI [0.45-0.90], p = 0.01). The predictive model-treatment interaction was significant (p-interaction = 0.01). In predictive model positive patients (n = 543, 34%), ADT significantly reduced the risk of distant metastasis compared to radiotherapy alone (sHR = 0.34, 95%CI [0.19-0.63], p < 0.001). There were no significant differences between treatment arms in the predictive model negative subgroup (n = 1,051, 66%; sHR = 0.92, 95%CI [0.59-1.43], p = 0.71). Conclusions: Our data, derived and validated from completed randomized phase III trials, show that an AI-based predictive model was able to identify prostate cancer patients, with predominately intermediate-risk disease, who are likely to benefit from short-term ADT.

Artificial Intelligence Predictive Model for Hormone Therapy Use in Prostate Cancer.

Predictive Model for Hormone Therapy in Prostate CancerDigital pathology images and clinical data from pretreatment prostate tissue were used to generate a predictive model to determine patients who would benefit from androgen deprivation therapy (ADT). In model-positive patients, ADT significantly reduced the risk of distant metastasis compared with radiotherapy alone.

Detection of signs of disease in external photographs of the eyes via deep learning.

Retinal fundus photographs can be used to detect a range of retinal conditions. Here we show that deep-learning models trained instead on external photographs of the eyes can be used to detect diabetic retinopathy (DR), diabetic macular oedema and poor blood glucose control. We developed the models using eye photographs from 145,832 patients with diabetes from 301 DR screening sites and evaluated the models on four tasks and four validation datasets with a total of 48,644 patients from 198 additional screening sites. For all four tasks, the predictive performance of the deep-learning models was significantly higher than the performance of logistic regression models using self-reported demographic and medical history data, and the predictions generalized to patients with dilated pupils, to patients from a different DR screening programme and to a general eye care programme that included diabetics and non-diabetics. We also explored the use of the deep-learning models for the detection of elevated lipid levels. The utility of external eye photographs for the diagnosis and management of diseases should be further validated with images from different cameras and patient populations.

Task-specific employment of sensory signals underlies rapid task switching.

Much of our flexible behavior is dependent on responding efficiently to relevant information while discarding irrelevant information. Little is known, however, about how neural pathways governing sensory-motor associations can rapidly switch to accomplish such flexibility. Here, we addressed this question by electrically microstimulating middle temporal (MT) neurons selective for both motion direction and binocular disparity in monkeys switching between direction and depth discrimination tasks. Surprisingly, we frequently found that the observed psychophysical bias precipitated by delivering microstimulation to neurons whose preferred direction and depth were related to opposite choices in the two tasks was substantially shifted toward a specific movement. Furthermore, these effects correlated with behavioral switching performance. Our findings suggest that the outputs of sensory signals are task specific and that irrelevant sensory-motor pathways are gated depending on task demand so as to accomplish rapid attentional switching.

Predicting the risk of developing diabetic retinopathy using deep learning.

BACKGROUND: Diabetic retinopathy screening is instrumental to preventing blindness, but scaling up screening is challenging because of the increasing number of patients with all forms of diabetes. We aimed to create a deep-learning system to predict the risk of patients with diabetes developing diabetic retinopathy within 2 years. METHODS: We created and validated two versions of a deep-learning system to predict the development of diabetic retinopathy in patients with diabetes who had had teleretinal diabetic retinopathy screening in a primary care setting. The input for the two versions was either a set of three-field or one-field colour fundus photographs. Of the 575 431 eyes in the development set 28 899 had known outcomes, with the remaining 546 532 eyes used to augment the training process via multitask learning. Validation was done on one eye (selected at random) per patient from two datasets: an internal validation (from EyePACS, a teleretinal screening service in the USA) set of 3678 eyes with known outcomes and an external validation (from Thailand) set of 2345 eyes with known outcomes. FINDINGS: The three-field deep-learning system had an area under the receiver operating characteristic curve (AUC) of 0·79 (95% CI 0·77-0·81) in the internal validation set. Assessment of the external validation set-which contained only one-field colour fundus photographs-with the one-field deep-learning system gave an AUC of 0·70 (0·67-0·74). In the internal validation set, the AUC of available risk factors was 0·72 (0·68-0·76), which improved to 0·81 (0·77-0·84) after combining the deep-learning system with these risk factors (p<0·0001). In the external validation set, the corresponding AUC improved from 0·62 (0·58-0·66) to 0·71 (0·68-0·75; p<0·0001) following the addition of the deep-learning system to available risk factors. INTERPRETATION: The deep-learning systems predicted diabetic retinopathy development using colour fundus photographs, and the systems were independent of and more informative than available risk factors. Such a risk stratification tool might help to optimise screening intervals to reduce costs while improving vision-related outcomes. FUNDING: Google.

Author Correction: Detection of anaemia from retinal fundus images via deep learning.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Detection of anaemia from retinal fundus images via deep learning.

Owing to the invasiveness of diagnostic tests for anaemia and the costs associated with screening for it, the condition is often undetected. Here, we show that anaemia can be detected via machine-learning algorithms trained using retinal fundus images, study participant metadata (including race or ethnicity, age, sex and blood pressure) or the combination of both data types (images and study participant metadata). In a validation dataset of 11,388 study participants from the UK Biobank, the fundus-image-only, metadata-only and combined models predicted haemoglobin concentration (in g dl-1) with mean absolute error values of 0.73 (95% confidence interval: 0.72-0.74), 0.67 (0.66-0.68) and 0.63 (0.62-0.64), respectively, and with areas under the receiver operating characteristic curve (AUC) values of 0.74 (0.71-0.76), 0.87 (0.85-0.89) and 0.88 (0.86-0.89), respectively. For 539 study participants with self-reported diabetes, the combined model predicted haemoglobin concentration with a mean absolute error of 0.73 (0.68-0.78) and anaemia an AUC of 0.89 (0.85-0.93). Automated anaemia screening on the basis of fundus images could particularly aid patients with diabetes undergoing regular retinal imaging and for whom anaemia can increase morbidity and mortality risks.

Disengagement of motor cortex from movement control during long-term learning.

Motor learning involves reorganization of the primary motor cortex (M1). However, it remains unclear how the involvement of M1 in movement control changes during long-term learning. To address this, we trained mice in a forelimb-based motor task over months and performed optogenetic inactivation and two-photon calcium imaging in M1 during the long-term training. We found that M1 inactivation impaired the forelimb movements in the early and middle stages, but not in the late stage, indicating that the movements that initially required M1 became independent of M1. As previously shown, M1 population activity became more consistent across trials from the early to middle stage while task performance rapidly improved. However, from the middle to late stage, M1 population activity became again variable despite consistent expert behaviors. This later decline in activity consistency suggests dissociation between M1 and movements. These findings suggest that long-term motor learning can disengage M1 from movement control.

Real-Time Processing of Two-Photon Calcium Imaging Data Including Lateral Motion Artifact Correction.

Two-photon calcium imaging has been extensively used to record neural activity in the brain. It has been long used solely with post-hoc analysis, but the recent efforts began to include closed-loop experiments. Closed-loop experiments pose new challenges because they require fast, real-time image processing without iterative parameter tuning. When imaging awake animals, one of the crucial steps of post hoc image analysis is correction of lateral motion artifacts. In most of the closed-loop experiments, this step has not been implemented and ignored due to technical difficulties. We recently reported the first experiments with real-time processing of calcium imaging that included lateral motion correction. Here, we report the details of the implementation of fast motion correction and present performance analysis across several algorithms with different parameters. Additionally, we introduce a novel method to estimate baseline calcium signal using kernel density estimate, which reduces the number of parameters to be tuned. Combined, we propose a novel software pipeline of real-time image processing suited for closed-loop experiments. The pipeline is also useful for rapid post hoc image processing.

Brain-Computer Interface with Inhibitory Neurons Reveals Subtype-Specific Strategies.

Brain-computer interfaces have seen an increase in popularity due to their potential for direct neuroprosthetic applications for amputees and disabled individuals. Supporting this promise, animals-including humans-can learn even arbitrary mapping between the activity of cortical neurons and movement of prosthetic devices [1-4]. However, the performance of neuroprosthetic device control has been nowhere near that of limb control in healthy individuals, presenting a dire need to improve the performance. One potential limitation is the fact that previous work has not distinguished diverse cell types in the neocortex, even though different cell types possess distinct functions in cortical computations [5-7] and likely distinct capacities to control brain-computer interfaces. Here, we made a first step in addressing this issue by tracking the plastic changes of three major types of cortical inhibitory neurons (INs) during a neuron-pair operant conditioning task using two-photon imaging of IN subtypes expressing GCaMP6f. Mice were rewarded when the activity of the positive target neuron (N+) exceeded that of the negative target neuron (N-) beyond a set threshold. Mice improved performance with all subtypes, but the strategies were subtype specific. When parvalbumin (PV)-expressing INs were targeted, the activity of N- decreased. However, targeting of somatostatin (SOM)- and vasoactive intestinal peptide (VIP)-expressing INs led to an increase of the N+ activity. These results demonstrate that INs can be individually modulated in a subtype-specific manner and highlight the versatility of neural circuits in adapting to new demands by using cell-type-specific strategies.

Whisker row deprivation affects the flow of sensory information through rat barrel cortex.

Whisker trimming causes substantial reorganization of neuronal response properties in barrel cortex. However, little is known about experience-dependent rerouting of sensory processing following sensory deprivation. To address this, we performed in vivo intracellular recordings from layers 2/3 (L2/3), layer 4 (L4), layer 5 regular-spiking (L5RS), and L5 intrinsically bursting (L5IB) neurons and measured their multiwhisker receptive field at the level of spiking activity, membrane potential, and synaptic conductance before and after sensory deprivation. We used Chernoff information to quantify the "sensory information" contained in the firing patterns of cells in response to spared and deprived whisker stimulation. In the control condition, information for flanking-row and same-row whiskers decreased in the order L4, L2/3, L5IB, L5RS. However, after whisker-row deprivation, spared flanking-row whisker information was reordered to L4, L5RS, L5IB, L2/3. Sensory information from the trimmed whiskers was reduced and delayed in L2/3 and L5IB neurons, whereas sensory information from spared whiskers was increased and advanced in L4 and L5RS neurons. Sensory information from spared whiskers was increased in L5IB neurons without a latency change. L5RS cells exhibited the largest changes in sensory information content through an atypical plasticity combining a significant decrease in spontaneous activity and an increase in a short-latency excitatory conductance. NEW & NOTEWORTHY: Sensory cortical plasticity is usually quantified by changes in evoked firing rate. In this study we quantified plasticity by changes in sensory detection performance using Chernoff information and receiver operating characteristic analysis. We found that whisker deprivation causes a change in information flow within the cortical layers and that layer 5 regular-spiking cells, despite showing only a small potentiation of short-latency input, show the greatest increase in information content for the spared input partly by decreasing their spontaneous activity.

A leaky-integrator model as a control mechanism underlying flexible decision making during task switching.

The ability to switch between tasks is critical for animals to behave according to context. Although the association between the prefrontal cortex and task switching has been well documented, the ultimate modulation of sensory-motor associations has yet to be determined. Here, we modeled the results of a previous study showing that task switching can be accomplished by communication from distinct populations of sensory neurons. We proposed a leaky-integrator model where relevant and irrelevant information were stored separately in two integrators and task switching was achieved by leaking information from the irrelevant integrator. The model successfully explained both the behavioral and neuronal data. Additionally, the leaky-integrator model showed better performance than an alternative model, where irrelevant information was discarded by decreasing the weight on irrelevant information, when animals initially failed to commit to a task. Overall, we propose that flexible switching is, in part, achieved by actively controlling the amount of leak of relevant and irrelevant information.

Apolipoprotein E, especially apolipoprotein E4, increases the oligomerization of amyloid ß peptide.

Alzheimer's disease (AD) is the most common progressive neurodegenerative disorder causing dementia. Massive deposition of amyloid ß peptide (Aß) as senile plaques in the brain is the pathological hallmark of AD, but oligomeric, soluble forms of Aß have been implicated as the synaptotoxic component. The apolipoprotein E ε 4 (apoE ε4) allele is known to be a genetic risk factor for developing AD. However, it is still unknown how apoE impacts the process of Aß oligomerization. Here, we found that the level of Aß oligomers in APOE ε4/ε4 AD patient brains is 2.7 times higher than those in APOE ε3/ε3 AD patient brains, matched for total plaque burden, suggesting that apoE4 impacts the metabolism of Aß oligomers. To test this hypothesis, we examined the effect of apoE on Aß oligomer formation. Using both synthetic Aß and a split-luciferase method for monitoring Aß oligomers, we observed that apoE increased the level of Aß oligomers in an isoform-dependent manner (E2 < E3 < E4). This effect appears to be dependent on the ApoE C-terminal domain. Moreover, these results were confirmed using endogenous apoE isolated from the TBS-soluble fraction of human brain, which increased the formation of Aß oligomers. Together, these data show that lipidated apoE, especially apoE4, increases Aß oligomers in the brain. Higher levels of Aß oligomers in the brains of APOE ε4/ε4 carriers compared with APOE ε3/ε3 carriers may increase the loss of dendritic spines and accelerate memory impairments, leading to earlier cognitive decline in AD.