Better insights with ISC when using a multi-sensory fMRI paradigm

SfN, Nov 6, 2018

1Electrical Engin., 2Computer Sci. & Engin., Indian Inst. of Technology, Delhi, New Delhi, India; 3All India Inst. of Med. Sci., Delhi, India; 4Mahajan Imaging Pvt. Ltd., Delhi, India

Cue reactivity tasks have been widely employed in fMRI studies. Due to ease of use and compatibility with General Linear Model (GLM), visual cues are predominantly adopted despite their limitation in terms of replicating real life scenario. We propose using Intersubject Correlation Analysis (ISC) to analyse multi sensory paradigms over GLM based analysis and demonstrate advantages of ISC in a multi sensory paradigm using a case study of craving for alcohol in subjects with heavy alcohol use.

Four male young adults (mean age of 24) with heavy alcohol use whose score on Alcohol Use Disorder Identification Test (AUDIT) was greater than 8, were scanned using a 3T GE MRI Scanner while undergoing a multi sensory craving paradigm. The paradigm included 20 blocks with short videos with fixation cross after every block. Ten videos contained alcohol which were matched with neutral videos based on colour, background, presence of faces, emotions, etc. The order of blocks was randomized once and then kept same across all subjects.

Preprocessing of fMRI data included BET extraction, slice timing correction (ascending interleaved), spatial smoothing (FWHM of 5mm) and temporal filtering of 0.01Hz using FSL. Contrast between alcohol cues and fixation was computed using GLM analysis and compared with statistical maps obtained using ISC analysis. Both the statistical maps were corrected for multiple comparisons using False Discovery Rate (FDR) of 0.05.

With GLM analysis, both visual and auditory regions were observed to be activated along with thalamus. With ISC analysis, regions previously known to be involved in craving such as insula, amygdala, hippocampus, caudate, putamen, anterior cingulate cortex (ACC), posterior cingulate cortex (PCC), orbitofrontal cortex (OFC) were also activated. Refer to the attached figure for the two statistical maps and the activated areas.

We hypothesize that craving is nonlinear in nature. Linear Time Invariant (LTI) assumption of GLM makes it harder to capture craving regions when applied to multi-sensory cues. ISC analysis is a better option in this case.

Read more here

Deep learning algorithms for detection of critical findings in head CT scans: a retrospective study

The Lancet; Published online October 11, 2018

Sasank Chilamkurthy, Rohit Ghosh, Swetha Tanamala, Mustafa Biviji, Norbert G Campeau, Vasantha Kumar Venugopal, Vidur Mahajan, Pooja Rao, Prashant Warier


Non-contrast head CT scan is the current standard for initial imaging of patients with head trauma or stroke symptoms. We aimed to develop and validate a set of deep learning algorithms for automated detection of the following key findings from these scans: intracranial haemorrhage and its types (ie, intraparenchymal, intraventricular, subdural, extradural, and subarachnoid); calvarial fractures; midline shift; and mass effect.


We retrospectively collected a dataset containing 313 318 head CT scans together with their clinical reports from around 20 centres in India between Jan 1, 2011, and June 1, 2017. A randomly selected part of this dataset (Qure25k dataset) was used for validation and the rest was used to develop algorithms. An additional validation dataset (CQ500 dataset) was collected in two batches from centres that were different from those used for the development and Qure25k datasets. We excluded postoperative scans and scans of patients younger than 7 years. The original clinical radiology report and consensus of three independent radiologists were considered as gold standard for the Qure25k and CQ500 datasets, respectively. Areas under the receiver operating characteristic curves (AUCs) were primarily used to assess the algorithms.


The Qure25k dataset contained 21 095 scans (mean age 43 years; 9030 [43%] female patients), and the CQ500 dataset consisted of 214 scans in the first batch (mean age 43 years; 94 [44%] female patients) and 277 scans in the second batch (mean age 52 years; 84 [30%] female patients). On the Qure25k dataset, the algorithms achieved an AUC of 0·92 (95% CI 0·91–0·93) for detecting intracranial haemorrhage (0·90 [0·89–0·91] for intraparenchymal, 0·96 [0·94–0·97] for intraventricular, 0·92 [0·90–0·93] for subdural, 0·93 [0·91–0·95] for extradural, and 0·90 [0·89–0·92] for subarachnoid). On the CQ500 dataset, AUC was 0·94 (0·92–0·97) for intracranial haemorrhage (0·95 [0·93–0·98], 0·93 [0·87–1·00], 0·95 [0·91–0·99], 0·97 [0·91–1·00], and 0·96 [0·92–0·99], respectively). AUCs on the Qure25k dataset were 0·92 (0·91–0·94) for calvarial fractures, 0·93 (0·91–0·94) for midline shift, and 0·86 (0·85–0·87) for mass effect, while AUCs on the CQ500 dataset were 0·96 (0·92–1·00), 0·97 (0·94–1·00), and 0·92 (0·89–0·95), respectively.


Our results show that deep learning algorithms can accurately identify head CT scan abnormalities requiring urgent attention, opening up the possibility to use these algorithms to automate the triage process.

Read more here