Prognostic value of machine learning for acute heart failure

Funding Acknowledgements

Type of funding sources: None.

Background

At onset of acute heart failure (AHF), various clinical fundamental parameters including vital sign, laboratory data, or initial treatment were investigated, and we can roughly estimate the prognosis. However, machine learning method for prediction of the prognosis was not studied.

Purpose

To elucidate prognostic value of machine learning for AHF comparing conventional statistical model.

Methods

We enrolled consecutive 300 patients with AHF (79.5 ± 12.1 years, 158 Males). Patients with acute coronary syndrome, mechanical circulatory support cases, and cardio-pulmonary arrest cases were excluded. The patients were randomly divided into 80% (240 cases) and 20% (60 cases), and the former was used as train data, and the latter as validation data. Objective variable was set as cardiac death in one year. First, logistic regression analysis with Akaike’s information criterion (AIC) was performed, and extracted predictive parameters. The predictive model for the cardiac prognosis was constructed by cut-off value of ROC curve analysis of propensity score was calculated. Next, machine learning (random forest method and deep learning) to build predictive model was performed with the predictors. Finally, accuracy of each predictive model was compared.

Results

Thirty cases showed cardiac death in one year. Logistic regression with AIC extracted 8  predictors, and the cut off-value of propensity score with the 6 parameters was 0.110. The accuracy was 0.714 and area under ROC (AUROC) was 0.836. Conversely, random forest method demonstrated the accuracy as 0.927, AUROC 0.860. On deep learning, the accuracy was 0.937 and AUROC 0.901. 

The top 4 high feature importance of random forest were Cl/red blood cell count/pH/Anion Gap. However, accuracy of those predictors was lower than that of machine learning.

Conclusion

Machine learning was a powerful tool to predict cardiac prognosis of AHF, comparing with conventional statistical model. Abstract Figure. Statistical model

Duchenne Muscular Dystrophy Successfully Treated with Aripiprazole in a Patient with Autism Spectrum Disorder Symptoms Including Irritability

Duchenne muscular dystrophy (DMD) is associated with neuropsychiatric disorders, and patients often present with autism spectrum disorder (ASD). We herein report a case of DMD accompanied by ASD that was successfully treated with aripiprazole, an atypical antipsychotic that has been used for treating irritability in child and early adolescent patients with ASD. The patient was diagnosed as having DMD at 3 years of age. Although he developed severe psychotic symptoms including irritability, insomnia, hallucinations, and delusions at 17 years of age, all the symptoms were successfully treated with aripiprazole without any detectable side effects.

ET-9 Development of photosensitive antibodies for near-infrared light immunotherapy targeting EGFR and IL13Rα2 of malignant gliomas and investigation of their photodynamic cytotoxic activity in vitro

Introduction: Near-Infrared Photoimmunotherapy (NIR-PIT) is a recently developed hybrid cancer therapy based on photodynamic cytotoxicity and anti-tumor immunopotentiation, utilizing a photosensitive antibody drug (PSAD). A global Phase III trial of NIR-PIT with an anti-EGFR-PSAD in patients with recurrent head and neck squamous cell carcinoma (HNSCC) is already underway, and NIR-PIT is expected to have therapeutic applications also in malignant gliomas. Methods: In this study, monoclonal antibodies to EGFR and IL13Rα2 were conjugated to the photosensitive dye IRDye700DX (IR700) to produce PSADs (EGFR-Ab/IR700 and IL13Rα2-Ab/IR700) and in vitro PDT assays using these PSADs were performed on four human glioma cell lines (U87MG, U251, U138, A172).Five groups were studied: EGFR-Ab/IR700 monotherapy: 5 μg/ml or 10 μg/ml, IL13Rα2-Ab/IR700 monotherapy: 5 μg/ml or 10 μg/ml, and EGFR-Ab/IR700: 5 μg/ml + IL13Rα2-Ab/IR700: 5 μg/ml combination therapy. The cytotoxic activity of each group was compared after irradiation with 690 nm light at 16 J/cm2. Results: Significantly higher cytotoxic activity was observed in all four glioma cell lines when EGFR-Ab/IR700 and IL13Rα2-Ab/IR700 were used in combination at 5 μg/ml each, than when each PSAD was treated with a doubled dose (10 μg/ml).Conclusion: Malignant gliomas show extensive cellular heterogeneity with diverse expression of cell surface antigens. The present results suggest that a therapeutic strategy using several different photosensitive antibodies simultaneously may lead to the release of tumor antigens from a greater number of tumor cells, resulting in a more efficient host immune response for therapeutic purposes.

Diagnostic performance of deep learning on 12-leads electrocardiography for recurrence after pulmonary vein isolation in patients with persistent atrial fibrillation

Background

Several patients with persistent atrial fibrillation (per-AF) suffer from recurrence after pulmonary vein isolation (PVI). Various methods to predict the recurrence were tried, but deep learning on 12-leads electrocardiography (ECG) after PVI was not studied.

Purpose

To elucidate diagnostic performance of deep learning on 12-leads ECG after PVI in patients with per-AF

Methods

We enrolled consecutive 109 patients with per-AF who underwent PVI (68.8±10.0 years, 83 males) excluding failure cases. We defined recurrence in 3–12 months after PVI. From the ECG just after PVI, five beats of each lead were sampled separately. Deep learning (convolutional neural network on bitmap ECG image) was performed by transfer learning of Inception-Resnet-V2 model. Gradient weighted class activation color mapping (GradCam) was performed to detect convolutional importance in the lead.

Results

Thirty-six patients showed recurrence in the period. Lead II (accuracy 0.701), aVR (0.690) were the top 2 leads of prediction, which showed larger accuracy than statistical accuracies of Non PV foci = SVC (accuracy = 0.541) and left atrial diameter >50mm (0.596). In lead II, GradCam spotlighted strong convolution of latter half of P wave in recurrent case, and former half of P wave and T wave in no-recurrent case.

Conclusions

Deep learning on ECG was a powerful tool to predict recurrence of per-AF after PVI.

Funding Acknowledgement

Type of funding sources: None. Results of deep learningResults of GradCam

A case of sporadic late-onset nemaline myopathy with monoclonal gammopathy of undetermined significance: long-term observation of neurological symptoms after autologous stem-cell transplantation

A 47-year-old woman presented with progressive limb weakness. A neurological examination revealed proximal dominant symmetrical muscle weakness in her limbs, and electromyography revealed complex repetitive discharges and short motor unit potentials with positive sharp waves in the biceps. We observed early recruitment in the quadriceps, and laboratory tests revealed normal creatine kinase. Serum protein electrophoresis showed monoclonal IgG-lambda, but the bone marrow aspiration specimen was normal. A muscle biopsy revealed nemaline rod accumulations in the muscle fibers; based on the results, we diagnosed the patient with sporadic late-onset nemaline myopathy with monoclonal gammopathy of undetermined significance (SLONM-MGUS). We administered repeated intravenous immunoglobulin, but her limb weakness continued, and she developed a restrictive ventilatory defect. The patient received melphalan, followed by autologous stem-cell transplantation (ASCT). Her upper extremity strength and respiratory capability improved within one year after ASCT; however, it was not until six years after ASCT that her atrophied lower extremities strengthened. A discrepancy in the timeline of treatment response between the upper or respiratory muscles and the atrophied lower limb was characteristic in the patient, suggesting that the efficacy of ASCT on SLONM-MGUS should be evaluated in the long term, especially in severely atrophied muscles. In addition, this case showed that ASCT for SLOMN-MGUS is an effective treatment option in Asian populations.

E_{8} Spectra of Quasi-One-Dimensional Antiferromagnet BaCo_{2}V_{2}O_{8} under Transverse Field

We report ^{51}V NMR and inelastic neutron scattering (INS) measurements on a quasi-1D antiferromagnet BaCo_{2}V_{2}O_{8} under transverse field along the [010] direction. The scaling behavior of the spin-lattice relaxation rate above the Néel temperatures unveils a 1D quantum critical point (QCP) at H_{c}^{1D}≈4.7  T, which is masked by the 3D magnetic order. With the aid of accurate analytical analysis and numerical calculations, we show that the zone center INS spectrum at H_{c}^{1D} is precisely described by the pattern of the 1D quantum Ising model in a magnetic field, a class of universality described in terms of the exceptional E_{8} Lie algebra. These excitations are nondiffusive over a certain field range when the system is away from the 1D QCP. Our results provide an unambiguous experimental realization of the massive E_{8} phase in the compound, and open a new experimental route for exploring the dynamics of quantum integrable systems as well as physics beyond integrability.

Intra-Aneurysmal Coil Embolization of a Ruptured Distal Posterior Inferior Temporal Artery Aneurysm: A Case Report

Objective

We report a case of a ruptured aneurysm at the posterior inferior temporal artery (PITA) of the posterior cerebral artery (PCA) treated by intra-aneurysmal coil embolization.

Case Presentation

A 93-year-old man presented with disturbance of consciousness. Angiography revealed a 3-mm aneurysm in the distal PITA of the left PCA. He was diagnosed with subarachnoid hemorrhage and intracerebral hemorrhage due to a ruptured aneurysm. This aneurysm was occluded by intra-aneurysmal coil embolization with preservation of the PITA.

Conclusion

Distal PITA aneurysm of the PCA is rare. Complete occlusion and preservation of the parent artery were achieved by intra-aneurysmal coil embolization, which may be an effective therapeutic option for such aneurysms.

Influence of rotational speed on torque/force generation and shaping ability during root canal instrumentation of extracted teeth with continuous rotation and optimum torque reverse motion

AIM

To evaluate how different rotational speeds affect the torque/force generation and shaping ability of rotary root canal instrumentation using JIZAI (MANI, Utsunomiya, Japan) nickel-titanium instruments in continuous rotation and optimum torque reverse (OTR) motion.

METHODOLOGY

Mesial root canals of extracted mandibular molars were instrumented up to size 25, 0.04 taper using JIZAI instruments, and anatomically matched canals were selected based on geometric features of the canal [canal volume (mm³ ), surface area (mm² ), length, 15°-20° curvature and radius of curvature (4-8 mm)] after micro-computed tomographic scanning. An automated root canal instrumentation and torque/force analysing device was programmed to permit a simulated pecking motion (2 s downward and 1 s upward at 50 mm min^-1 ). The selected canals were prepared with size 25, 0.06 taper JIZAI instruments using continuous rotation or OTR motion and further subdivided according to the rotational speed (300 or 500 rpm, n = 10 each). Real-time clockwise/counterclockwise torque and downward/upward force were recorded using a custom-made torque/force analysing device. Then, the registered pre- and post-operative micro-computed tomographic datasets were examined to evaluate the canal volume changes and centring ratios at 1, 3, 5 and 7 mm from the apical foramen. Data were analysed using two-way analysis of variance or the Kruskal-Wallis test and Mann-Whitney U test with Bonferroni correction (α = 5%).

RESULTS

Maximum upward force and clockwise torque were significantly smaller in 500 rpm groups than in 300 rpm groups (P < .05); however, no significant difference was found between continuous rotation and OTR motion (P > .05). OTR motion developed higher maximum counterclockwise torque than continuous rotation (P < .05). Maximum downward force, canal volume changes and centring ratios were not significantly different among all groups (P > .05). There was no file fracture in any of the groups.

CONCLUSIONS

Under laboratory conditions using JIZAI instruments, a rotational speed of 500 rpm generated significantly lower maximum screw-in forces and torque values than rotational speed of 300 rpm. Continuous rotation and OTR motion performed similarly in shaping the canals.

In vitro evidence of propagation of superoxide dismutase-1 protein aggregation in canine degenerative myelopathy

Canine degenerative myelopathy (DM) is a progressive and fatal neurodegenerative disorder that has been linked to mutations in the superoxide dismutase 1 (SOD1) gene. The accumulation of misfolded protein aggregates in spinal neurons and astrocytes is implicated as an important pathological process in DM; however, the mechanism of protein aggregate formation is largely unknown. In human neurodegenerative diseases, such as amyotrophic lateral sclerosis (ALS), cell-to-cell propagation of disease-relevant proteins has been demonstrated. Therefore, in this study, propagation of aggregation-forming property of mutant SOD1 protein in DM in vitro was investigated. This study demonstrated that aggregates composed of canine wild type SOD1 protein were increased by co-transfection with canine mutant SOD1 (E40K SOD1), indicating intracellular propagation of SOD1 aggregates. Further, aggregated recombinant SOD1 proteins were released from the cells, taken up by other cells, and induced further aggregate formation of normally folded SOD1 proteins. These results suggest intercellular propagation of SOD1 aggregates. The hypothesis of cell-to-cell propagation of SOD1 aggregates proposed in this study may underly the progressive nature of DM pathology.

A case of Infective Native (Abdominal) Aortic Aneurysm Caused by Streptococcus Agalactiae: An Updated Literature Review Based on New Nomenclature

The management of abdominal aortic aneurysms (AAA) has evolved significantly with the advent of endovascular strategies. Thus, there has been a decline in the number of open AAA repairs once an endovascular option is available. There have also been reports of successful endovascular management of infective native aortic aneurysms (INAA)¹, previously called mycotic aneurysms². The rarity of this condition makes its management a challenging one as there are no standard guidelines. The European Society of Vascular Surgery has suggested that the nomenclature be changed from mycotic aneurysms as this can be misleading to standardise reporting¹. The authors' present a case of a 67-year old male who presented during the peak of the Corona Virus pandemic with constitutional gastrointestinal symptoms. He was subsequently diagnosed with an INAA and successfully managed with open Neo-Aorto Iliac System reconstruction with a homograft³. The report highlights various strategies used in the surgical approach and their benefits in the management of INAA. Furthermore, a literature review of Streptococcus (Streptococcus agalactiae) species as a rare cause of INAA and how these cases were managed are also highlighted.

Clinical Result of Mechanical Thrombectomy Using Sofia Plus with Acute Ischemic Stroke Compared with the Stent Retriever

OBJECTIVE

The Sofia catheter is a new large-bore aspiration catheter that allows easy access and good reperfusion. In this study, we analyzed the efficacy and safety of the Sofia catheter in comparison with stent retrievers as a contact aspiration thrombectomy (CAT) tool for large vessel occlusion.

METHODS

We enrolled patients with acute ischemic stroke who underwent endovascular thrombectomy from April 2017 and April 2020 in our hospital. Patients were retrospectively reviewed and divided into the stent retriever group (SR), the Sofia group, and all cases group.

RESULTS

A total of 114 patients were treated during the study period, including 27 in the stent retriever group and 15 in the Sofia group. Higher rates of the first-pass effect (FPE) (37% vs. 47%, P = 0.12) and significantly higher modified FPE (44% vs. 67%, P = 0.001) were observed in patients with SR and Sofia, respectively. Functional independence (modified Rankin Scale ≤2) at 30 days after onset was observed in 30% versus 47% (P = 0.47) of SR and Sofia, respectively.

CONCLUSIONS

The Sofia Plus is a large-bore aspiration catheter with high FPE and good accessibility. Use of this catheter resulted better angiographic outcome compared with the stent retriever, but there was no difference in clinical outcomes in this study. Further studies are needed to compare in new generation aspiration catheters.

GALA-Modified Lipid Nanoparticles for the Targeted Delivery of Plasmid DNA to the Lungs

This study describes the development of lipid nanoparticles (LNPs) for the efficient and selective delivery of plasmid DNA (pDNA) to the lungs. The GALA peptide was used as a ligand to target the lung endothelium and as an endosomal escape device. Transfection activity in the lungs was significantly improved when pDNA was encapsulated in double-coated LNPs. The inner coat was composed of dioleoylphsophoethanolamine and a stearylated octaarginine (STR-R8) peptide, while the outer coat was largely a cationic lipid, di-octadecenyl-trimethylammonium propane, mixed with YSK05, a pH-sensitive lipid, and cholesterol. Optimized amounts of YSK05 and GALA were used to achieve an efficient and lung-selective system. The optimized system produced a high gene expression level in the lungs (>10⁷ RLU/mg protein) with high lung/liver and lung/spleen ratios. GALA/R8 modification and the double-coating design were indispensable for efficient gene expression in the lungs. Despite the fact that NPs prepared with 1-step or 2-step coating have the same lipid amount and composition and the same pDNA dose, the transfection activity was dramatically higher in the lungs in the case of 2-step coating. Surprisingly, 1-step or 2-step coatings had no effect on the amount of nanoparticles that were delivered to the lungs, suggesting that the double-coating strategy substantially improved the efficiency of gene expression at the intracellular level.

Novel lipid combination for delivery of plasmid DNA to immune cells in the spleen

This study reports on the development of a novel lipid combination that permits the efficient and highly selective delivery of plasmid DNA (pDNA) to immune cells in the spleen. Using DODAP, an ionizable lipid that was previously thought to be inefficient for gene delivery, we show for the first time, that this ignored lipid can be successfully used for efficient and targeted gene delivery in vivo, but only when combined with DOPE, a specific helper lipid. Using certain DODAP and DOPE ratios resulted in the formation of lipid nanoparticles (LNPs) with a ~ 1000-fold higher gene expression, and this expression was specific for the spleen, making it the most spleen-selective system for transfection using pDNA. The developed DODAP/DOPE-LNPs target immune cells in the spleen via receptors for complement C3 and this pathway is critical for efficient gene expression. We hypothesize that the high spleen transfection activity of DODAP/DOPE-LNPs is caused by the promotion of gene expression associated with B cell activation via complement receptors. LNPs encapsulating tumor-antigen encoding pDNA showed both prophylactic and therapeutic anti-tumor effects. The optimized LNPs resulted in the production of different cytokines and antigen-specific antibodies as well as exerting antigen-specific cytotoxic effects. This study revives the use of DODAP in gene delivery and highlights the importance of using appropriate lipid combinations for delivering genes to specific cells.

Lipid Nanoparticles for Cell-Specific in Vivo Targeted Delivery of Nucleic Acids

The last few years have witnessed a great advance in the development of nonviral systems for in vivo targeted delivery of nucleic acids. Lipid nanoparticles (LNPs) are the most promising carriers for producing clinically approved products in the future. Compared with other systems used for nonviral gene delivery, LNPs provide several advantages including higher stability, low toxicity, and greater efficiency. Additionally, systems based on LNPs can be modified with ligands and devices for controlled biodistribution and internalization into specific cells. Efforts are ongoing to improve the efficiency of lipid-based gene vectors. These efforts depend on the appropriate design of nanocarriers as well as the development of new lipids with improved gene delivery ability. Several ionizable lipids have recently been developed and have shown dramatically improved efficiency. However, enhancing the ability of nanocarriers to target specific cells in the body remains the most difficult challenge. Systemically administered LNPs can access organs in which the capillaries are characterized by the presence of fenestrations, such as the liver and spleen. The liver has received the most attention to date, although targeted delivery to the spleen has recently emerged as a promising tool for modulating the immune system. In this review, we discuss recent advances in the use of LNPs for cell-specific targeted delivery of nucleic acids. We focus mainly on targeting liver hepatocytes and spleen immune cells as excellent targets for gene therapy. We also discuss the potential of endothelial cells as an alternate approach for targeting organs with a continuous endothelium.

STMO-07 Adverse events related to the glioblastoma surgery and their perioperative management in our hospital

Background: The standard treatment for glioblastoma is surgical resection following chemoradiation therapy. The rate of removal or the amount of residual tumor has some impact on the prognosis of patients with glioblastoma, but the highly invasive nature of this tumor makes complete removal limited to the contrast-enhanced lesions difficult due to its localization. Furthermore, when postoperative seizures and venous thrombosis are included in surgery-related complications, these perioperative adverse events can cause delays in the initiation of chemoradiotherapy and delay the return to work and home, such as prolonged hospitalization and rehabilitation time.

Methods: We retrospectively reviewed the perioperative status of the recent 50 consecutive cases with histologically confirmed as glioblastoma at our hospital, the patient background, tumor localization, and perioperative treatment, and so on. Results: The major perioperative complications were ischemic or hemorrhagic complications, epileptic seizures, venous thrombosis, and pneumonia; CTCAE grade 2 or higher, grade 3 or higher, and grade 4 occurred in about 40%, 20%, and 10%, respectively, with some patients having multiple complications. Discussion: Although there was a tendency for ischemic changes around the cavity of the resection as the resection rate increased, most cases were asymptomatic and it seemed to be acceptable if residual brain function could be preserved. Residual tumors tended to show hemorrhagic changes and epileptic seizures because this is thought to be that the tumor was deliberately left in place to preserve function, based on the localization of the tumor. Postoperative FDP levels were useful in predicting the development of deep vein thrombosis and pulmonary artery thromboembolism. Conclusion: Because glioblastoma has short survival time and patient PS before and after surgery varies greatly depending on tumor localization, it is important to consider risk-benefit strategies for each case and to establish a scheme for a seamless transition from perioperative management to the introduction of postoperative therapy and maintenance therapy.

The wide-ranging clinical and genetic features in Japanese families with valosin-containing protein proteinopathy

Mutations in the valosin-containing protein (VCP) gene are known to cause various neurodegenerative disorders. Here, we report 8 Japanese patients [6 men, 2 women; median age at onset: 49.5 (range, 35-58) years] from 5 unrelated families with VCP missense mutations. Although 7 of 8 patients were diagnosed with either inclusion body myopathy or amyotrophic lateral sclerosis, 1 patient showed demyelinating polyneuropathy, which was confirmed by longitudinal nerve conduction studies. Sural nerve biopsy of the patient revealed intranuclear ubiquitin staining in Schwann cells. Three known pathogenic VCP mutations (p.Arg191Gln, p.Arg155Cys, and p.Ile126Phe) were detected. A novel mutation, c.293 A>T (p.Asp98Val), was also identified in a patient with amyotrophic lateral sclerosis and frontotemporal dementia. This mutation was predicted to be "deleterious" or "disease causing" using in silico mutation analyses. In conclusion, demyelinating polyneuropathy may be a novel phenotype caused by VCP mutations. The p.Asp98Val mutation was found to be a novel pathogenic mutation of VCP proteinopathy. We believe our cases represent a wide clinical spectrum of VCP mutations.

Optic Nerve Root Enhancement in Gadolinium-Enhanced Magnetic Resonance Imaging of Rathke's Cleft Cyst

BACKGROUND

Although most patients with Rathke's cleft cysts (RCCs) remain asymptomatic throughout their lives, visual impairment in RCCs usually indicates surgical interventions, including endoscopic drainage of the cyst. We report a case of RCC with unique findings in the optic nerve root (ONR) by magnetic resonance imaging (MRI).

CASE DESCRIPTION

A 58-year-old man admitted to our hospital complained of progressive left visual impairment. Preoperative computed tomography MRI revealed a suprasellar cystic lesion that extended anteriorly, and the bilateral ONRs were lateralized with the cyst and were partially enhanced with gadolinium administration. The cyst content was drained, and the cyst wall was partially removed by endoscopic transsphenoidal surgery. Postoperatively, the patient was administered corticosteroid intravenously for 3 days, and his visual acuity improved dramatically. Postoperative MRI revealed that the volume of the cyst decreased, and the position of the bilateral ONRs normalized. Additionally, enhancement of the ONRs with gadolinium administration was not observed. Histological examination of the surgical specimen revealed a ciliated single-layer columnar epithelium with inflammatory cell infiltration.

CONCLUSIONS

To our knowledge, we report the first case of a patient with RCC with ONR enhancement with gadolinium administration on MRI. This unique finding might prove that inflammation is one of the causes of visual impairment in RCC as in optic neuritis.

Prediction for cardiac prognosis in patients with congestive heart failure by machine learning on dual-isotope myocardial semiconductor SPECT

Background

Dual-isotope (low doze 201TlCl and 123I-β-methyl-P-iodophenyl-pentadecanoic acid (BMIPP)) single photon emission computed tomography (SPECT) is utilized to estimate myocardial damage in patients with congestive heart failure (CHF). However, predictive model construction on the SPECT for cardiac death by machine learning was not studied.

Purpose

To elucidate predictive value of machine learning model on dual-isotope SPECT for CHF.

Methods

We enrolled consecutive 310 patients who admitted with CHF (77.1±3.1 years, 164 males). After initial treatment, they underwent electrocardiography gated SPECT and observed in median 507 days [IQR: 165, 1032]. Multivariate Cox regression analysis for cardiac death was performed, and predictive model was constructed by ROC curve analysis and machine learning (Random Forest and Deep Learning). The accuracies (= [True positive + True negative] / Total) of the prediction models were compared with ROC curve model.

Results

Thirty-six patients fell into cardiac death. Cox analysis showed Age, left ventricular ejection fraction (LVEF), summed rest score (SRS) of BMIPP, and mismatch score were significant predictors (Hazard ratio: 1.068, 0.970, 1.032, 1.092, P value: &lt;0.001, 0.014, 0.002, &lt;0.001, respectively). ROC curve analysis of them revealed the accuracy of the cut-off value was 0.479–0.773. Conversely, machine learning model demonstrated higher accuracy for cardiac death (Random Forest: 0.895, Deep Learning: 0.935). The top 4 feature importance of the random forest were LVEF (0.299), SRS BMIPP (0.263), Age (0.262), and mismatch score (0.160).

Conclusion

Machine learning model on SPECT was superior to conventional statistic model for predicting cardiac death in patients with CHF.

Funding Acknowledgement

Type of funding source: None